JP3819237B2 - Tracking device - Google Patents

Description

【０００７】
上記（１）式において、ｏは観測値を示す添字で、ｋはサンプリング時刻を示す添字で、ｊは探知データ数を示す添字である。つまり、探知データＤ_kjはサンプリング時刻ｋにおけるｊ番目の探知データを示している。また、ｘ_okj は探知データＤ_kjのＸ座標成分を示し、同様にｙ_okj は探知データＤ_kjのＹ座標成分、ｚ_okj は探知データＤ_kjのＺ座標成分をそれぞれ示している。さらに、記号「’」はベクトル及び行列の転置を示す。
【０００８】
各Ｓ／Ｎ比残差共分散行列算出手段４は、観測手段１から得られる探知データＤ_kjと探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを入力し、各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出して出力する。ここで、各Ｓ／Ｎ比残差共分散行列をＳ_kj（ｔ）の添字ｋはサンプリング時刻を示し、添字ｊはｊ番目の探知データＤ_kjからのものであることを示し、（ｔ）は航跡番号を示している。
【０００９】
各Ｓ／Ｎ比残差共分散行列算出手段４は、各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を、カルマンフィルタの理論に基づき次の（２）式より算出する。
【数２】

上記（２）式において、Ｈは観測行列、Ｐ_pk（ｔ）は後述の予測手段７から遅延手段８を介して得られる航跡番号（ｔ）の予測誤差共分散行列、Ｒ_kj（ｔ）は北基準直交座標における航跡番号（ｔ）の観測誤差共分散行列を示している。
【００１０】
上記（２）式における観測行列Ｈは、次の（３）式のように示される。
【数３】

【００１１】
また、上記（２）式における北基準直交座標における観測誤差共分散行列Ｒ_kj（ｔ）は、次の（４）式で示される。
【数４】

上記（４）式において、Λ_kjは極座標における観測誤差共分散行列であり、Γ_kj（ｔ）は、航跡番号（ｔ）のΛ_kjの変換行列である。
【００１２】
上記（４）式に示される、極座標における観測誤差共分散行列Λ_kjは、次の（５）式で示される。
【数５】

【００１３】
上記（５）式において、σｒ_kjは距離の観測雑音の標準偏差、σｅ_kjは仰角の観測雑音の標準偏差、σｂ_kjは方位角の観測雑音の標準偏差を示す。距離観測雑音は平均０、分散σｒ² _kjの正規分布に従う。同様にして、仰角観測雑音は平均０、分散σｅ² _kjの正規分布に従い、方位角観測雑音は平均０、分散σｂ² _kjの正規分布に従う。
【００１４】
上記（４）式における、航跡番号（ｔ）のΛ_kjの変換行列Γ_kj（ｔ）は、次の（６）式で示される。
【数６】

上記（６）式におけるｒ_pk（ｔ），ｅ_pk（ｔ），ｂ_pk（ｔ）は、それぞれ、後述の予測手段７から遅延手段８を介して得られる予測ベクトルＶＸ_pk（ｔ）の位置成分ｘ_pk（ｔ），ｙ_pk（ｔ），ｚ_pk（ｔ）の極座標成分を示したものである。
【００１５】
また、ｒ_pk（ｔ），ｅ_pk（ｔ），ｂ_pk（ｔ）と、ｘ_pk（ｔ），ｙ_pk（ｔ），ｚ_pk（ｔ）の関係は、次の（７）式で示される。
【数７】

【００１６】
上記（５）式における距離の観測雑音の標準偏差σｒ_kj、仰角の観測雑音の標準偏差σｅ_kj、方位角の観測雑音の標準偏差σｂ_kjは、Ｓ／Ｎ比ＳＮＲ_kjを用いて、次の（８）式、（９）式、（１０）式のように示すことができる。
【数８】

上記（８）式、（９）式、（１０）式におけるα_r ，α_e ，α_b は、レーダの送信信号の周波数、レーダの送信信号のバンド幅、レーダアンテナの有効開口径、及び光速から決定される正の定数である。
【００１７】
精ゲート内外判定手段５で使用する精ゲートは次の（１１）式で示される領域である。
【数９】

上記（１１）式において、ＶＺ_pk（ｔ）は予測ベクトルＶＸ_pk（ｔ）の位置成分を示し、（１１）式における上添字の「−１」は、行列の逆行列を示す記号である。
【００１８】
さらに、（１１）式におけるｄεは精ゲートの大きさを決めるゲートサイズパラメータである。このゲートサイズパラメータｄεは定数であり、ゲートサイズパラメータｄεの添字εは、精ゲート内に目標が存在する確率を示す。この精ゲート内に目標が存在する確率を、ゲート内確率、又はゲート内目標存在期待確率という。ここで、ゲートサイズパラメータｄεは、ゲート内確率εのときのゲートの大きさを示し、統計学でいうカイ２乗分布により、自由度と自分が設定したいゲート内確率εを事前に決め、カイ２乗密度関数、又はカイ２乗分布表より求める。
【００１９】
また、予測ベクトルＶＸ_pk（ｔ）は、次の（１２）式のように示される。
【数１０】

（１２）式において、ｘ_pk（ｔ），ｙ_pk（ｔ），ｚ_pk（ｔ）は、Ｘ，Ｙ，Ｚの予測位置、ｄｘ_pk（ｔ），ｄｙ_pk（ｔ），ｄｚ_pk（ｔ）は、Ｘ，Ｙ，Ｚの予測位置における予測速度である。
【００２０】
上記（１１）式における予測ベクトルＶＸ_pk（ｔ）の位置成分ＶＺ_pk（ｔ）と、上記（１２）式の予測ベクトルＶＸ_pk（ｔ）は、上記（３）式の観測行列Ｈを用いて、次の（１３）式の関係がある。
【数１１】

【００２１】
上記（１１）式で示される精ゲートの大きさは、上記ゲートサイズパラメータｄεと各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）により決まる。次に、何故、ゲートサイズパラメータｄεと残差共分散行列Ｓ_kj（ｔ）により精ゲートの大きさが決まるかを、例を用いて説明する。以下、説明を簡便化するため、次の（１４）式、（１５）式のように、行列Ａ及びベクトルｘを仮定する。そうすると、上記（１１）式は、（１４）式、（１５）式より、次の（１６）式のように決まる。
【００２２】
【数１２】

【００２３】
ここで、上記（１４）式における行列Ａは３行３列の正値対称行列なので、次の（１７）式、（１８）式が成立する。
【数１３】

【００２４】
ここで、行列Ａを次の（１９）式のように、代数学による対角行列Ｂ、直交行列Ｃで対角化できたとする。
【数１４】

【００２５】
上記（１９）式における対角行列Ｂは、次の（２０）式のように、行列Ａの固有値λ₁ ，λ₂ ，λ₃ で示されるとする。ここで、上記（１８）式の関係より、固有値λ₁ ，λ₂ ，λ₃ は、各々、λ₁ ＞０，λ₂ ＞０，λ₃ ＞０である。
【数１５】

【００２６】
上記（１９）式における直交行列Ｂは、行列Ａの固有ベクトルｖ₁ ，ｖ₂ ，ｖ₃ を用いて、次の（２１）式のように示される。ここで、ｖ₁ ，ｖ₂ ，ｖ₃は各々３行１列のベクトルである。
【数１６】

【００２７】
上記（２１）式における行列Ａの固有ベクトルｖ₁ ，ｖ₂ ，ｖ₃ には、次の（２２）式の関係がある。
【数１７】

【００２８】
また、上記（２０）式は、上記（１９）式を用いて、次の（２３）式のように示される。
【数１８】

【００２９】
ここで、上記（２３）式におけるＣｘの成分を、次の（２４）式のようにした場合には、上記（２３）式は、上記（２０）式、次の（２４）式を用いて、次の（２５）式のように示される。
【数１９】

【００３０】
ここで、上記（１４）式、（２０）式より、行列Ａの固有値はλ₁ ，λ₂ ，λ₃ なので、各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）の固有値は１／λ₁ ，１／λ₂ ，１／λ₃ である。よって、上記（２５）式は、誤差楕円の半径が、各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）の固有値１／λ₁ ，１／λ₂ ，１／λ₃ ，及びゲートサイズパラメータｄεにより決まることを示している。つまり、精ゲートの大きさは、各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）とゲートサイズパラメータｄεにより決まる。そして、各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）が大きいほど、また、各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）の逆行列が小さいほど、精ゲートの大きさは大きくなる。さらに、ゲートサイズパラメータｄεが大きいほど、精ゲートの大きさは大きくなる。
【００３１】
また、上記（１１）式の予測ベクトルＶＸ_pk（ｔ）の位置成分ＶＺ_pk（ｔ）は精ゲートの中心を示し、（１１）式の右辺は、探知データＤ_kjと精ゲートの中心である予測ベクトルＶＸ_pk（ｔ）の位置成分ＶＺ_pk（ｔ）の各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）で正規化した距離を示している。
【００３２】
したがって、精ゲート内外判定手段５では、探知データＤ_kjと精ゲートの中心である予測ベクトルＶＸ_pk（ｔ）の位置成分ＶＺ_pk（ｔ）の各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）で正規化した距離が、ゲートサイズパラメータｄεより小さい場合、つまり、上記（１１）式を満たす探知データＤ_kjを目標信号の候補ａＤ_kjとして出力する。さらに、その目標信号の候補をａＤ_kjに対応する各Ｓ／Ｎ比残差共分散行列をａＳ_kj（ｔ）として出力する。さらに、予測ベクトＶＸ_pk（ｔ）ルの位置成分ＶＺ_pk（ｔ）を出力する。
【００３３】
ここで、探知データＤ_kjは観測手段１から得られ、精ゲートの中心である予測ベクトルＶＸ_pk（ｔ）の位置成分ＶＺ_pk（ｔ）は、後述の予測手段７の出力である現サンプリング時刻ｋに対する１サンプリング後の予測ベクトルＶＸ_pk+1（ｔ）を、後述の遅延手段８を介して現時刻のＶＸ_pk（ｔ）とし後、上記（１３）式を用いて、予測ベクトルＶＸ_pk（ｔ）の位置成分ＶＺ_pk（ｔ）を算出する。各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）は、各Ｓ／Ｎ比残差共分散行列算出手段４より得られる。
【００３４】
データ更新手段６は、カルマンフィルタの理論に基づき、ゲイン行列Ｋ_k （ｔ）を次の（２６）式から算出する。また、平滑ベクトルＶＸ_sk（ｔ）を（２６）式のゲイン行列Ｋ_k （ｔ）を用いて、次の（２７）式により算出する。さらに、平滑誤差共分散行列Ｐ_sk（ｔ）を（２６）式のゲイン行列Ｋ_k （ｔ）を用いて、次の（２８）式により算出する。そして、データ更新手段６は、（２７）式、（２８）式により算出された平滑ベクトルＶＸ_sk（ｔ）及び平滑誤差共分散行列Ｐ_sk（ｔ）を出力する。
【００３５】
【数２０】

【００３６】
予測手段７は、カルマンフィルタの理論に基づき、次の（２９）式により、現サンプリング時刻ｋより１サンプリング後の時刻ｋ＋１における予測ベクトルＶＸ_pk+1（ｔ）を算出して出力する。また、予測手段７は、次の（３０）式により、現サンプリング時刻ｋより１サンプリング後の時刻ｋ＋１における予測誤差共分散行列Ｐ_pk+1（ｔ）を算出し出力する。
【数２１】

【００３７】
上記（２９）式において、１サンプリング外挿を行う状態遷移行列Φ_k は、次の（３１）式で示される。
【数２２】

上記（３１）式におけるΔｔは、現時刻ｋと現時刻から１サンプリング後の時刻ｋ＋１までのサンプリング間隔を示す。
【００３８】
また、上記（３０）式における駆動雑音の共分散行列Ｑ_k は、次の（３２）式で示される。
【数２３】

上記（３２）式におけるσ_a は駆動雑音の標準偏差を示し、カルマンフィルタの目標の運動モデルの曖昧さを示す定数である。
【００３９】
遅延手段８では、予測手段７より入力された、予測ベクトルＶＸ_pk+1（ｔ）及び予測誤差共分散行列Ｐ_pk+1（ｔ）を１サンプリング分遅延させて、現時刻の予測ベクトルＶＸ_pk（ｔ）及び現時刻の予測誤差共分散行列Ｐ_pk（ｔ）を各Ｓ／Ｎ比残差共分散行列手段４に出力する。
【００４０】
上記のような追尾装置では、全ての探知データＤ_kjに関して、上記（１１）式における各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）の逆行列計算を行うことになる。例えば、現時刻より１サンプリング前の航跡の数をＬとし、現時刻における探知データＤ_kjの数をＭとすると、航跡数がＬになることは、航跡番号（ｔ）がｔ＝１，・・・，Ｌとなるということであり、また探知データＤ_kjの数（探知データ数ｊ）がＭということは、探知データＤ_kjにおける探知データＤ_kjの数を示す添字ｊがｊ＝１，・・・，Ｍになるということである。
【００４１】
そうすると、上記（１１）式を用いて、これらの全ての組み合わせで、精ゲートの処理を行うとＬＭ個の逆行列計算が必要になり、計算機の演算負荷が膨大となる。精ゲート内外判定手段５における精ゲートの意味は、上記全ての組み合わせで精ゲートの処理を行うことを意味する。
【００４２】
ここで、航跡とは各時刻から高々１個の探知データＤ_kjを選んで構成した時系列データである。航跡の種類は、既存航跡、新航跡の２通りの航跡がある。既存航跡は前時刻までに作成された航跡であり、新航跡は現時刻のある探知データＤ_kjより新たに開始される航跡である。
【００４３】
【発明が解決しようとする課題】
従来の追尾装置は以上のように構成されているので、全ての探知データＤ_kjにに関して、上記（１１）式における各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）の逆行列計算を行う場合に、現時刻より１サンプリング前の航跡数をＬとし、現時刻における探知データＤ_kjの数をＭとすると、（１１）式を用いて、これらの全ての組み合わせで精ゲートの処理を行うには、ＬＭ個の逆行列計算が必要になり、計算機の演算負荷が膨大となるという課題があった。
【００４４】
この発明は上記のような課題を解決するためになされたもので、目標信号にとり不要な信号を除去するための精ゲートの演算において、演算負荷をかけずに目標信号を精ゲートの中に捕獲できる追尾装置を得ることを目的とする。
【００４５】
【課題を解決するための手段】
この発明に係る追尾装置は、目標を探知するためのしきい値であるＳ／Ｎ比最小値を用いて、内部に備えているセンサから探知した探知データ及び探知データに付随するＳ／Ｎ比を出力する観測手段と、上記観測手段の出力を入力し、上記Ｓ／Ｎ比最小値を用いて計算した探知データ数に依存しない現時刻の観測誤差共分散行列Ｒ_kmin（ｔ）により、探知データ数に依存しない残差共分散行列Ｓ_kmin（ｔ）を算出し、算出した残差共分散行列Ｓ_kmin（ｔ）、並びに入力した探知データ及びＳ／Ｎ比を出力するＳ／Ｎ比最小値残差共分散行列算出手段と、上記Ｓ／Ｎ比最小値残差共分散行列算出手段から入力した残差共分散行列Ｓ_kmin（ｔ）により、目標存在期待領域である粗ゲートを算出し、入力した探知データ及びＳ／Ｎ比について、粗ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、判定した粗ゲート内に入っている探知データ及びＳ／Ｎ比を出力する粗ゲート内外判定手段と、上記粗ゲート内外判定手段の出力を用いて、サンプリング時刻、探知データ数、航跡数に依存する、粗ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出して、算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）、並びに入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比を出力する各Ｓ／Ｎ比残差共分散行列算出手段と、上記各Ｓ／Ｎ比残差共分散行列算出手段の出力を用いて、粗ゲートより狭い目標存在期待領域である精ゲートを算出し、入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比について、精ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、サンプリング時刻、探知データ数、航跡数に依存する、精ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出し、判定した精ゲート内に入っている探知データ及びＳ／Ｎ比、並びに算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を出力する精ゲート内外判定手段とを備えたものである。
【００４６】
この発明に係る追尾装置は、目標を探知するためのしきい値であるＳ／Ｎ比最小値を用いて、内部に備えているセンサから探知した探知データ及び探知データに付随するＳ／Ｎ比を出力する観測手段と、上記観測手段の出力と、後述の遅延手段が出力した現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを入力し、上記Ｓ／Ｎ比最小値を用いて計算した探知データ数に依存しない現時刻の観測誤差共分散行列Ｒ_kmin（ｔ）と、入力した現時刻の予測誤差共分散行列とを用いて、探知データ数に依存しない残差共分散行列Ｓ_kmin（ｔ）を算出し、算出した残差共分散行列Ｓ_kmin（ｔ）、並びに入力した探知データ、Ｓ／Ｎ比、現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力するＳ／Ｎ比最小値残差共分散行列算出手段と、上記Ｓ／Ｎ比最小値残差共分散行列算出手段から入力した残差共分散行列Ｓ_kmin（ｔ）及び現時刻の予測ベクトルにより、目標存在期待領域である粗ゲートを算出し、入力した探知データ及びＳ／Ｎ比について、粗ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、判定した粗ゲート内に入っている探知データ及びＳ／Ｎ比、並びに入力した現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力する粗ゲート内外判定手段と、上記粗ゲート内外判定手段の出力を用いて、サンプリング時刻、探知データ数、航跡数に依存する、粗ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出して、算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）、並びに入力した粗ゲート内に入っている探知データ、粗ゲート内に入っているＳ／Ｎ比、現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力する各Ｓ／Ｎ比残差共分散行列算出手段と、上記各Ｓ／Ｎ比残差共分散行列算出手段の出力を用いて、粗ゲートより狭い目標存在期待領域である精ゲートを算出し、入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比について、精ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、サンプリング時刻、探知データ数、航跡数に依存する、精ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出し、判定した精ゲート内に入っている探知データ及びＳ／Ｎ比、算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）、並びに入力した現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力する精ゲート内外判定手段と、上記精ゲート内外判定手段の出力を用いて、現時刻の平滑誤差共分散行列と現時刻の平滑ベクトルを算出して出力するデータ更新手段と、上記データ更新手段の出力を用いて、現時刻より１サンプリング後の予測誤差共分散行列と現時刻より１サンプリング後の予測ベクトルを算出して出力する予測手段と、上記予測手段の出力を用いて、現時刻の予測誤差共分散行列と現時刻の予測ベクトルを出力する遅延手段とを備えたものである。
【００４７】
この発明に係る追尾装置は、精ゲート内外判定手段から出力される各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）の数と、観測手段が出力した探知データ数により、上記観測手段及びＳ／Ｎ比最小値残差共分散行列算出手段が使用しているＳ／Ｎ比最小値の値を変更し、上記観測手段及び上記Ｓ／Ｎ比最小値残差共分散行列算出手段に通知するＳ／Ｎ比最小値制御手段とを備え、上記観測手段及び上記Ｓ／Ｎ比最小値残差共分散行列算出手段が、変更されたＳ／Ｎ比最小値を使用して処理を行うものである。
【００４８】
この発明に係る追尾装置は、目標を探知するためのしきい値であるＳ／Ｎ比最小値を用いて、内部に備えているセンサから探知した探知データ及び探知データに付随するＳ／Ｎ比を出力する観測手段と、レーダの最大探知距離に基づき観測誤差共分散行列成分の内の最大値成分を選択し、選択した観測誤差共分散行列の最大値成分を用いて、サンプリング時刻、探知データ数、航跡数に依存しない観測誤差共分散行列Ｒ_min を算出し、上記観測手段からの出力、及び算出した観測誤差共分散行列Ｒ_min を用いて、探知データ数に依存しない簡易な残差共分散行列Ｓ’_kmin（ｔ）を算出し、算出した簡易な残差共分散行列Ｓ’_kmin（ｔ）、並びに入力した探知データ及びＳ／Ｎ比を出力するＳ／Ｎ比最小値残差共分散行列算出手段と、上記Ｓ／Ｎ比最小値残差共分散行列算出手段から入力した簡易な残差共分散行列Ｓ’_kmin（ｔ）により、目標存在期待領域である粗ゲートを算出し、入力した探知データ及びＳ／Ｎ比について、粗ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、判定した粗ゲート内に入っている探知データ及びＳ／Ｎ比を出力する粗ゲート内外判定手段と、上記粗ゲート内外判定手段の出力を用いて、サンプリング時刻、探知データ数、航跡数に依存する、粗ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出して、算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）、並びに入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比を出力する各Ｓ／Ｎ比残差共分散行列算出手段と、上記各Ｓ／Ｎ比残差共分散行列算出手段の出力を用いて、粗ゲートより狭い目標存在期待領域である精ゲートを算出し、入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比について、精ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、サンプリング時刻、探知データ数、航跡数に依存する、精ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出し、判定した精ゲート内に入っている探知データ及びＳ／Ｎ比、並びに算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を出力する精ゲート内外判定手段とを備えたものである。
【００４９】
この発明に係る追尾装置は、目標を探知するためのしきい値であるＳ／Ｎ比最小値を用いて、内部に備えているセンサから探知した探知データ及び探知データに付随するＳ／Ｎ比を出力する観測手段と、レーダの最大探知距離に基づき観測誤差共分散行列成分の内の最大値成分を選択し、選択した観測誤差共分散行列の最大値成分を用いて、サンプリング時刻、探知データ数、航跡数に依存しない観測誤差共分散行列Ｒ_min を算出し、上記観測手段からの出力、後述の遅延手段からの現時刻の予測誤差共分散行列及び現時刻の予測ベクトル、及び算出した観測誤差共分散行列Ｒ_min を用いて、探知データ数に依存しない簡易な残差共分散行列Ｓ’_kmin（ｔ）を算出し、算出した簡易な残差共分散行列Ｓ’_kmin（ｔ）、並びに入力した探知データ、Ｓ／Ｎ比、現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力するＳ／Ｎ比最小値残差共分散行列算出手段と、上記Ｓ／Ｎ比最小値残差共分散行列算出手段から入力した簡易な残差共分散行列Ｓ’_kmin（ｔ）及び現時刻の予測ベクトルにより、目標存在期待領域である粗ゲートを算出し、入力した探知データ及びＳ／Ｎ比について、粗ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、判定した粗ゲート内に入っている探知データ及びＳ／Ｎ比、並びに入力した現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力する粗ゲート内外判定手段と、上記粗ゲート内外判定手段の出力を用いて、サンプリング時刻、探知データ数、航跡数に依存する、粗ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出して、算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）、並びに入力した粗ゲート内に入っている探知データ、粗ゲート内に入っているＳ／Ｎ比、現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力する各Ｓ／Ｎ比残差共分散行列算出手段と、上記各Ｓ／Ｎ比残差共分散行列算出手段の出力を用いて、粗ゲートより狭い目標存在期待領域である精ゲートを算出し、入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比について、精ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、サンプリング時刻、探知データ数、航跡数に依存する、精ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出し、判定した精ゲート内に入っている探知データ及びＳ／Ｎ比、算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）、並びに入力した現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力する精ゲート内外判定手段と、上記精ゲート内外判定手段の出力を用いて、現時刻の平滑誤差共分散行列と現時刻の平滑ベクトルを算出して出力するデータ更新手段と、上記データ更新手段の出力を用いて、現時刻より１サンプリング後の予測誤差共分散行列と現時刻より１サンプリング後の予測ベクトルを算出して出力する予測手段と、上記予測手段の出力を用いて、現時刻の予測誤差共分散行列と現時刻の予測ベクトルを出力する遅延手段とを備えたものである。
【００５０】
この発明に係る追尾装置は、目標を探知するためのしきい値であるＳ／Ｎ比最小値を用いて、内部に備えているセンサから探知した探知データ及び探知データに付随するＳ／Ｎ比を出力する観測手段と、レーダの最大探知距離に基づき観測誤差共分散行列成分の内の最大値成分を選択し、選択した観測誤差共分散行列の最大値成分を用いて、サンプリング時刻、探知データ数、航跡数に依存しない観測誤差共分散行列Ｒ_min を算出し、上記観測手段からの出力、及び算出した観測誤差共分散行列Ｒ_min を用いて、探知データ数、航跡数に依存しない残差共分散行列Ｅ_k を算出し、算出した残差共分散行列Ｅ_k 、並びに入力した探知データ及びＳ／Ｎ比を出力するＳ／Ｎ比最小値残差共分散行列算出手段と、上記Ｓ／Ｎ比最小値残差共分散行列算出手段から入力した残差共分散行列Ｅ_kにより、目標存在期待領域である粗ゲートを算出し、入力した探知データ及びＳ／Ｎ比について、粗ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、判定した粗ゲート内に入っている探知データ及びＳ／Ｎ比を出力する粗ゲート内外判定手段と、上記粗ゲート内外判定手段の出力を用いて、サンプリング時刻、探知データ数、航跡数に依存する、粗ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出して、算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）、並びに入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比を出力する各Ｓ／Ｎ比残差共分散行列算出手段と、上記各Ｓ／Ｎ比残差共分散行列算出手段の出力を用いて、粗ゲートより狭い目標存在期待領域である精ゲートを算出し、入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比について、精ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、サンプリング時刻、探知データ数、航跡数に依存する、精ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出し、判定した精ゲート内に入っている探知データ及びＳ／Ｎ比、並びに算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を出力する精ゲート内外判定手段とを備えたものである。
【００５１】
この発明に係る追尾装置は、目標を探知するためのしきい値であるＳ／Ｎ比最小値を用いて、内部に備えているセンサから探知した探知データ及び探知データに付随するＳ／Ｎ比を出力する観測手段と、レーダの最大探知距離に基づき観測誤差共分散行列成分の内の最大値成分を選択し、選択した観測誤差共分散行列の最大値成分を用いて、サンプリング時刻、探知データ数、航跡数に依存しない観測誤差共分散行列Ｒ_min を算出し、上記観測手段からの出力、後述の遅延手段からの現時刻の予測誤差共分散行列及び現時刻の予測ベクトル、並びに算出した観測誤差共分散行列Ｒ_min を用いて、探知データ数、航跡数に依存しない残差共分散行列Ｅ_k を算出し、算出した残差共分散行列Ｅ_k 、並びに入力した探知データ、Ｓ／Ｎ比、現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力するＳ／Ｎ比最小値残差共分散行列算出手段と、上記Ｓ／Ｎ比最小値残差共分散行列算出手段から入力した残差共分散行列Ｅ_k 及び現時刻の予測ベクトルにより、目標存在期待領域である粗ゲートを算出し、入力した探知データ及びＳ／Ｎ比について、粗ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、判定した粗ゲート内に入っている探知データ及びＳ／Ｎ比、並びに入力した現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力する粗ゲート内外判定手段と、上記粗ゲート内外判定手段の出力を用いて、サンプリング時刻、探知データ数、航跡数に依存する、粗ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出して、算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）、並びに入力した粗ゲート内に入っている探知データ、粗ゲート内に入っているＳ／Ｎ比、現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力する各Ｓ／Ｎ比残差共分散行列算出手段と、上記各Ｓ／Ｎ比残差共分散行列算出手段の出力を用いて、粗ゲートより狭い目標存在期待領域である精ゲートを算出し、入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比について、精ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、サンプリング時刻、探知データ数、航跡数に依存する、精ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出し、判定した精ゲート内に入っている探知データ及びＳ／Ｎ比、算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）、並びに入力した現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力する精ゲート内外判定手段と、上記精ゲート内外判定手段の出力を用いて、現時刻の平滑誤差共分散行列と現時刻の平滑ベクトルを算出して出力するデータ更新手段と、上記データ更新手段の出力を用いて、現時刻より１サンプリング後の予測誤差共分散行列と現時刻より１サンプリング後の予測ベクトルを算出して出力する予測手段と、上記予測手段の出力を用いて、現時刻の予測誤差共分散行列と現時刻の予測ベクトルを出力する遅延手段とを備えたものである。
【００５２】
この発明に係る追尾装置は、目標を探知するためのしきい値であるＳ／Ｎ比最小値を用いて、内部に備えているセンサから探知した探知データ及び探知データに付随するＳ／Ｎ比を出力する観測手段と、レーダの最大探知距離に基づき観測誤差共分散行列成分の内の最大値成分を選択し、選択した観測誤差共分散行列の最大値成分、及び上記観測手段から入力した探知データ及びＳ／Ｎ比を出力する観測誤差共分散行列最大値選択手段と、上記観測誤差共分散行列最大値選択手段の出力を用いて、固定のサンプリング間隔Ｔ_int により、サンプリング時刻、探知データ数、航跡数に依存しない半径ｒａｄｉｕｓ０の粗ゲートを算出し、入力した探知データ及びＳ／Ｎ比について、粗ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、判定した粗ゲート内に入っている探知データ及びＳ／Ｎ比を出力する粗ゲート内外判定手段と、上記粗ゲート内外判定手段の出力を用いて、サンプリング時刻、探知データ数、航跡数に依存する、粗ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出して、算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）、並びに入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比を出力する各Ｓ／Ｎ比残差共分散行列算出手段と、上記各Ｓ／Ｎ比残差共分散行列算出手段の出力を用いて、粗ゲートより狭い目標存在期待領域である精ゲートを算出し、入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比について、精ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、サンプリング時刻、探知データ数、航跡数に依存する、精ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出し、判定した精ゲート内に入っている探知データ及びＳ／Ｎ比、並びに算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を出力する精ゲート内外判定手段とを備えたものである。
【００５３】
この発明に係る追尾装置は、目標を探知するためのしきい値であるＳ／Ｎ比最小値を用いて、内部に備えているセンサから探知した探知データ及び探知データに付随するＳ／Ｎ比を出力する観測手段と、レーダの最大探知距離に基づき観測誤差共分散行列成分の内の最大値成分を選択し、選択した観測誤差共分散行列の最大値成分、上記観測手段から入力した探知データ及びＳ／Ｎ比、並びに後述の遅延手段からの現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力する観測誤差共分散行列最大値選択手段と、上記観測誤差共分散行列最大値選択手段の出力を用いて、固定のサンプリング間隔Ｔ_int により、サンプリング時刻、探知データ数、航跡数に依存しない半径ｒａｄｉｕｓ０の粗ゲートを算出し、入力した探知データ及びＳ／Ｎ比について、粗ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、判定した粗ゲート内に入っている探知データ及びＳ／Ｎ比、並びに入力した現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力する粗ゲート内外判定手段と、上記粗ゲート内外判定手段の出力を用いて、サンプリング時刻、探知データ数、航跡数に依存する、粗ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出して、算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）、並びに入力した粗ゲート内に入っている探知データ、粗ゲート内に入っているＳ／Ｎ比、現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力する各Ｓ／Ｎ比残差共分散行列算出手段と、上記各Ｓ／Ｎ比残差共分散行列算出手段の出力を用いて、粗ゲートより狭い目標存在期待領域である精ゲートを算出し、入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比について、精ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、サンプリング時刻、探知データ数、航跡数に依存する、精ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出し、判定した精ゲート内に入っている探知データ及びＳ／Ｎ比、算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）、並びに入力した現時刻の予測誤差共分散行列及び現時刻の予測ベクトルを出力する精ゲート内外判定手段と、上記精ゲート内外判定手段の出力を用いて、現時刻の平滑誤差共分散行列と現時刻の平滑ベクトルを算出して出力するデータ更新手段と、上記データ更新手段の出力を用いて、現時刻より１サンプリング後の予測誤差共分散行列と現時刻より１サンプリング後の予測ベクトルを算出して出力する予測手段と、上記予測手段の出力を用いて、現時刻の予測誤差共分散行列と現時刻の予測ベクトルを出力する遅延手段とを備えたものである。
【００５４】
【発明の実施の形態】
以下、この発明の実施の一形態を説明する。
実施の形態１．
図１はこの発明の実施の形態１による追尾装置の構成を示すブロック図で、図において、２は、観測手段１が出力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、並びに遅延手段８が出力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を入力し、目標を探知するためのしきい値である、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存しないＳ／Ｎ比最小値ＳＮＲ_min を用いて計算した、探知データ数ｊに依存しない現時刻の観測誤差共分散行列Ｒ_kmin（ｔ）と、入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）とを用いて、探知データ数ｊに依存しない残差共分散行列Ｓ_kmin（ｔ）を算出し、算出した残差共分散行列Ｓ_kmin（ｔ）、並びに入力した探知データＤ_kj、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を出力するＳ／Ｎ比最小値残差共分散行列算出手段である。
【００５５】
また、図１において、３は、Ｓ／Ｎ比最小値残差共分散行列算出手段２からの残差共分散行列Ｓ_kmin（ｔ）、探知データＤ_kj、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を入力し、入力した残差共分散行列Ｓ_kmin（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）により、目標存在期待領域である粗ゲートを算出し、入力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjについて、粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを判定し、判定した粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、並びに入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を出力する粗ゲート内外判定手段であり、その他の構成は従来の図６に示すものと同等である。
【００５６】
次にＳ／Ｎ比最小値残差共分散行列算出手段２について詳細に説明する。
観測手段１のレーダ信号処理における信号検出では、雑音電力δに対して一定以上の電力を有する信号をもって目標が探知されたと判定する。この目標が探知されたと判定するためのしきい値であるＳ／Ｎ比を一定のＳ／Ｎ最小値ＳＮＲ_min とする。すなわち、δ・ＳＮＲ_min 以上の電力の信号を探知したとする。このとき、探知データＤ_kjのＳ／Ｎ比ＳＮＲ_kjは次の（３３）式を満たす。
【数２４】

【００５７】
また、Ｓ／Ｎ最小値ＳＮＲ_min から算出される距離、仰角、方位角の観測雑音を各々、σｒ_min ，σｅ_min ，σｂ_min とすると、σｒ_min ，σｅ_min ，σｂ_min は、上記（８）式、（９）式、（１０）式から、次の（３４）式、（３５）式、（３６）式のようになる。
【数２５】

【００５８】
上記（３３）式、（３４）式、（３５）式、（３６）式から、次の（３７）式、（３８）式、（３９）式の関係が成立する。
【数２６】

【００５９】
また、上記（３４）式、（３５）式、（３６）式より、Ｓ／Ｎ比最小値ＳＮＲ_min から算出される極座標による観測誤差共分散行列Λ_min は、上記（５）式より次の（４０）式で示される。
【数２７】

【００６０】
ここで、上記（５）式と上記（４０）式を比較した場合、上記（３７）式、（３８）式、（３９）式より、次の（４１）が成立する。
【数２８】

上記（４１）式は、極座標で表されるΛ_min −Λ_kjが、代数学による正値対称行列であることを示している。
【００６１】
さらに、Ｓ／Ｎ比最小値ＳＮＲ_min から算出される（４１）式の結果より、探知データ数ｊに依存しない、北基準直交座標における観測誤差共分散行列Ｒ_kmin（ｔ）は、上記（４）式より次の（４２）式となる。
【数２９】

【００６２】
上記（４１）式、（４）式、（４２）式から、次の（４３）式の関係が得られる。
【数３０】

【００６３】
さらに、探知データ数ｊに依存しないＳ／Ｎ比最小値残差共分散行列Ｓ_kmin（ｔ）は、次の（４４）式により算出される。
【数３１】

【００６４】
上記（２）式、（４４）式は、代数学による正値対称行列であるため、（２）式と（４４）式を比較すると、上記（４３）式の関係から、次の（４５）式が得られる。
【数３２】

上記（４５）式の結果は、探知データＤ_kjのＳ／Ｎ比ＳＮＲ_kjから算出される精ゲート内の探知データＤ_kjは、Ｓ／Ｎ比最小値ＳＮＲ_min から算出される粗ゲート内に必ず存在することを示している。
【００６５】
したがって、Ｓ／Ｎ比最小値残差共分散行列算出手段２では、上記（３４）式、（３５）式、（３６）式、（４０）式、（４２）式、（４４）式を使用して、Ｓ／Ｎ比最小値ＳＮＲ_min のときの、探知データ数ｊに依存しないＳ／Ｎ比最小値残差共分散行列Ｓ_kmin（ｔ）を算出して出力する。
【００６６】
次に粗ゲート内外判定手段３について詳細に説明する。
粗ゲート内外判定手段３では、Ｓ／Ｎ比最小値残差共分散行列算出手段２で算出されたＳ／Ｎ比最小値ＳＮＲ_min のときの、探知データ数ｊに依存しないＳ／Ｎ比最小値残差共分散行列Ｓ_kmin（ｔ）を用いて、次の（４６）式により粗ゲート判定を行ない、（４６）式を満たす探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを出力する。
【数３３】

【００６７】
従って、従来の追尾装置では、上記（１１）式を用いて、全ての探知データＤ_kjに関して、（１１）式における各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）の逆行列計算を行うことになっていたため、計算機の演算負荷が膨大となっている。しかし、上記のように、現時刻より１サンプリング前の航跡数をＬとし、現時刻における探知データ数をＭとしたとき、上記（４６）式のＳ／Ｎ比最小値残差共分散行列Ｓ_kmin（ｔ）は、サンプリング時刻ｋ、航跡数には依存するが、探知データ数ｊには依存しない値なので、Ｌ個の逆行列の計算で済む。また、（４６）式による粗ゲート判定の後、探知データ数ｊが絞れるだけではなく、航跡数も減少するため、精ゲート内外判定手段５、データ更新手段６、予測手段７の各処理における計算の演算負荷が軽減される。
【００６８】
次に動作について説明する。
図２はこの発明の実施の形態１による追尾装置の処理の流れを示すフローチャートである。ステップＳＴ１において、観測手段１は、目標を探知するためのしきい値である、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存しないＳ／Ｎ比最小値ＳＮＲ_min を用いて、内部に備えているセンサから探知した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを、Ｓ／Ｎ比最小値残差共分散行列算出手段２に出力する。
【００６９】
ステップＳＴ２において、Ｓ／Ｎ比最小値残差共分散行列算出手段２は、観測手段１が出力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、並びに遅延手段８が出力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を入力し、Ｓ／Ｎ比最小値ＳＮＲ_min を用いて計算した、探知データ数ｊに依存しない現時刻の観測誤差共分散行列Ｒ_kmin（ｔ）と、入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）とを用いて、探知データ数ｊに依存しない残差共分散行列Ｓ_kmin（ｔ）を算出し、算出した残差共分散行列Ｓ_kmin（ｔ）、並びに入力した探知データＤ_kj、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を粗ゲート内外判定手段３に出力する。
【００７０】
ステップＳＴ３において、粗ゲート内外判定手段３は、入力した残差共分散行列Ｓ_kmin（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）により、目標存在期待領域である粗ゲートを算出し、入力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjについて、粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを判定し、判定した粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、並びに入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を各Ｓ／Ｎ比残差共分散行列算出手段４に出力する。
【００７１】
ステップＳＴ４において、各Ｓ／Ｎ比残差共分散行列算出手段４は、入力した粗ゲート内に入っている探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを用いて計算した、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存する現時刻の観測誤差共分散行列Ｒ_kj（ｔ）と、入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）とを用いて、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存する、粗ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出して、算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）、並びに入力した粗ゲート内に入っている探知データＤ_kj、粗ゲート内に入っている探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を精ゲート内外判定手段５に出力する。
【００７２】
ステップＳＴ５において、精ゲート内外判定手段５は、入力した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）により、粗ゲートより狭い目標存在期待領域である精ゲートを算出し、入力した粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjについて、精ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを判定し、精ゲート内に入っている探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを用いて計算した、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存する現時刻の観測誤差共分散行列Ｒ_kj（ｔ）と、入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）とを用いて、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存する、精ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出し、判定した精ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）、並びに入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）をデータ更新手段６に出力する。
【００７３】
ステップＳＴ６において、データ更新手段６は、入力した精ゲート内に入っている探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）及び現時刻の予測誤差共分散行列Ｐ_pk（ｔ）により平滑情報である現時刻の平滑誤差共分散行列Ｐ_sk（ｔ）を算出すると共に、入力した精ゲート内に入っている探知データＤ_kj及び現時刻の予測ベクトルＶＸ_pk（ｔ）により平滑情報である現時刻の平滑ベクトルＶＸ_sk（ｔ）を算出して、算出した現時刻の平滑誤差共分散行列Ｐ_sk（ｔ）、現時刻の平滑ベクトルＶＸ_sk（ｔ）を予測手段７に出力する。
【００７４】
ステップＳＴ７において、予測手段７は、入力した現時刻の平滑誤差共分散行列Ｐ_sk（ｔ）により、予測情報である現時刻より１サンプリング後の予測誤差共分散行列Ｐ_pk+1（ｔ）を算出すると共に、入力した現時刻の平滑ベクトルＶＸ_sk（ｔ）により、予測情報である現時刻より１サンプリング後の予測ベクトルＶＸ_pk+1（ｔ）を算出する。
【００７５】
ステップＳＴ８において、オペレータの指示又は自動判定により処理を継続する場合には、予測手段７が算出した現時刻より１サンプリング後の予測誤差共分散行列Ｐ_pk+1（ｔ）、現時刻より１サンプリング後の予測ベクトルＶＸ_pk+1（ｔ）を遅延手段８に出力する。
【００７６】
ステップＳＴ９において、遅延手段８は、入力した現時刻より１サンプリング後の予測誤差共分散行列Ｐ_pk+1（ｔ）を１サンプリング遅延させて、現時刻の予測誤差共分散行列Ｐ_pk（ｔ）をＳ／Ｎ比最小値残差共分散行列算出手段２に出力すると共に、入力した現時刻より１サンプリング後の予測ベクトルＶＸ_pk+1（ｔ）を１サンプリング遅延させて、現時刻の予測ベクトルＶＸ_pk（ｔ）をＳ／Ｎ比最小値残差共分散行列算出手段２に出力し、上記ステップＳＴ１からの処理を繰り返す。
【００７７】
以上のように、この実施の形態１によれば、粗ゲート内外判定手段５が、Ｓ／Ｎ比最小値残差共分散行列算出手段２が算出した探知データ数ｊに依存しない残差共分散行列Ｓ_kmin（ｔ）により、目標存在期待領域である粗ゲートを算出し、入力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjについて、粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを判定し、精ゲート内外判定手段５が、粗ゲートより狭い目標存在期待領域である精ゲートを算出し、入力した粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjについて、精ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを判定することにより、精ゲート内外判定手段５における、目標信号にとり不要な信号を除去するための精ゲートの演算負荷が軽減されると共に、データ更新手段６、予測手段７の各処理における計算の演算負荷も軽減されるという効果が得られる。
【００７８】
実施の形態２．
図３はこの発明の実施の形態２による追尾装置の構成を示すブロック図で、図において、１２は、観測手段１が出力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、並びに遅延手段８が出力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を入力し、レーダの最大探知距離ＲＡＮＧＥ_max により観測誤差共分散行列Ｒ_kj成分の最大値成分を選択し、選択した観測誤差共分散行列Ｒ_kj成分の最大値成分により、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存しない、北基準直交座標における観測誤差共分散行列Ｒ_min を算出し、算出した観測誤差共分散行列Ｒ_min と、入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）とを用いて、探知データ数ｊに依存しない簡易な残差共分散行列Ｓ’_kmin（ｔ）を算出し、算出した簡易な残差共分散行列Ｓ’_kmin（ｔ）、並びに入力した探知データＤ_kj、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を出力するＳ／Ｎ比最小値残差共分散行列算出手段である。
【００７９】
また、図３において、１３は、Ｓ／Ｎ比最小値残差共分散行列算出手段１２からの簡易な残差共分散行列Ｓ’_kmin（ｔ）、探知データＤ_kj、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を入力し、入力した簡易な残差共分散行列Ｓ’_kmin（ｔ）、及び現時刻の予測ベクトルＶＸ_pk（ｔ）により、目標存在期待領域である粗ゲートを算出し、入力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjについて、粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを判定し、判定した粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、並びに入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を出力する粗ゲート内外判定手段であり、その他の構成は実施の形態１の図１と同等である。
【００８０】
次にＳ／Ｎ比最小値残差共分散行列算出手段１２について詳細に説明する。
ここで、レーダでは、処理する最大探知距離ＲＡＮＧＥ_max の値が既知であるため、次の（４７）式のように、観測誤差共分散行列Ｒ_kj成分の最大値成分が定義できる。
【数３４】

上記（４７）式において、σｃｒ_min ，σｃｅ_min ，σｃｂ_min は、それぞれ、正の定数とする。また、ｍａｘｓｅｌｅｃｔ（Ａ，Ｂ，Ｃ）は、Ａ，Ｂ，Ｃ，．．．の中の最大値１つを選びだす関数である。
【００８１】
ここで、次の（４８）式のように、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存しない、北基準直交座標における観測誤差共分散行列Ｒ_min を定義したとき、上記（４４）式における、探知データ数ｊに依存しない観測誤差共分散行列Ｒ_kmin（ｔ）とは、次の（４９）式の関係がある。
【数３５】

【００８２】
したがって、上記（４４）式のＲ_kmin（ｔ）を上記（４８）式のＲ_min で置き換えると、次の（５０）式となる。
【数３６】

上記（５０）式における予測誤差共分散行列Ｐ_pk（ｔ）は、航跡番号（ｔ）ごとに異なる値を持つため、（５０）式におけるＨＰ_pk（ｔ）Ｈ’及び残差共分散行列Ｓ_min （ｔ）も航跡番号（ｔ）ごとに異なる値を持つ。
【００８３】
観測誤差共分散行列Ｒ_min から上記（５０）式を用いて算出される簡易な残差共分散行列Ｓ’_kmin（ｔ）と、各サンプリング時刻ｋのＳ／Ｎ比最小値ＳＮＲ_min から上記（４４）式を用いて、逐次算出される残差共分散行列Ｓ_kmin（ｔ）の関係は、次の（５１）式となる。
【数３７】

【００８４】
上記（５１）式は、観測誤差共分散行列Ｒ_min から（５０）式を用いて算出される簡易な残差共分散行列Ｓ’_kmin（ｔ）により算出された粗ゲートが、各サンプリング時刻ｋのＳ／Ｎ比最小値ＳＮＲ_min から（４４）式を用いて、逐次算出されるＳ／Ｎ比最小値残差共分散行列Ｓ_kmin（ｔ）により算出された粗ゲートより同じ又は広いことを示している。
【００８５】
そこで、Ｓ／Ｎ比最小値残差共分散行列算出手段１２では、上記（４８）式を用いて算出した観測誤差共分散行列Ｒ_min を使用する。つまり上記（４４）式における観測誤差共分散行列Ｒ_kmin（ｔ）は、探知データ数ｊ、航跡数に依存するので、逐次計算を行う必要があるのに対し、上記（５０）式における観測誤差共分散行列Ｒ_min は、サンプリング時刻ｋ、探知データＤ_kj、航跡（ｔ）に依存しない定数であるので、（５０）式を用いた場合の計算機の演算負荷は、（４４）式を用いた場合の計算機の演算負荷に比べてさらに軽くなる。
【００８６】
次に動作について説明する。
図４はこの発明の実施の形態２による追尾装置の処理の流れを示すフローチャートである。ステップＳＴ１は実施の形態１の図２と同様である。
【００８７】
ステップＳＴ１１において、Ｓ／Ｎ比最小値残差共分散行列算出手段１２は、レーダの最大探知距離ＲＡＮＧＥ_max に基づき、上記（４７）式を用いて、観測誤差共分散行列Ｒ_kj成分の距離方向の誤差分散σｃｒ² _min、仰角方向の誤差分散ＲＡＮＧＥ² _maxσｃｅ_min 、方位角方向の誤差分散ＲＡＮＧＥ² _maxσｃｂ_min の内、観測誤差共分散行列Ｒ_kj成分の内の最大値成分を選択する。
【００８８】
ステップＳＴ１２において、Ｓ／Ｎ比最小値残差共分散行列算出手段１２は、選択した観測誤差共分散行列Ｒ_kj成分の最大値成分を用いて、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存しない、北基準直交座標における観測誤差共分散行列Ｒ_min を（４８）式を用いて算出する。
【００８９】
ステップＳＴ１３において、Ｓ／Ｎ比最小値残差共分散行列算出手段１２は、観測手段１が出力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、並びに遅延手段８が出力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を入力し、算出した観測誤差共分散行列Ｒ_min と、入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）とを用いて、探知データ数ｊに依存しない簡易な残差共分散行列Ｓ’_kmin（ｔ）を算出し、算出した簡易な残差共分散行列Ｓ’_kmin（ｔ）、並びに入力した探知データＤ_kj、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を粗ゲート内外判定手段１３に出力する。
【００９０】
ステップＳＴ１４において、粗ゲート内外判定手段１３は、入力した簡易な残差共分散行列Ｓ’_kmin（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）により、目標存在期待領域である粗ゲートを算出し、入力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjについて、粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを判定し、判定した粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、並びに入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を各Ｓ／Ｎ比残差共分散行列算出手段４に出力する。
【００９１】
ステップＳＴ４〜ＳＴ９までの処理は、実施の形態１の図２と同様である。したがって、以上のステップにより、上記（５０）式を用いた場合の計算機の演算負荷は、上記（４４）式を用いた場合の計算機の演算負荷に比べ、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存しない測誤差共分散行列Ｒ_min を使用しているので、さらに軽くなる。
【００９２】
以上のように、この実施の形態２によれば、Ｓ／Ｎ比最小値残差共分散行列算出手段１２が、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存しない観測誤差共分散行列Ｒ_min を算出し、算出した観測誤差共分散行列Ｒ_min を使用して、探知データ数ｊに依存しない簡易な残差共分散行列Ｓ’_kmin（ｔ）を算出するので、実施の形態１に比べ、粗ゲート内外判定手段１３、各Ｓ／Ｎ比残差共分散行列算出手段４、精ゲート内外判定手段５、データ更新手段６、予測手段７の各処理における演算負荷が軽減されるという効果が得られる。
【００９３】
実施の形態３．
図５はこの発明の実施の形態３による追尾装置の構成を示すブロック図で、図において、２２は、観測手段１が出力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、並びに遅延手段８が出力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を入力し、レーダの最大探知距離ＲＡＮＧＥ_max により観測誤差共分散行列Ｒ_kj成分の最大値成分を選択し、選択した観測誤差共分散行列Ｒ_kj成分の最大値成分により、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存しない、北基準直交座標における観測誤差共分散行列Ｒ_min を算出し、算出した観測誤差共分散行列Ｒ_min と、入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）とを用いて、探知データ数ｊ、航跡数に依存しない残差共分散行列Ｅ_k を算出し、算出した残差共分散行列Ｅ_k 、並びに入力した探知データＤ_kj、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を出力するＳ／Ｎ比最小値残差共分散行列算出手段である。
【００９４】
また、図５において、２３は、Ｓ／Ｎ比最小値残差共分散行列算出手段２２からの残差共分散行列Ｅ_k 、探知データＤ_kj、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を入力し、入力した残差共分散行列Ｅ_k 及び現時刻の予測ベクトルＶＸ_pk（ｔ）により、目標存在期待領域である粗ゲートを算出し、入力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjについて、粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを判定し、判定した粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、並びに入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を出力する粗ゲート内外判定手段であり、その他の構成は実施の形態１の図１と同等である。
【００９５】
次にＳ／Ｎ比最小値残差共分散行列算出手段２２について詳細に説明する。
まず、カルマンフィルタの定義より、平滑誤差共分散行列Ｐ_sk（ｔ）は、目標の真値を表す状態ベクトルＶＸ_k （ｔ）、平滑ベクトルＶＸ_sk（ｔ）を用いて、次の（５２）式のように示される。
【数３８】

上記（５２）式において、Ｅ［ ］は、平均を表す記号である。
【００９６】
ここで、平滑ベクトルＶＸ_sk（ｔ）及び平滑誤差共分散行列Ｐ_sk（ｔ）の簡易な推定値として、探知データＤ_kjから直接算出する平滑ベクトルＶＶＸ_skj （ｔ）及び探知データＤ_kjから直接算出する平滑誤差共分散行列ＰＰ_sk（ｔ）をそれぞれ次の（５３）式、（５４）式のように定義する。
【数３９】

上記（５３）式において、ＶＺ_kjは、粗ゲート内のクラッタの探知データを除いた目標の探知データを示す。またｔ_k はサンプリング時刻ｋを示す。
【００９７】
ここで、カルマンフィルタの理論により、運動モデルは（５５）式のように示される。
【数４０】

上記（５５）式において、Φ_k-1 は時刻ｔ_k-1 からｔ_k への状態ベクトルの既知の推移行列であり、次の（５６）式のように示される。
【数４１】

次の（５６）式において、Ｉは３×３の単位行列である。
【００９８】
上記（５５）式において、Ｗ_k は、平均ゼロベクトル、誤差共分散行列Ｑ_k の３変量正規分布に従う駆動雑音ベクトルで、加速度ベクトルに相当する。また、Δ_k-1 は駆動雑音ベクトルの既知の変換行列で、次の（５７）式のように示される。
【数４２】

【００９９】
上記（５５）式における状態ベクトルＶＸ_k （ｔ）は、目標の位置ベクトル及び速度ベクトルからなるベクトルで、３次元空間の場合、次の（５８）式のように示される。
【数４３】

上記（５８）式において、ｘ_k ，ｙ_k ，ｚ_k は北基準直交座標における任意の位置を示し、ｘ_k ，ｙ_k ，ｚ_k の上部の記号「・」は１次の時間微分を示し、上部の記号「・」を付与したｘ_k ，ｙ_k ，ｚ_k はその位置における速度を示す。
【０１００】
また、カルマンフィルタの理論より、粗ゲート内のクラッタの探知データを除いた目標の探知データＶＺ_kjは、次の（５９）式のようになる。
【数４４】

上記（５９）式において、また、Ｈは既知の観測行列であり、次の（６０）式のように示される。
【数４５】

【０１０１】
また、上記（５９）式において、Ｕ_kjは平均ゼロベクトルで、共分散行列Λ_kjの３変量白色正規分布に従う時刻ｔ_k における極座標による観測雑音ベクトルであり、Γ_kjは観測雑音ベクトルＵ_kjの既知の変換行列である。ここで、上記（６）式におけるΓ_kj（ｔ）との関係であるが、観測雑音ベクトルＵ_kjの既知の変換行列Γ_kjを算出する際は真値を用いるが、実際は目標の真値がわからないため、（６）式のように真値の代りに予測値を入れる。したがって、Γ_kjとΓ_kj（ｔ）を同じとして扱う。
【０１０２】
よって、上記（５４）式に、上記（５３）式、（５５）式、（５９）式、（４）式を適用すると、次の（６１）式が得られる。
【数４６】

上記（６１）式におけるＱ_k は、上記（５５）式における駆動雑音Ｗ_k の共分散行列である。
【０１０３】
ここで、カルマンフィルタの理論より、平滑ベクトルＶＸ_sk（ｔ）は、平滑誤差共分散行列Ｐ_sk（ｔ）の対角成分の和を最小にするという意味での最適値であるのに対し、探知データＤ_kjから直接算出する平滑ベクトルＶＶＸ_sk（ｔ）は、探知データＤ_kjから直接算出する平滑誤差共分散行列Ｐ_sk（ｔ）の対角成分の和を最小にするという意味での最適値ではない。したがって、次の（６２）式の関係が成り立つ。
【数４７】

【０１０４】
上記（６２）式の関係から、次の（６３）式の関係が成り立つ。
【数４８】

【０１０５】
ここで、カルマンフィルタの理論より、予測誤差共分散行列Ｐ_pk+1（ｔ）は、次の（６４）式により算出される。
【数４９】

【０１０６】
探知データＤ_kjから直接算出される平滑誤差共分散行列ＰＰ_sk（ｔ）を用いて、１サンプリング後の探知データＤ_kjから直接算出される予測誤差共分散行列はＰＰ_pk+1（ｔ）は、次の（６５）式のように算出される。
【数５０】

【０１０７】
上記（６５）式、（６１）式、（５６）式、（５７）式、及び観測行列Ｈの定義（６０）式を用いると、次の（６６）式を得る。
【数５１】

【０１０８】
ここで、上記（４３）式及び（４９）式の関係より、次の（６７）式が成立する。
【数５２】

【０１０９】
従って、上記（６６）式を（６７）式に代入すると、次の（６８）式が得られる。
【数５３】

【０１１０】
また、上記（６６）式、（６７）式、（６８）式を用いて、次の（６９）式が得られる。
【数５４】

【０１１１】
上記（６９）式の左辺は、上記（２）式の右辺における予測誤差共分散行列Ｐ_pk（ｔ）を、探知データＤ_kjから直接算出される予測誤差共分散行列ＰＰ_pk（ｔ）に置き換えたものとみなすことができる。ここで、予測誤差共分散行列ＰＰ_pk（ｔ）から算出される残差共分散行列ＳＳ_k （ｔ）を、次の（７０）式のように定める。また、上記（６９）式の右辺を次の（７１）式のように定める。
【数５５】

上記（７１）式において、Ｅ_k は探知データ数ｊ、航跡数に依存しない残差共分散行列を示す。
【０１１２】
そうすると、上記（６９）式、（７０）式、（７１）式から、次の（７２）式が算出される。
【数５６】

【０１１３】
また、上記（２）式、（６３）式、（６４）式、（６５）式から、次の（７３）式が成立する。
【数５７】

【０１１４】
よって、上記（７２）式、（７３）式から、次の（７４）式が成立する。
【数５８】

【０１１５】
上記（７４）式は、次の（７５）式で定義される粗ゲートが、上記（１１）式で定義される精ゲートよりも広いことを示す。つまり、（１１）式で定義される精ゲート内の探知データＤ_kjは、（７５）式で定義される粗ゲート内に必ず存在する。
【数５９】

【０１１６】
したがって、Ｓ／Ｎ比最小値残差共分散行列算出手段２２では、演算負荷が重いとユーザが判断したとき、上記（７１）式を用いて探知データ数ｊ、航跡数に依存しない残差共分散行列Ｅ_k を算出する。
【０１１７】
ここで、上記（７１）式におけるＱ_k は、運用前に決めておく固定値であるため、（７１）式における残差共分散行列Ｅ_k は、探知データ数ｊ、航跡数に依存しない固定値により求まることを示している。つまり、上記（７５）式の粗ゲートを用いた場合の計算機の演算負荷は、上記（４６）式のＳ_kmin（ｔ）をＳ’_kmin（ｔ）に置き換えた粗ゲートを用いた場合の計算機の演算負荷に比べて、さらに軽くなる。
【０１１８】
次に動作について説明する。
図６はこの発明の実施の形態３による追尾装置の処理の流れを示すフローチャートである。ステップＳＴ１は実施の形態１の図２と同様である。
【０１１９】
ステップＳＴ２１において、Ｓ／Ｎ比最小値残差共分散行列算出手段２２は、レーダの最大探知距離ＲＡＮＧＥ_max に基づき、上記（４７）式を用いて、観測誤差共分散行列Ｒ_kj成分の距離方向の誤差分散σｃｒ² _min、仰角方向の誤差分散ＲＡＮＧＥ² _maxσｃｅ_min 、方位角方向の誤差分散ＲＡＮＧＥ² _maxσｃｂ_min の内、観測誤差共分散行列Ｒ_kj成分の内の最大値成分を選択する。
【０１２０】
ステップＳＴ２２において、Ｓ／Ｎ比最小値残差共分散行列算出手段２２は、選択した観測誤差共分散行列Ｒ_kj成分の最大値成分を用いて、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存しない、北基準直交座標における観測誤差共分散行列Ｒ_min を上記（４８）式を用いて算出する。
【０１２１】
ステップＳＴ２３において、Ｓ／Ｎ比最小値残差共分散行列算出手段２２は、観測手段１が出力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、遅延手段８が出力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を入力し、算出した観測誤差共分散行列Ｒ_min と、入力した探知データＤ_kjから直接算出される予測誤差共分散行列ＰＰ_pk（ｔ）とを用いて、探知データ数ｊ、航跡数に依存しない残差共分散行列Ｅ_k を算出し、算出した残差共分散行列Ｅ_k、並びに入力した探知データＤ_kj、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を粗ゲート内外判定手段２３に出力する。
【０１２２】
ステップＳＴ２４において、粗ゲート内外判定手段２３は、入力した残差共分散行列Ｅ_k、及び現時刻の予測ベクトルＶＸ_pk（ｔ）により、目標存在期待領域である粗ゲートを算出し、入力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjについて、粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを判定し、判定した粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、並びに入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を各Ｓ／Ｎ比残差共分散行列算出手段４に出力する。
【０１２３】
ステップＳＴ４〜ＳＴ９までの処理は、実施の形態１の図２と同様である。したがって、以上のステップにより、上記（７１）式を用いた場合の計算機の演算負荷は、上記（５０）式を用いた場合の計算機の演算負荷に比べ、探知データ数ｊ、航跡数に依存しない残差共分散行列Ｅ_k を使用しているので、さらに軽くなる。
【０１２４】
以上のように、この実施の形態３によれば、探知データ数ｊ、航跡数に依存しない残差共分散行列Ｅ_k により粗ゲート内外判定を行うことにより、実施の形態２に比べて、粗ゲート内外判定手段２３、各Ｓ／Ｎ比残差共分散行列算出手段４、精ゲート内外判定手段５、データ更新手段６、予測手段７の各処理における計算の演算負荷が軽減されるという効果が得られる。
【０１２５】
実施の形態４．
図７はこの発明の実施の形態４による追尾装置の構成を示すブロック図で、図において、３２は、観測手段１が出力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、並びに遅延手段８が出力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を入力し、レーダの最大探知距離ＲＡＮＧＥ_max により観測誤差共分散行列Ｒ_kj成分の最大値成分を選択し、選択した観測誤差共分散行列Ｒ_kj成分の最大値成分、並びに入力した探知データＤ_kj、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を出力する観測誤差共分散行列最大値選択手段である。
【０１２６】
また、図７において、３３は、観測誤差共分散行列最大値選択手段３２からの観測誤差共分散行列Ｒ_kj成分の最大値成分、探知データＤ_kj、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を入力し、固定のサンプリング間隔Ｔ_int により、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存しない半径ｒａｄｉｕｓ０の粗ゲートを算出し、入力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjについて、粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを判定し、判定した粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、並びに入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を出力する粗ゲート内外判定手段であり、その他の構成は実施の形態１の図１と同等である。
【０１２７】
ここで、半径ｒａｄｉｕｓ０の粗ゲートについて説明する。
次の（７６）式のようにサンプリング間隔を一定値Ｔ_int とし、駆動雑音共分散行列Ｑ_k を座標軸及び時刻に無関係な対角行列として、次の（７７）式のように示す。
【数６０】

上記（７７）式におけるｑは駆動雑音定数である。
【０１２８】
そうすると、上記（４８）式、（７１）式、（７６）式より、上記（７５）式の粗ゲートは、次の（７８）式のように書き直すことができる。
【数６１】

【０１２９】
上記（７８）式において、半径ｒａｄｉｕｓ０は、次の（７９）式のように示される。
【数６２】

【０１３０】
ここで、上記（７５）式で定義される粗ゲートに対し、上記（７８）式で定義される半径ｒａｄｉｕｓ０の粗ゲートは、サンプリング間隔が固定されているため、サンプリング間隔によらない。したがって、（７８）式の半径ｒａｄｉｕｓ０の粗ゲートを用いた場合の計算機の演算負荷は、上記（７５）式の粗ゲートを用いた場合の計算機負荷に比べて、さらに軽くなる。
【０１３１】
次に動作について説明する。
図８はこの発明の実施の形態４による追尾装置の処理の流れを示すフローチャートである。ステップＳＴ１は実施の形態１の図１と同様である。
【０１３２】
ステップＳＴ３１において、観測誤差共分散行列最大値選択手段３２は、観測手段１が出力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、並びに遅延手段８が出力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を入力し、レーダの最大探知距離ＲＡＮＧＥ_max により観測誤差共分散行列Ｒ_kj成分の最大値成分を選択し、選択した観測誤差共分散行列Ｒ_kj成分の最大値成分、並びに入力した探知データＤ_kj、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を粗ゲート内外判定手段３３に出力する。
【０１３３】
ステップＳＴ３２において、粗ゲート内外判定手段３３は、観測誤差共分散行列最大値選択手段３２からの観測誤差共分散行列Ｒ_kj成分の最大値成分、探知データＤ_kj、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を入力し、固定のサンプリング間隔Ｔ_int により上記（７９）式を用いて、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存しない半径ｒａｄｉｕｓ０の粗ゲートを算出する。
【０１３４】
ステップＳＴ３３において、粗ゲート内外判定手段３３は、入力した探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjについて、（７８）式を用いて、算出した半径ｒａｄｉｕｓ０の粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを判定し、判定した粗ゲート内に入っている探知データＤ_kj及び探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kj、並びに入力した現時刻の予測誤差共分散行列Ｐ_pk（ｔ）及び現時刻の予測ベクトルＶＸ_pk（ｔ）を各Ｓ／Ｎ比残差共分散行列算出手段４に出力する。
【０１３５】
ステップＳＴ４〜ＳＴ９までの処理は、実施の形態１の図２と同様である。したがって、以上のステップにより、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存しない半径ｒａｄｉｕｓ０の粗ゲートにより、（７８）式を用いて粗ゲート内外判定をする場合の計算機の演算負荷は、探知データ数ｊ、航跡数に依存しない残差共分散行列Ｅ_k により、上記（７１）式を用いて粗ゲート内外判定を行う場合の計算機の演算負荷に比べて、さらに軽くなる。
【０１３６】
以上のように、この実施の形態４によれば、サンプリング時刻ｋ、探知データ数ｊ、航跡数に依存しない半径ｒａｄｉｕｓ０の粗ゲートにより粗ゲート内外判定を行うことにより、実施の形態３に比べて、粗ゲート内外判定手段３３、各Ｓ／Ｎ比残差共分散行列算出手段４、精ゲート内外判定手段５、データ更新手段６、予測手段７の各処理における計算の演算負荷が軽減されるという効果が得られる。
【０１３７】
実施の形態５．
図９はこの発明の実施の形態５による追尾装置の構成を示すブロック図で、図において、４１は、精ゲート内外判定手段５から出力される各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）の数、すなわち航跡数と、観測手段１が探知した現サンプリング時刻の探知データ数ｊを比較し、観測手段１及びＳ／Ｎ比最小値残差共分散行列算出手段２が使用しているＳ／Ｎ比最小値ＳＮＲ_min の値を変更し、観測手段１及びＳ／Ｎ比最小値残差共分散行列算出手段２に通知するＳ／Ｎ比最小値制御手段であり、その他の構成は実施の形態１の図１と同等である。
【０１３８】
観測手段１における信号検出では、雑音電力δに対して一定以上の電力を有する信号により目標が探知されたと判定を行い、この目標が探知されたと判定するためのしきい値であるＳ／Ｎ比をＳ／Ｎ比最小値ＳＮＲ_min としたとき、観測手段１は、δ・ＳＮＲ_min 以上の電力の信号を、探知された目標の探知データＤ_kj、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjとして出力する。ここで、Ｓ／Ｎ比最小値ＳＮＲ_min は、誤警報確率を決めるパラメータでもある。
【０１３９】
従って、Ｓ／Ｎ比最小値ＳＮＲ_min の値を大きくして、誤警報確率を小さくすれば、得られる探知データＤ_kjの総数を削減できる。このため、１サンプリング前の精ゲート内外判定手段５から出力される各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）の数、すなわち航跡数に比べ、現サンプリング時刻の探知データ数ｊが著しく多い場合、次の（８０）式のように、観測手段１が、使用しているＳ／Ｎ比最小値ＳＮＲ_min の値より大きいＳ／Ｎ比最小値ＳＳＮＲ_min を使用すれば、探知データ数ｊを絞り込むことができる。
【数６３】

【０１４０】
すなわち、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjが、次の（８１）式を満たすときのみ、Ｓ／Ｎ比最小値ＳＳＮＲ_min を追尾に使用して、それ以外は追尾に使用しない。この処理により、追尾に使用する探知データ数ｊが削減でき、計算機の演算負荷が軽くなる。
【数６４】

【０１４１】
このように、観測手段１は所定のＳ／Ｎ比最小値ＳＮＲ_min を用いて探知データＤ_kjを得ているが、精ゲート内外判定手段５の判定結果により、目標の探知データＤ_kj以外のクラッタの探知データＤ_kjが多くなると、Ｓ／Ｎ比最小値ＳＮＲ_min を大きくする必要がある。逆に目標の探知データＤ_kjが少ない場合には、Ｓ／Ｎ比最小値ＳＮＲ_min を小さくする必要がある。
【０１４２】
次に動作について説明する。
図１０はこの発明の実施の形態５による追尾装置の処理の流れを示すフローチャートである。ステップＳＴ１〜ＳＴ９までの処理は、実施の形態１の図１と同等である。
【０１４３】
ステップＳＴ４１において、Ｓ／Ｎ比最小値制御手段４１は、精ゲート内外判定手段５から出力される各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）の数、すなわち航跡数と、観測手段１が探知した現サンプリング時刻における探知データ数ｊを比較する。
【０１４４】
ステップＳＴ４２において、Ｓ／Ｎ比最小値制御手段４１は、比較した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）の数、すなわち航跡数と、観測手段１が探知した現サンプリング時刻における探知データ数ｊが著しく異なる場合に、観測手段１及びＳ／Ｎ比最小値残差共分散行列算出手段２が使用していたＳ／Ｎ比最小値ＳＮＲ_min の値を変更して、観測手段１及びＳ／Ｎ比最小値残差共分散行列算出手段２に通知する。
【０１４５】
例えば、各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）の数、すなわち航跡数に比べて、観測手段１が探知した現サンプリング時刻における探知データ数ｊが著しく多い場合に、観測手段１及びＳ／Ｎ比最小値残差共分散行列算出手段２が使用しているたＳ／Ｎ比最小値ＳＮＲ_min の値より大きいＳ／Ｎ比最小値ＳＳＮＲ_min に変更して、観測手段１及びＳ／Ｎ比最小値残差共分散行列算出手段２に通知する。
【０１４６】
上記ステップＳＴ４２で、Ｓ／Ｎ比最小値を変更した後、ステップＳＴ１に戻り、観測手段１は、変更されたＳ／Ｎ比最小値ＳＮＲ_min を用いて、実施の形態１と同様にして探知データＤ_kj、探知データＤ_kjに付随するＳ／Ｎ比ＳＮＲ_kjを出力し、ステップＳＴ２において、Ｓ／Ｎ比最小値残差共分散行列算出手段２は、変更されたＳ／Ｎ比最小値ＳＮＲ_min を用いて、実施の形態１と同様にして観測誤差共分散行列Ｒ_kmin（ｔ）を算出する。以降の処理は、実施の形態１の図２と同様である。
【０１４７】
以上のように、この実施の形態５によれば、精ゲート内外判定手段５から出力される各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）の数、すなわち航跡数と、観測手段１が探知した現サンプリング時刻の探知データ数ｊが著しく異なる場合に、Ｓ／Ｎ比最小値制御手段４１が、観測手段１及びＳ／Ｎ比最小値残差共分散行列算出手段２が使用しているＳ／Ｎ比最小値ＳＮＲ_min の値を変更し、観測手段１及びＳ／Ｎ比最小値残差共分散行列算出手段２が、変更されたＳ／Ｎ比最小値ＳＮＲ_min を使用して処理を行うことにより、例えば、探知データ数ｊが著しく多い場合には、粗ゲート内の探知データ数ｊを減らすことができ、精ゲート内外判定手段５、データ更新手段６、予測手段７の各処理における計算の演算負荷が軽減されるという効果が得られる。
【０１４８】
【発明の効果】
以上のように、この発明によれば、Ｓ／Ｎ比最小値を用いて計算した探知データ数に依存しない現時刻の観測誤差共分散行列Ｒ_kmin（ｔ）により、探知データ数に依存しない残差共分散行列Ｓ_kmin（ｔ）を算出するＳ／Ｎ比最小値残差共分散行列算出手段と、残差共分散行列Ｓ_kmin（ｔ）により、目標存在期待領域である粗ゲートを算出し、入力した探知データ及びＳ／Ｎ比について、粗ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、判定した粗ゲート内に入っている探知データ及びＳ／Ｎ比を出力する粗ゲート内外判定手段と、粗ゲート内外判定手段からの出力を用いて、サンプリング時刻、探知データ数、航跡数に依存する、粗ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出する各Ｓ／Ｎ比残差共分散行列算出手段と、各Ｓ／Ｎ比残差共分散行列算出手段からの出力を用いて、粗ゲートより狭い目標存在期待領域である精ゲートを算出し、入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比について、精ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、サンプリング時刻、探知データ数、航跡数に依存する、精ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出する精ゲート内外判定手段とを備えたことにより、目標信号にとり不要な信号を除去するための精ゲートの演算負荷が軽減されるという効果がある。
【０１４９】
この発明によれば、精ゲート内外判定手段から出力される各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）の数と、観測手段が出力した探知データ数により、観測手段及びＳ／Ｎ比最小値残差共分散行列算出手段が使用しているＳ／Ｎ比最小値の値を変更し、観測手段及びＳ／Ｎ比最小値残差共分散行列算出手段に通知するＳ／Ｎ比最小値制御手段とを備え、観測手段及びＳ／Ｎ比最小値残差共分散行列算出手段が、変更されたＳ／Ｎ比最小値を使用して処理を行うことにより、例えば、探知データ数が著しく多い場合には、粗ゲート内の探知データ数を減らすことができ、演算負荷が軽くなるという効果がある。
【０１５０】
この発明によれば、レーダの最大探知距離に基づき観測誤差共分散行列成分の内の最大値成分を選択し、選択した観測誤差共分散行列の最大値成分を用いて、サンプリング時刻、探知データ数、航跡数に依存しない観測誤差共分散行列Ｒ_min を算出し、観測手段からの出力、及び算出した観測誤差共分散行列Ｒ_min を用いて、探知データ数に依存しない簡易な残差共分散行列Ｓ’_kmin（ｔ）を算出し、算出した簡易な残差共分散行列Ｓ’_kmin（ｔ）を出力するＳ／Ｎ比最小値残差共分散行列算出手段と、簡易な残差共分散行列Ｓ’_kmin（ｔ）により、目標存在期待領域である粗ゲートを算出し、入力した探知データ及びＳ／Ｎ比について、粗ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、判定した粗ゲート内に入っている探知データ及びＳ／Ｎ比を出力する粗ゲート内外判定手段と、粗ゲート内外判定手段からの出力を用いて、サンプリング時刻、探知データ数、航跡数に依存する、粗ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出して、算出した各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を出力する各Ｓ／Ｎ比残差共分散行列算出手段と、各Ｓ／Ｎ比残差共分散行列算出手段からの出力を用いて、粗ゲートより狭い目標存在期待領域である精ゲートを算出し、入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比について、精ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、サンプリング時刻、探知データ数、航跡数に依存する、精ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出する精ゲート内外判定手段とを備えたことにより、粗ゲート内外判定手段、各Ｓ／Ｎ比残差共分散行列算出手段、精ゲート内外判定手段の各処理における計算の演算負荷が軽減されるという効果がある。
【０１５１】
この発明によれば、レーダの最大探知距離に基づき観測誤差共分散行列成分の内の最大値成分を選択し、選択した観測誤差共分散行列の最大値成分を用いて、サンプリング時刻、探知データ数、航跡数に依存しない観測誤差共分散行列Ｒ_min を算出し、観測手段からの出力、及び算出した観測誤差共分散行列Ｒ_min を用いて、探知データ数、航跡数に依存しない残差共分散行列Ｅ_k を算出するＳ／Ｎ比最小値残差共分散行列算出手段と、残差共分散行列Ｅ_kにより、目標存在期待領域である粗ゲートを算出し、入力した探知データ及びＳ／Ｎ比について、粗ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、判定した粗ゲート内に入っている探知データ及びＳ／Ｎ比を出力する粗ゲート内外判定手段と、粗ゲート内外判定手段からの出力を用いて、サンプリング時刻、探知データ数、航跡数に依存する、粗ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出する各Ｓ／Ｎ比残差共分散行列算出手段と、各Ｓ／Ｎ比残差共分散行列算出手段からの出力を用いて、粗ゲートより狭い目標存在期待領域である精ゲートを算出し、入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比について、精ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、サンプリング時刻、探知データ数、航跡数に依存する、精ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出する精ゲート内外判定手段とを備えたことにより、粗ゲート内外判定手段、各Ｓ／Ｎ比残差共分散行列算出手段、精ゲート内外判定手段の各処理における計算の演算負荷が軽減されるという効果がある。
【０１５２】
この発明によれば、レーダの最大探知距離に基づき観測誤差共分散行列成分の内の最大値成分を選択する観測誤差共分散行列最大値選択手段と、観測誤差共分散行列最大値選択手段からの出力を用いて、固定のサンプリング間隔Ｔ_int により、サンプリング時刻、探知データ数、航跡数に依存しない半径ｒａｄｉｕｓ０の粗ゲートを算出し、入力した探知データ及びＳ／Ｎ比について、粗ゲート内に入っている探知データ及びＳ／Ｎ比を判定する粗ゲート内外判定手段と、粗ゲート内外判定手段からの出力を用いて、サンプリング時刻、探知データ数、航跡数に依存する、粗ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出する各Ｓ／Ｎ比残差共分散行列算出手段と、各Ｓ／Ｎ比残差共分散行列算出手段からの出力を用いて、粗ゲートより狭い目標存在期待領域である精ゲートを算出し、入力した粗ゲート内に入っている探知データ及びＳ／Ｎ比について、精ゲート内に入っている探知データ及びＳ／Ｎ比を判定し、サンプリング時刻、探知データ数、航跡数に依存する、精ゲートによる各Ｓ／Ｎ比残差共分散行列Ｓ_kj（ｔ）を算出する精ゲート内外判定手段とを備えたことにより、粗ゲート内外判定手段、各Ｓ／Ｎ比残差共分散行列算出手段、精ゲート内外判定手段の各処理における計算の演算負荷が軽減されるという効果がある。
【図面の簡単な説明】
【図１】 この発明の実施の形態１による追尾装置の構成を示すブロック図である。
【図２】 この発明の実施の形態１による追尾装置の処理の流れを示すフローチャートである。
【図３】 この発明の実施の形態２による追尾装置の構成を示すブロック図である。
【図４】 この発明の実施の形態２による追尾装置の処理の流れを示すフローチャートである。
【図５】 この発明の実施の形態３による追尾装置の構成を示すブロック図である。
【図６】 この発明の実施の形態３による追尾装置の処理の流れを示すフローチャートである。
【図７】 この発明の実施の形態４による追尾装置の構成を示すブロック図である。
【図８】 この発明の実施の形態４による追尾装置の処理の流れを示すフローチャートである。
【図９】 この発明の実施の形態５による追尾装置の構成を示すブロック図である。
【図１０】 この発明の実施の形態５による追尾装置の処理の流れを示すフローチャートである。
【図１１】 従来の追尾装置の構成を示すブロック図である。
【図１２】 従来の観測手段に係る座標系を説明する図である。
【符号の説明】
１ 観測手段、２ Ｓ／Ｎ比最小値残差共分散行列算出手段、３ 粗ゲート内外判定手段、４ 各Ｓ／Ｎ比残差共分散行列算出手段、５ 精ゲート内外判定手段、６ データ更新手段、７ 予測手段、８ 遅延手段、１２ Ｓ／Ｎ比最小値残差共分散行列算出手段、１３ 粗ゲート内外判定手段、２２ Ｓ／Ｎ比最小値残差共分散行列算出手段、２３ 粗ゲート内外判定手段、３２ 観測誤差共分散行列最大値選択手段、３３ 粗ゲート内外判定手段、４１ Ｓ／Ｎ比最小値制御手段。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a tracking device that reduces a calculation load in a high-density environment for target tracking.
[0002]
[Prior art]
FIG. 11 is shown, for example, in the formula “Samuel S. Blackman, Multiple-Target Tracking with Radar Applications, Artech House, Dedham, 1986”, p83-p92, especially p88, 4.2.2 (4.6). It is a block diagram which shows the structure of the conventional tracking device with a gate determination method.
[0003]
In FIG. 11, 1 is an observation means, 4 is each S / N ratio residual covariance matrix calculation means, 5 is a fine gate inside / outside determination means, 6 is a data update means, 7 is a prediction means, and 8 is a delay means.
[0004]
FIG. 12 is a diagram for explaining a coordinate system related to the observation means 1 in FIG. In FIG. 12, O is a sensor provided in the observation means 1, T is a tracking target, R is a distance between the tracking target T and the sensor O, E is a line segment OT connecting the sensor O and the tracking target T is X−. The elevation angle formed with the Y plane, and B is the azimuth angle formed by the orthogonal projection vector of the line segment OT connecting the sensor O and the tracking target T onto the XY plane with the X axis. Further, [R, E, B] indicate polar coordinates, and [X, Y, Z] indicate north reference orthogonal coordinates. Hereinafter, unless otherwise specified, the coordinates indicate the north reference orthogonal coordinates.
[0005]
Next, the operation will be described.
The observation means 1 uses detection data and the S / N ratio that is a signal-to-noise ratio (noise) associated with the detection data, and the S / N ratio minimum value that is a threshold value for detecting a target. Then, detection is performed from the sensor provided inside, and the detection data and the S / N ratio associated with the detection data are output. Here, the types of detection data include a target signal for a target to be tracked and other unnecessary signals. The unnecessary signals may be referred to as clutter in the future.
[0006]
Here, the detection data is D_kjDetecting data D_kjSNR of SNR_kjAnd The subscript k indicates the sampling time, and the subscript j indicates the number of detection data at the sampling time k. Detection data D_kjIs represented by the following equation (1).
[Expression 1]

[0007]
In the above equation (1), o is a subscript indicating the observed value, k is a subscript indicating the sampling time, and j is a subscript indicating the number of detected data. That is, detection data D_kjIndicates j-th detection data at the sampling time k. X_okjIs detection data D_kjX-coordinate component of y_okjIs detection data D_kjY coordinate component of z_okjIs detection data D_kjThe Z coordinate components of are respectively shown. Furthermore, the symbol “′” indicates transposition of vectors and matrices.
[0008]
Each S / N ratio residual covariance matrix calculation means 4 has detection data D obtained from the observation means 1._kjAnd detection data D_kjS / N ratio SNR associated with_kjAnd each S / N ratio residual covariance matrix S_kj(T) is calculated and output. Here, each S / N ratio residual covariance matrix is expressed as S_kjThe subscript k in (t) indicates the sampling time, and the subscript j is the jth detection data D._kj(T) indicates the track number.
[0009]
Each S / N ratio residual covariance matrix calculating means 4 is provided with each S / N ratio residual covariance matrix S._kj(T) is calculated from the following equation (2) based on the Kalman filter theory.
[Expression 2]

In the above equation (2), H is an observation matrix, P_pk(T) is a prediction error covariance matrix of the track number (t) obtained from the prediction means 7 described later via the delay means 8, R_kj(T) shows the observation error covariance matrix of the track number (t) in the north reference orthogonal coordinates.
[0010]
The observation matrix H in the above equation (2) is shown as the following equation (3).
[Equation 3]

[0011]
In addition, the observation error covariance matrix R in the north reference orthogonal coordinates in the above equation (2)_kj(T) is expressed by the following equation (4).
[Expression 4]

In the above equation (4), Λ_kjIs the observation error covariance matrix in polar coordinates and Γ_kj(T) is Λ of track number (t)_kjIs the transformation matrix.
[0012]
Observation error covariance matrix Λ in polar coordinates, shown in equation (4) above_kjIs expressed by the following equation (5).
[Equation 5]

[0013]
In the above equation (5), σr_kjIs the standard deviation of distance observation noise, σe_kjIs the standard deviation of the observation noise of the elevation angle, σb_kjIndicates the standard deviation of azimuth observation noise. Distance observation noise is 0 on average, variance σr² _kjFollow the normal distribution of. Similarly, the elevation observation noise has an average of 0 and variance σe.² _kjThe average azimuth observation noise is 0 and the variance σb² _kjFollow the normal distribution of.
[0014]
Λ of track number (t) in the above equation (4)_kjTransformation matrix Γ_kj(T) is expressed by the following equation (6).
[Formula 6]

R in the above equation (6)_pk(T), e_pk(T), b_pk(T) is a prediction vector VX obtained from the prediction means 7 described later via the delay means 8, respectively._pkPosition component x of (t)_pk(T), y_pk(T), z_pkThe polar coordinate component of (t) is shown.
[0015]
R_pk(T), e_pk(T), b_pk(T) and x_pk(T), y_pk(T), z_pkThe relationship (t) is expressed by the following equation (7).
[Expression 7]

[0016]
Standard deviation σr of distance observation noise in equation (5)_kj, Standard deviation σe of elevation noise_kj, Standard deviation σb of observation noise of azimuth_kjIs the S / N ratio SNR_kjThe following equation (8), equation (9), and equation (10) can be used.
[Equation 8]

Α in the above equations (8), (9), and (10)_r, Α_e, Α_bIs a positive constant determined from the frequency of the radar transmission signal, the bandwidth of the radar transmission signal, the effective aperture diameter of the radar antenna, and the speed of light.
[0017]
The precision gate used in the precision gate inside / outside determination means 5 is an area represented by the following equation (11).
[Equation 9]

In the above equation (11), VZ_pk(T) is the prediction vector VX_pkThe position component of (t) is shown, and the superscript “−1” in the equation (11) is a symbol indicating the inverse matrix.
[0018]
Furthermore, dε in equation (11) is a gate size parameter that determines the size of the fine gate. The gate size parameter dε is a constant, and the subscript ε of the gate size parameter dε indicates the probability that the target exists in the fine gate. The probability that a target exists in this fine gate is referred to as an in-gate probability or an in-gate target existence expected probability. Here, the gate size parameter dε indicates the size of the gate at the in-gate probability ε, and the degree of freedom and the in-gate probability ε that the user wants to set are determined in advance by the chi-square distribution referred to in statistics. Obtained from a square density function or a chi-square distribution table.
[0019]
Also, the prediction vector VX_pk(T) is expressed by the following equation (12).
[Expression 10]

In equation (12), x_pk(T), y_pk(T), z_pk(T) is the predicted position of X, Y, Z, dx_pk(T), dy_pk(T), dz_pk(T) is a predicted speed at predicted positions of X, Y, and Z.
[0020]
Prediction vector VX in equation (11) above_pkPosition component VZ of (t)_pk(T) and the prediction vector VX of the above equation (12)_pk(T) has the relationship of the following equation (13) using the observation matrix H of the above equation (3).
## EQU11 ##

[0021]
The size of the fine gate expressed by the above equation (11) is the gate size parameter dε and each S / N ratio residual covariance matrix S._kjDetermined by (t). Next, why the gate size parameter dε and the residual covariance matrix S_kjWhether the size of the fine gate is determined by (t) will be described using an example. Hereinafter, in order to simplify the explanation, a matrix A and a vector x are assumed as in the following equations (14) and (15). Then, the above equation (11) is determined as the following equation (16) from the equations (14) and (15).
[0022]
[Expression 12]

[0023]
Here, since the matrix A in the equation (14) is a positive symmetric matrix of 3 rows and 3 columns, the following equations (17) and (18) are established.
[Formula 13]

[0024]
Here, it is assumed that the matrix A can be diagonalized by a diagonal matrix B and an orthogonal matrix C by algebra as shown in the following equation (19).
[Expression 14]

[0025]
The diagonal matrix B in the above equation (19) is the eigenvalue λ of the matrix A as in the following equation (20).₁, Λ₂, Λ_ThreeIt is assumed that Here, from the relationship of the above equation (18), the eigenvalue λ₁, Λ₂, Λ_ThreeAre respectively λ₁> 0, λ₂> 0, λ_Three> 0.
[Expression 15]

[0026]
The orthogonal matrix B in the above equation (19) is the eigenvector v of the matrix A.₁, V₂, V_ThreeThe following equation (21) is used. Where v₁, V₂, V_ThreeAre vectors of 3 rows and 1 column each.
[Expression 16]

[0027]
Eigenvector v of matrix A in equation (21) above₁, V₂, V_ThreeHas the following relationship (22).
[Expression 17]

[0028]
Further, the above equation (20) is expressed as the following equation (23) using the above equation (19).
[Formula 18]

[0029]
Here, when the component of Cx in the above equation (23) is represented by the following equation (24), the above equation (23) is obtained by using the above equation (20) and the following equation (24). The following equation (25) is shown.
[Equation 19]

[0030]
Here, from the above equations (14) and (20), the eigenvalue of the matrix A is λ₁, Λ₂, Λ_ThreeSo each S / N ratio residual covariance matrix S_kjThe eigenvalue of (t) is 1 / λ₁, 1 / λ₂, 1 / λ_ThreeIt is. Therefore, the above equation (25) indicates that the radius of the error ellipse is equal to each S / N ratio residual covariance matrix S._kjEigenvalue 1 / λ of (t)₁, 1 / λ₂, 1 / λ_Three, And the gate size parameter dε. That is, the size of the fine gate is determined by each S / N ratio residual covariance matrix S._kjIt is determined by (t) and the gate size parameter dε. And each S / N ratio residual covariance matrix S_kjThe larger (t), the more the S / N ratio residual covariance matrix S_kjThe smaller the inverse matrix of (t), the larger the size of the fine gate. Furthermore, the larger the gate size parameter dε, the larger the size of the fine gate.
[0031]
Further, the prediction vector VX of the above equation (11)_pkPosition component VZ of (t)_pk(T) indicates the center of the precision gate, and the right side of the equation (11) indicates the detection data D_kjAnd the prediction vector VX which is the center of the precision gate_pkPosition component VZ of (t)_pkEach S / N ratio residual covariance matrix S of (t)_kjThe distance normalized by (t) is shown.
[0032]
Therefore, the precision gate inside / outside determination means 5 detects the detection data D._kjAnd the prediction vector VX which is the center of the precision gate_pkPosition component VZ of (t)_pkEach S / N ratio residual covariance matrix S of (t)_kjWhen the distance normalized by (t) is smaller than the gate size parameter dε, that is, the detection data D satisfying the above equation (11)_kjTarget signal candidate aD_kjOutput as. Further, the target signal candidate is aD_kjEach S / N ratio residual covariance matrix corresponding to_kjOutput as (t). Furthermore, the prediction vector VX_pk(T) Le position component VZ_pk(T) is output.
[0033]
Here, detection data D_kjIs obtained from the observation means 1 and is the prediction vector VX which is the center of the fine gate_pkPosition component VZ of (t)_pk(T) is a prediction vector VX after one sampling with respect to the current sampling time k which is an output of the prediction means 7 described later_{pk + 1}(T) is changed to VX at the current time via delay means 8 described later._pk(T) and then using the above equation (13), the prediction vector VX_pkPosition component VZ of (t)_pk(T) is calculated. Each S / N ratio residual covariance matrix S_kj(T) is obtained from each S / N ratio residual covariance matrix calculating means 4.
[0034]
The data updating means 6 is based on the Kalman filter theory, and the gain matrix K_k(T) is calculated from the following equation (26). Also, the smooth vector VX_sk(T) is the gain matrix K of equation (26)_kUsing (t), the following equation (27) is used for calculation. Furthermore, the smooth error covariance matrix P_sk(T) is the gain matrix K of equation (26)_kUsing (t), the following equation (28) is used for calculation. The data updating means 6 then calculates the smoothed vector VX calculated by the equations (27) and (28)._sk(T) and the smoothing error covariance matrix P_sk(T) is output.
[0035]
[Expression 20]

[0036]
Based on the theory of the Kalman filter, the prediction means 7 calculates the prediction vector VX at time k + 1 one sampling after the current sampling time k according to the following equation (29)._{pk + 1}(T) is calculated and output. Further, the prediction means 7 calculates the prediction error covariance matrix P at time k + 1 after one sampling from the current sampling time k according to the following equation (30)._{pk + 1}(T) is calculated and output.
[Expression 21]

[0037]
In the above equation (29), the state transition matrix Φ for performing one sampling extrapolation_kIs expressed by the following equation (31).
[Expression 22]

Δt in the equation (31) indicates the current time k and the sampling interval from the current time to time k + 1 after one sampling.
[0038]
Further, the covariance matrix Q of the drive noise in the above equation (30)_kIs expressed by the following equation (32).
[Expression 23]

Σ in the above equation (32)_aIndicates a standard deviation of driving noise and is a constant indicating the ambiguity of the target motion model of the Kalman filter.
[0039]
In the delay means 8, the prediction vector VX input from the prediction means 7 is displayed._{pk + 1}(T) and the prediction error covariance matrix P_{pk + 1}(T) is delayed by one sampling, and the prediction vector VX at the current time_pk(T) and the prediction error covariance matrix P at the current time_pk(T) is output to each S / N ratio residual covariance matrix means 4.
[0040]
In the tracking device as described above, all detection data D_kj, Each S / N ratio residual covariance matrix S in equation (11) above_kjInverse matrix calculation of (t) is performed. For example, if the number of wakes one sampling before the current time is L, the detection data D at the current time_kjIf the number of is M, the number of tracks becomes L means that the track number (t) becomes t = 1,..., L, and the detection data D_kjThe number of detection data (number of detection data j) is M._kjDetection data D_kjThat is, the subscript j indicating the number of j becomes j = 1,.
[0041]
Then, using the above equation (11) and performing precision gate processing with all these combinations, LM inverse matrix calculations are required, and the computational load on the computer becomes enormous. The meaning of the precision gate in the precision gate inside / outside determination means 5 means that the precision gate is processed in all the combinations described above.
[0042]
Here, the wake is at most one detection data D from each time._kjIt is time series data composed by selecting. There are two types of wakes: existing wakes and new wakes. The existing track is a track created by the previous time, and the new track is detection data D with the current time._kjThis is a new wake.
[0043]
[Problems to be solved by the invention]
Since the conventional tracking device is configured as described above, all the detection data D_kj, Each S / N ratio residual covariance matrix S in equation (11) above_kjWhen the inverse matrix calculation of (t) is performed, the number of wakes one sampling before the current time is L, and the detection data D at the current time_kjIf M is the number of, the LM inverse matrix calculation is required to perform precision gate processing with all of these combinations using equation (11), and the computational load of the computer becomes enormous. There was a problem.
[0044]
The present invention has been made to solve the above-described problems. In the calculation of a fine gate for removing unnecessary signals from the target signal, the target signal is captured in the fine gate without applying a calculation load. It aims at obtaining the tracking device which can be done.
[0045]
[Means for Solving the Problems]
The tracking device according to the present invention uses the S / N ratio minimum value, which is a threshold value for detecting a target, to detect detection data from an internal sensor and the S / N ratio associated with the detection data. And the observation error covariance matrix R at the current time independent of the number of detection data calculated using the S / N ratio minimum value._kminFrom (t), the residual covariance matrix S independent of the number of detected data_kmin(T) is calculated, and the calculated residual covariance matrix S_kmin(T), S / N ratio minimum value residual covariance matrix calculating means for outputting the input detection data and S / N ratio, and a residual input from the S / N ratio minimum value residual covariance matrix calculating means. Difference covariance matrix S_kminThe coarse gate which is the target existence expected area is calculated from (t), the detected data and S / N ratio contained in the coarse gate are determined for the input detection data and S / N ratio, and the determined coarse gate is determined. The coarse gate inside / outside judgment means for outputting the detection data and S / N ratio contained in the inside, and the output of the coarse gate inside / outside judgment means, the coarse gate depends on the sampling time, the number of detection data, and the number of tracks. Each S / N ratio residual covariance matrix S_kj(T) is calculated, and each calculated S / N ratio residual covariance matrix S_kj(T), each S / N ratio residual covariance matrix calculating means for outputting detection data and S / N ratio contained in the input coarse gate, and each S / N ratio residual covariance matrix calculation The output of the means is used to calculate a fine gate which is a target existence expectation area narrower than the coarse gate, and the detection data contained in the coarse gate and the S / N ratio inputted in the fine gate are detected data. Each S / N ratio residual covariance matrix S by a precision gate, which depends on the sampling time, the number of detected data, and the number of wakes._kj(T) is calculated and the detected detection data and S / N ratio in the determined gate, and each calculated S / N ratio residual covariance matrix S_kjAnd a fine gate inside / outside determination means for outputting (t).
[0046]
The tracking device according to the present invention uses the S / N ratio minimum value, which is a threshold value for detecting a target, to detect detection data from an internal sensor and the S / N ratio associated with the detection data. Is input using the observation means, the output of the observation means, the prediction error covariance matrix of the current time and the prediction vector of the current time output by the delay means described later, and the calculation is performed using the minimum value of the S / N ratio. Current observation error covariance matrix R independent of the number of detected data_kmin(T) and the input prediction error covariance matrix at the current time, the residual covariance matrix S independent of the number of detected data_kmin(T) is calculated, and the calculated residual covariance matrix S_kmin(T), S / N ratio minimum value residual covariance matrix calculating means for outputting the input detection data, the S / N ratio, the prediction error covariance matrix at the current time, and the prediction vector at the current time; Residual covariance matrix S input from N ratio minimum residual covariance matrix calculation means_kminBased on (t) and the prediction vector at the current time, a rough gate which is a target existence expectation region is calculated, and the detection data and S / N ratio contained in the rough gate are determined for the input detection data and S / N ratio. Coarse gate inside / outside determination means for outputting the detected detection data and S / N ratio contained in the determined coarse gate, the input current time prediction error covariance matrix and the current time prediction vector, and the rough gate inside / outside Each S / N ratio residual covariance matrix S by the coarse gate, which depends on the sampling time, the number of detected data, and the number of tracks, using the output of the judging means_kj(T) is calculated, and each calculated S / N ratio residual covariance matrix S_kj(T), and each S / N that outputs the detection data contained in the input coarse gate, the S / N ratio contained in the coarse gate, the prediction error covariance matrix at the current time, and the prediction vector at the current time By using the output of the ratio residual covariance matrix calculation means and the output of each of the S / N ratio residual covariance matrix calculation means, a fine gate that is a target existence expectation area narrower than the coarse gate is calculated, and the input within the input coarse gate The detection data and S / N ratio in the fine gate are determined with respect to the detection data and S / N ratio that are included, and each S / N by the fine gate depends on the sampling time, the number of detection data, and the number of tracks. N-ratio residual covariance matrix S_kj(T) is calculated and the detected data and S / N ratio included in the determined fine gate, and each calculated S / N ratio residual covariance matrix S_kj(T), and a fine gate inside / outside determination means for outputting the input current prediction error covariance matrix and the current time prediction vector, and the output of the fine gate inside / outside determination means, the smooth error covariance at the current time Data updating means for calculating and outputting a matrix and a smooth vector at the current time, and using the output of the data updating means, a prediction error covariance matrix after one sampling from the current time and a prediction vector after one sampling from the current time And a delay means for outputting a prediction error covariance matrix at the current time and a prediction vector at the current time using the output of the prediction means.
[0047]
In the tracking device according to the present invention, each S / N ratio residual covariance matrix S output from the precision gate inside / outside determining means is provided._kjThe value of the S / N ratio minimum value used by the observation means and the S / N ratio minimum value residual covariance matrix calculation means is changed according to the number of (t) and the number of detection data output by the observation means. S / N ratio minimum value control means for notifying the observation means and the S / N ratio minimum value residual covariance matrix calculating means, and the observation means and the S / N ratio minimum value residual covariance matrix. The calculation means performs processing using the changed S / N ratio minimum value.
[0048]
The tracking device according to the present invention uses the S / N ratio minimum value, which is a threshold value for detecting a target, to detect detection data from an internal sensor and the S / N ratio associated with the detection data. Is selected from the observation error covariance matrix components based on the maximum detection distance of the radar and the sampling time and detection data using the maximum value components of the selected observation error covariance matrix. Observation error covariance matrix R independent of number and number of tracks_min, The output from the observation means, and the calculated observation error covariance matrix R_minAnd a simple residual covariance matrix S ′ independent of the number of detected data_kmin(T) is calculated, and the calculated simple residual covariance matrix S ′_kmin(T), S / N ratio minimum value residual covariance matrix calculating means for outputting the input detection data and S / N ratio, and simple input from the S / N ratio minimum value residual covariance matrix calculating means Residual covariance matrix S '_kminThe coarse gate which is the target existence expected area is calculated from (t), the detected data and S / N ratio contained in the coarse gate are determined for the input detection data and S / N ratio, and the determined coarse gate is determined. The coarse gate inside / outside judgment means for outputting the detection data and S / N ratio contained in the inside, and the output of the coarse gate inside / outside judgment means, the coarse gate depends on the sampling time, the number of detection data, and the number of tracks. Each S / N ratio residual covariance matrix S_kj(T) is calculated, and each calculated S / N ratio residual covariance matrix S_kj(T), each S / N ratio residual covariance matrix calculating means for outputting detection data and S / N ratio contained in the input coarse gate, and each S / N ratio residual covariance matrix calculation The output of the means is used to calculate a fine gate which is a target existence expectation area narrower than the coarse gate, and the detection data contained in the coarse gate and the S / N ratio inputted in the fine gate are detected data. Each S / N ratio residual covariance matrix S by a precision gate, which depends on the sampling time, the number of detected data, and the number of wakes._kj(T) is calculated and the detected detection data and S / N ratio in the determined gate, and each calculated S / N ratio residual covariance matrix S_kjAnd a fine gate inside / outside determination means for outputting (t).
[0049]
The tracking device according to the present invention uses the S / N ratio minimum value, which is a threshold value for detecting a target, to detect detection data from an internal sensor and the S / N ratio associated with the detection data. Is selected from the observation error covariance matrix components based on the maximum detection distance of the radar and the sampling time and detection data using the maximum value components of the selected observation error covariance matrix. Observation error covariance matrix R independent of number and number of tracks_min, The output from the observation means, the current time prediction error covariance matrix and the current time prediction vector from the delay means described later, and the calculated observation error covariance matrix R_minAnd a simple residual covariance matrix S ′ independent of the number of detected data_kmin(T) is calculated, and the calculated simple residual covariance matrix S ′_kmin(T), S / N ratio minimum value residual covariance matrix calculating means for outputting the input detection data, the S / N ratio, the prediction error covariance matrix at the current time, and the prediction vector at the current time; Simple residual covariance matrix S ′ input from the N ratio minimum value residual covariance matrix calculation means_kminBased on (t) and the prediction vector at the current time, a rough gate which is a target existence expectation region is calculated, and the detection data and S / N ratio contained in the rough gate are determined for the input detection data and S / N ratio. Coarse gate inside / outside determination means for outputting the detected detection data and S / N ratio contained in the determined coarse gate, the input current time prediction error covariance matrix and the current time prediction vector, and the rough gate inside / outside Each S / N ratio residual covariance matrix S by the coarse gate, which depends on the sampling time, the number of detected data, and the number of tracks, using the output of the judging means_kj(T) is calculated, and each calculated S / N ratio residual covariance matrix S_kj(T), and each S / N that outputs the detection data contained in the input coarse gate, the S / N ratio contained in the coarse gate, the prediction error covariance matrix at the current time, and the prediction vector at the current time By using the output of the ratio residual covariance matrix calculation means and the output of each of the S / N ratio residual covariance matrix calculation means, a fine gate that is a target existence expectation area narrower than the coarse gate is calculated, and the input within the input coarse gate The detection data and S / N ratio in the fine gate are determined with respect to the detection data and S / N ratio that are included, and each S / N by the fine gate depends on the sampling time, the number of detection data, and the number of tracks. N-ratio residual covariance matrix S_kj(T) is calculated and the detected data and S / N ratio included in the determined fine gate, and each calculated S / N ratio residual covariance matrix S_kj(T), and a fine gate inside / outside determination means for outputting the input current prediction error covariance matrix and the current time prediction vector, and the output of the fine gate inside / outside determination means, the smooth error covariance at the current time Data updating means for calculating and outputting a matrix and a smooth vector at the current time, and using the output of the data updating means, a prediction error covariance matrix after one sampling from the current time and a prediction vector after one sampling from the current time And a delay means for outputting a prediction error covariance matrix at the current time and a prediction vector at the current time using the output of the prediction means.
[0050]
The tracking device according to the present invention uses the S / N ratio minimum value, which is a threshold value for detecting a target, to detect detection data from an internal sensor and the S / N ratio associated with the detection data. Is selected from the observation error covariance matrix components based on the maximum detection distance of the radar and the sampling time and detection data using the maximum value components of the selected observation error covariance matrix. Observation error covariance matrix R independent of number and number of tracks_min, The output from the observation means, and the calculated observation error covariance matrix R_min, The residual covariance matrix E independent of the number of detected data and the number of tracks_kAnd the calculated residual covariance matrix E_kS / N ratio minimum value residual covariance matrix calculating means for outputting input detection data and S / N ratio, and residual covariance input from the S / N ratio minimum value residual covariance matrix calculating means Matrix E_kTo calculate the rough gate which is the target existence expected region, and for the input detection data and S / N ratio, determine the detection data and S / N ratio that are in the rough gate, and enter the determined rough gate. The coarse gate inside / outside determination means for outputting the detected detection data and the S / N ratio, and the output of the coarse gate inside / outside determination means, each S / R by the coarse gate depending on the sampling time, the number of detection data, and the number of wakes. N-ratio residual covariance matrix S_kj(T) is calculated, and each calculated S / N ratio residual covariance matrix S_kj(T), each S / N ratio residual covariance matrix calculating means for outputting detection data and S / N ratio contained in the input coarse gate, and each S / N ratio residual covariance matrix calculation The output of the means is used to calculate a fine gate which is a target existence expectation area narrower than the coarse gate, and the detection data contained in the coarse gate and the S / N ratio inputted in the fine gate are detected data. Each S / N ratio residual covariance matrix S by a precision gate, which depends on the sampling time, the number of detected data, and the number of wakes._kj(T) is calculated and the detected detection data and S / N ratio in the determined gate, and each calculated S / N ratio residual covariance matrix S_kjAnd a fine gate inside / outside determination means for outputting (t).
[0051]
The tracking device according to the present invention uses the S / N ratio minimum value, which is a threshold value for detecting a target, to detect detection data from an internal sensor and the S / N ratio associated with the detection data. Is selected from the observation error covariance matrix components based on the maximum detection distance of the radar and the sampling time and detection data using the maximum value components of the selected observation error covariance matrix. Observation error covariance matrix R independent of number and number of tracks_min, The output from the observation means, the current time prediction error covariance matrix and the current time prediction vector from the delay means described later, and the calculated observation error covariance matrix R_min, The residual covariance matrix E independent of the number of detected data and the number of tracks_kAnd the calculated residual covariance matrix E_kS / N ratio minimum value residual covariance matrix calculating means for outputting input detection data, S / N ratio, current time prediction error covariance matrix and current time prediction vector, and the above S / N ratio minimum Residual covariance matrix E input from value residual covariance matrix calculation means_kThen, based on the prediction vector at the current time, a rough gate that is a target existence expectation region is calculated, and the detection data and S / N ratio that are contained in the rough gate are determined for the input detection data and S / N ratio. Coarse gate inside / outside determination means for outputting the detection data and S / N ratio contained in the rough gate, the input current time prediction error covariance matrix and the current time prediction vector, and the rough gate inside / outside determination means Using the output, each S / N ratio residual covariance matrix S by the coarse gate depends on the sampling time, the number of detected data, and the number of tracks._kj(T) is calculated, and each calculated S / N ratio residual covariance matrix S_kj(T), and each S / N that outputs the detection data contained in the input coarse gate, the S / N ratio contained in the coarse gate, the prediction error covariance matrix at the current time, and the prediction vector at the current time By using the output of the ratio residual covariance matrix calculation means and the output of each of the S / N ratio residual covariance matrix calculation means, a fine gate that is a target existence expectation area narrower than the coarse gate is calculated, and the input within the input coarse gate The detection data and S / N ratio in the fine gate are determined with respect to the detection data and S / N ratio that are included, and each S / N by the fine gate depends on the sampling time, the number of detection data, and the number of tracks. N-ratio residual covariance matrix S_kj(T) is calculated and the detected data and S / N ratio included in the determined fine gate, and each calculated S / N ratio residual covariance matrix S_kj(T), and a fine gate inside / outside determination means for outputting the input current prediction error covariance matrix and the current time prediction vector, and the output of the fine gate inside / outside determination means, the smooth error covariance at the current time Data updating means for calculating and outputting a matrix and a smooth vector at the current time, and using the output of the data updating means, a prediction error covariance matrix after one sampling from the current time and a prediction vector after one sampling from the current time And a delay means for outputting a prediction error covariance matrix at the current time and a prediction vector at the current time using the output of the prediction means.
[0052]
The tracking device according to the present invention uses the S / N ratio minimum value, which is a threshold value for detecting a target, to detect detection data from an internal sensor and the S / N ratio associated with the detection data. Is selected from the observation error covariance matrix components based on the maximum detection distance of the radar, the maximum value component of the selected observation error covariance matrix, and the detection input from the observation means An observation error covariance matrix maximum value selection means for outputting data and an S / N ratio, and an output of the observation error covariance matrix maximum value selection means, and a fixed sampling interval T_intTo calculate a coarse gate with radius radius0 independent of sampling time, number of detected data, and number of tracks, and determine the detected data and S / N ratio in the coarse gate for the input detected data and S / N ratio Then, using the output of the rough gate inside / outside determination means for outputting the detected detection data and S / N ratio contained in the determined coarse gate, and the output of the rough gate inside / outside determination means, the sampling time, the number of detection data, and the number of tracks are calculated. Dependent S / N ratio residual covariance matrix S with coarse gate_kj(T) is calculated, and each calculated S / N ratio residual covariance matrix S_kj(T), each S / N ratio residual covariance matrix calculating means for outputting detection data and S / N ratio contained in the input coarse gate, and each S / N ratio residual covariance matrix calculation The output of the means is used to calculate a fine gate which is a target existence expectation area narrower than the coarse gate, and the detection data contained in the coarse gate and the S / N ratio inputted in the fine gate are detected data. Each S / N ratio residual covariance matrix S by a precision gate, which depends on the sampling time, the number of detected data, and the number of wakes._kj(T) is calculated and the detected detection data and S / N ratio in the determined gate, and each calculated S / N ratio residual covariance matrix S_kjAnd a fine gate inside / outside determination means for outputting (t).
[0053]
The tracking device according to the present invention uses the S / N ratio minimum value, which is a threshold value for detecting a target, to detect detection data from an internal sensor and the S / N ratio associated with the detection data. And the maximum value component of the observation error covariance matrix component based on the maximum detection distance of the radar, the maximum value component of the selected observation error covariance matrix, and the detection data input from the above observation means And an S / N ratio, and an observation error covariance matrix maximum value selection means for outputting a prediction error covariance matrix and a prediction vector at the current time from a delay means described later, and the observation error covariance matrix maximum value selection. Using the output of the means, a fixed sampling interval T_intTo calculate a coarse gate with radius radius0 independent of sampling time, number of detected data, and number of tracks, and determine the detected data and S / N ratio in the coarse gate for the input detected data and S / N ratio Coarse gate inside / outside determination means for outputting the detected detection data and S / N ratio contained in the determined coarse gate, the input current time prediction error covariance matrix and the current time prediction vector, and the rough gate inside / outside Each S / N ratio residual covariance matrix S by the coarse gate, which depends on the sampling time, the number of detected data, and the number of tracks, using the output of the judging means_kj(T) is calculated, and each calculated S / N ratio residual covariance matrix S_kj(T), and each S / N that outputs the detection data contained in the input coarse gate, the S / N ratio contained in the coarse gate, the prediction error covariance matrix at the current time, and the prediction vector at the current time By using the output of the ratio residual covariance matrix calculation means and the output of each of the S / N ratio residual covariance matrix calculation means, a fine gate that is a target existence expectation area narrower than the coarse gate is calculated, and the input within the input coarse gate The detection data and S / N ratio in the fine gate are determined with respect to the detection data and S / N ratio that are included, and each S / N by the fine gate depends on the sampling time, the number of detection data, and the number of tracks. N-ratio residual covariance matrix S_kj(T) is calculated and the detected data and S / N ratio included in the determined fine gate, and each calculated S / N ratio residual covariance matrix S_kj(T), and a fine gate inside / outside determination means for outputting the input current prediction error covariance matrix and the current time prediction vector, and the output of the fine gate inside / outside determination means, the smooth error covariance at the current time Data updating means for calculating and outputting a matrix and a smooth vector at the current time, and using the output of the data updating means, a prediction error covariance matrix after one sampling from the current time and a prediction vector after one sampling from the current time And a delay means for outputting a prediction error covariance matrix at the current time and a prediction vector at the current time using the output of the prediction means.
[0054]
DETAILED DESCRIPTION OF THE INVENTION
An embodiment of the present invention will be described below.
Embodiment 1 FIG.
FIG. 1 is a block diagram showing the configuration of a tracking device according to Embodiment 1 of the present invention. In FIG. 1, reference numeral 2 denotes detection data D output by the observation means 1._kjAnd detection data D_kjS / N ratio SNR associated with_kj, And the prediction error covariance matrix P of the current time output from the delay means 8_pk(T) and prediction vector VX at the current time_pkS / N ratio minimum value SNR independent of sampling time k, number of detected data j, number of tracks, which are threshold values for detecting a target by inputting (t)_minThe observation error covariance matrix R at the current time that does not depend on the number of detected data j, calculated using_kmin(T) and the input current prediction error covariance matrix P_pk(T) and the residual covariance matrix S independent of the number of detected data j_kmin(T) is calculated, and the calculated residual covariance matrix S_kmin(T) and input detection data D_kjDetecting data D_kjS / N ratio SNR associated with_kj, Prediction error covariance matrix P at the current time_pk(T) and prediction vector VX at the current time_pkS / N ratio minimum value residual covariance matrix calculating means for outputting (t).
[0055]
In FIG. 1, reference numeral 3 denotes a residual covariance matrix S from the S / N ratio minimum value residual covariance matrix calculation means 2._kmin(T), detection data D_kjDetecting data D_kjS / N ratio SNR associated with_kj, Prediction error covariance matrix P at current time_pk(T) and prediction vector VX at the current time_pk(T) is input, and the input residual covariance matrix S_kmin(T) and prediction vector VX at the current time_pkFrom (t), a rough gate which is a target existence expectation area is calculated, and the input detection data D_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetected data D in the coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetection data D in the determined coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kj, And the input prediction error covariance matrix P_pk(T) and prediction vector VX at the current time_pkThis is a coarse gate inside / outside determination means for outputting (t), and the other configuration is the same as that shown in FIG.
[0056]
Next, the S / N ratio minimum value residual covariance matrix calculating means 2 will be described in detail.
In the signal detection in the radar signal processing of the observation means 1, it is determined that the target has been detected with a signal having a certain level of power with respect to the noise power δ. The S / N ratio, which is a threshold value for determining that this target has been detected, is a constant S / N minimum value SNR._minAnd That is, δ · SNR_minSuppose that the signal of the above electric power was detected. At this time, detection data D_kjS / N ratio SNR_kjSatisfies the following equation (33).
[Expression 24]

[0057]
Also, the S / N minimum value SNR_minThe observation noise of the distance, elevation angle, and azimuth calculated from_min, Σe_min, Σb_minThen σr_min, Σe_min, Σb_minFrom the above formulas (8), (9), and (10), the following formulas (34), (35), and (36) are obtained.
[Expression 25]

[0058]
From the above expressions (33), (34), (35), and (36), the following expressions (37), (38), and (39) are established.
[Equation 26]

[0059]
Further, from the above equations (34), (35), and (36), the S / N ratio minimum value SNR._minObservation error covariance matrix Λ in polar coordinates calculated from_minIs expressed by the following equation (40) from the above equation (5).
[Expression 27]

[0060]
Here, when the equation (5) is compared with the equation (40), the following equation (41) is established from the equations (37), (38), and (39).
[Expression 28]

The above equation (41) is expressed as Λ expressed in polar coordinates._min−Λ_kjIs a positive symmetric matrix by algebra.
[0061]
Furthermore, the S / N ratio minimum value SNR_minFrom the result of the equation (41) calculated from the observation error covariance matrix R in the north reference orthogonal coordinates, which does not depend on the number of detected data j_kmin(T) becomes the following equation (42) from the above equation (4).
[Expression 29]

[0062]
From the above equations (41), (4), and (42), the relationship of the following equation (43) is obtained.
[30]

[0063]
Furthermore, the S / N ratio minimum residual covariance matrix S independent of the number of detected data j_kmin(T) is calculated by the following equation (44).
[31]

[0064]
Since the above formulas (2) and (44) are algebraic positive symmetric matrices, comparing the formulas (2) and (44), the following (45) The formula is obtained.
[Expression 32]

The result of the above equation (45) is the detection data D_kjS / N ratio SNR_kjDetection data D in the precision gate calculated from_kjIs the S / N ratio minimum value SNR_minThis means that it always exists in the coarse gate calculated from
[0065]
Therefore, the S / N ratio minimum value residual covariance matrix calculation means 2 uses the above-mentioned formulas (34), (35), (36), (40), (42), and (44). S / N ratio minimum value SNR_minS / N ratio minimum residual covariance matrix S independent of the number of detected data j_kmin(T) is calculated and output.
[0066]
Next, the rough gate inside / outside determination means 3 will be described in detail.
In the coarse gate inside / outside determination means 3, the S / N ratio minimum value SNR calculated by the S / N ratio minimum value residual covariance matrix calculation means 2 is used._minS / N ratio minimum residual covariance matrix S independent of the number of detected data j_kminUsing (t), coarse gate determination is performed according to the following equation (46), and detection data D satisfying equation (46)_kjAnd detection data D_kjS / N ratio SNR associated with_kjIs output.
[Expression 33]

[0067]
Therefore, in the conventional tracking device, all the detection data D are calculated using the above equation (11)._kj, Each S / N ratio residual covariance matrix S in equation (11)_kjSince the inverse matrix calculation of (t) is to be performed, the computational load on the computer is enormous. However, as described above, when the number of wakes one sampling before the current time is L and the number of detected data at the current time is M, the S / N ratio minimum residual covariance matrix S in the above equation (46)._kminSince (t) depends on the sampling time k and the number of tracks, but does not depend on the number of detected data j, it is only necessary to calculate L inverse matrices. In addition, after the rough gate determination according to the equation (46), not only the number of detected data j can be reduced, but also the number of wakes is reduced. The calculation load is reduced.
[0068]
Next, the operation will be described.
FIG. 2 is a flowchart showing a process flow of the tracking device according to the first embodiment of the present invention. In step ST1, the observation means 1 is the threshold value for detecting the target, which is the S / N ratio minimum value SNR independent of the sampling time k, the number of detected data j, and the number of tracks._minDetected data D detected from the sensor provided inside_kjAnd detection data D_kjS / N ratio SNR associated with_kjIs output to the S / N ratio minimum value residual covariance matrix calculating means 2.
[0069]
In step ST2, the S / N ratio minimum residual covariance matrix calculation means 2 detects the detection data D output by the observation means 1._kjAnd detection data D_kjS / N ratio SNR associated with_kj, And the prediction error covariance matrix P of the current time output from the delay means 8_pk(T) and prediction vector VX at the current time_pk(T) is input, and the S / N ratio minimum value SNR_minThe observation error covariance matrix R at the current time that does not depend on the number of detected data j, calculated using_kmin(T) and the input current prediction error covariance matrix P_pk(T) and the residual covariance matrix S independent of the number of detected data j_kmin(T) is calculated, and the calculated residual covariance matrix S_kmin(T) and input detection data D_kjDetecting data D_kjS / N ratio SNR associated with_kj, Prediction error covariance matrix P at the current time_pk(T) and prediction vector VX at the current time_pk(T) is output to the rough gate inside / outside determination means 3.
[0070]
In step ST3, the coarse gate inside / outside determination means 3 inputs the residual covariance matrix S_kmin(T) and prediction vector VX at the current time_pkFrom (t), a rough gate which is a target existence expectation area is calculated, and the input detection data D_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetected data D in the coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetection data D in the determined coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kj, And the input prediction error covariance matrix P_pk(T) and prediction vector VX at the current time_pk(T) is output to each S / N ratio residual covariance matrix calculating means 4.
[0071]
In step ST4, each S / N ratio residual covariance matrix calculation means 4 receives the detection data D contained in the input coarse gate._kjS / N ratio SNR associated with_kjThe observation error covariance matrix R at the current time depending on the sampling time k, the number of detected data j, and the number of tracks calculated using_kj(T) and the input current prediction error covariance matrix P_pk(T) and each S / N ratio residual covariance matrix S by the coarse gate depending on the sampling time k, the number of detected data j, and the number of tracks._kj(T) is calculated, and each calculated S / N ratio residual covariance matrix S_kj(T) and the detected data D contained in the input coarse gate_kjDetection data D in the rough gate_kjS / N ratio SNR associated with_kj, Prediction error covariance matrix P at the current time_pk(T) and prediction vector VX at the current time_pk(T) is output to the precision gate inside / outside determination means 5.
[0072]
In step ST5, the precision gate inside / outside determination means 5 inputs each input S / N ratio residual covariance matrix S._kj(T) and prediction vector VX at the current time_pkFrom (t), a fine gate which is a target existence expectation area narrower than the coarse gate is calculated, and the detected data D contained in the inputted coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetecting data D in the precision gate_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetecting data D in the precision gate_kjS / N ratio SNR associated with_kjThe observation error covariance matrix R at the current time depending on the sampling time k, the number of detected data j, and the number of tracks calculated using_kj(T) and the input current prediction error covariance matrix P_pk(T) and each S / N ratio residual covariance matrix S by the precision gate depending on the sampling time k, the number of detected data j, and the number of tracks._kj(T) is calculated and detected data D in the determined precision gate_kjAnd detection data D_kjS / N ratio SNR associated with_kj, Each calculated S / N ratio residual covariance matrix S_kj(T) and the input prediction error covariance matrix P at the current time_pk(T) and prediction vector VX at the current time_pk(T) is output to the data updating means 6.
[0073]
In step ST6, the data update means 6 detects the detection data D contained in the input precision gate._kjS / N ratio SNR associated with_kj, Each S / N ratio residual covariance matrix S_kj(T) and the prediction error covariance matrix P at the current time_pkThe smoothing error covariance matrix P at the current time as smoothing information by (t)_sk(T) is calculated and the detected data D contained in the input precision gate_kjAnd the prediction vector VX of the current time_pkThe smoothing vector VX at the current time which is smoothing information by (t)_sk(T) is calculated, and the calculated smoothing error covariance matrix P at the current time_sk(T), smooth vector VX at the current time_sk(T) is output to the prediction means 7.
[0074]
In step ST7, the prediction means 7 receives the input smoothing error covariance matrix P at the current time._skFrom (t), a prediction error covariance matrix P after one sampling from the current time as prediction information_{pk + 1}(T) is calculated and the input smoothing vector VX at the current time_skBy (t), the prediction vector VX after one sampling from the current time which is the prediction information_{pk + 1}(T) is calculated.
[0075]
In step ST8, when the processing is continued by an operator instruction or automatic determination, a prediction error covariance matrix P after one sampling from the current time calculated by the prediction means 7 is used._{pk + 1}(T), prediction vector VX after one sampling from the current time_{pk + 1}(T) is output to the delay means 8.
[0076]
In step ST9, the delay means 8 uses the prediction error covariance matrix P after one sampling from the input current time._{pk + 1}(T) is delayed by one sampling, and the prediction error covariance matrix P at the current time_pk(T) is output to the S / N ratio minimum residual covariance matrix calculating means 2 and the prediction vector VX after one sampling from the input current time_{pk + 1}(T) is delayed by one sampling, and the prediction vector VX at the current time_pk(T) is output to the S / N ratio minimum value residual covariance matrix calculating means 2, and the processing from step ST1 is repeated.
[0077]
As described above, according to the first embodiment, the coarse gate inside / outside determination means 5 performs the residual covariance independent of the number of detected data j calculated by the S / N ratio minimum value residual covariance matrix calculation means 2. Matrix S_kminFrom (t), a rough gate which is a target existence expectation area is calculated, and the input detection data D_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetected data D in the coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kjThe fine gate inside / outside determination means 5 calculates a fine gate which is a target existence expectation area narrower than the coarse gate, and the detection data D contained in the inputted coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetecting data D in the precision gate_kjAnd detection data D_kjS / N ratio SNR associated with_kjThis reduces the calculation load of the precision gate for removing unnecessary signals for the target signal in the precision gate inside / outside judgment means 5 and the calculation in each process of the data update means 6 and the prediction means 7. The effect that the calculation load is also reduced is obtained.
[0078]
Embodiment 2. FIG.
FIG. 3 is a block diagram showing the configuration of the tracking device according to Embodiment 2 of the present invention. In the figure, reference numeral 12 denotes detection data D output by the observation means 1._kjAnd detection data D_kjS / N ratio SNR associated with_kj, And the prediction error covariance matrix P of the current time output from the delay means 8_pk(T) and prediction vector VX at the current time_pk(T) is input, and the maximum detection distance RANGE of the radar_maxThe observation error covariance matrix R_kjThe maximum component of the component is selected, and the selected observation error covariance matrix R_kjThe observation error covariance matrix R in the north reference orthogonal coordinates independent of the sampling time k, the number of detected data j, and the number of tracks due to the maximum component of the component_minAnd the calculated observation error covariance matrix R_minAnd the input prediction error covariance matrix P_pk(T) and a simple residual covariance matrix S ′ that does not depend on the number of detected data j_kmin(T) is calculated, and the calculated simple residual covariance matrix S ′_kmin(T) and input detection data D_kjDetecting data D_kjS / N ratio SNR associated with_kj, Prediction error covariance matrix P at the current time_pk(T) and prediction vector VX at the current time_pkS / N ratio minimum value residual covariance matrix calculating means for outputting (t).
[0079]
In FIG. 3, reference numeral 13 denotes a simple residual covariance matrix S ′ from the S / N ratio minimum value residual covariance matrix calculation means 12._kmin(T), detection data D_kjDetecting data D_kjS / N ratio SNR associated with_kj, Prediction error covariance matrix P at the current time_pk(T) and prediction vector VX at the current time_pk(T) is input, and the input simple residual covariance matrix S ′_kmin(T) and the current time prediction vector VX_pkFrom (t), a rough gate which is a target existence expectation area is calculated, and the input detection data D_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetected data D in the coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetection data D in the determined coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kj, And the input prediction error covariance matrix P_pk(T) and prediction vector VX at the current time_pkThis is a rough gate inside / outside determination means for outputting (t), and other configurations are the same as those in FIG.
[0080]
Next, the S / N ratio minimum value residual covariance matrix calculating means 12 will be described in detail.
Here, the radar detects the maximum detection distance RANGE to be processed._maxSince the value of is known, the observation error covariance matrix R is expressed by the following equation (47)._kjThe maximum component of the component can be defined.
[Expression 34]

In the equation (47), σcr_min, Σce_min, Σcb_minIs a positive constant. Also, maxselect (A, B, C) is A, B, C,. . . Is a function that selects one of the maximum values.
[0081]
Here, as shown in the following equation (48), the observation error covariance matrix R in the north reference orthogonal coordinates that does not depend on the sampling time k, the number of detected data j, and the number of wakes._minIs defined, the observation error covariance matrix R that does not depend on the number of detected data j in the above equation (44)._kmin(T) has the relationship of the following equation (49).
[Expression 35]

[0082]
Therefore, R in the above equation (44)_kmin(T) is R in the above equation (48)_minIs replaced by the following equation (50).
[Expression 36]

Prediction error covariance matrix P in equation (50) above_pkSince (t) has a different value for each track number (t), HP in equation (50)_pk(T) H ′ and residual covariance matrix S_min(T) also has a different value for each track number (t).
[0083]
Observation error covariance matrix R_minFrom the simple residual covariance matrix S ′ calculated using the above equation (50)_kmin(T) and the S / N ratio minimum value SNR at each sampling time k_minTo the residual covariance matrix S calculated sequentially using the above equation (44)._kminThe relationship (t) is expressed by the following equation (51).
[Expression 37]

[0084]
The above equation (51) is the observation error covariance matrix R_minTo a simple residual covariance matrix S ′ calculated using the equation (50)_kminThe coarse gate calculated by (t) is the S / N ratio minimum value SNR at each sampling time k._minTo S / N ratio minimum value residual covariance matrix S calculated sequentially using equation (44)_kminIt shows that it is the same or wider than the coarse gate calculated by (t).
[0085]
Therefore, the S / N ratio minimum value residual covariance matrix calculating means 12 uses the observation error covariance matrix R calculated using the above equation (48)._minIs used. That is, the observation error covariance matrix R in the above equation (44)._kminSince (t) depends on the number of detected data j and the number of tracks, it is necessary to perform sequential calculation, whereas the observation error covariance matrix R in the above equation (50)._minIs sampling time k, detection data D_kjSince the constant does not depend on the wake (t), the calculation load of the computer when the equation (50) is used is further reduced as compared with the calculation load of the computer when the equation (44) is used.
[0086]
Next, the operation will be described.
FIG. 4 is a flowchart showing a process flow of the tracking device according to the second embodiment of the present invention. Step ST1 is the same as that in FIG. 2 of the first embodiment.
[0087]
In step ST11, the S / N ratio minimum residual covariance matrix calculation means 12 performs the radar maximum detection distance RANGE._maxBased on the observation error covariance matrix R using the above equation (47)_kjError variance σcr in distance direction of components² _min, Elevation error variance RANGE² _maxσce_min, Azimuthal error variance RANGE² _maxσcb_minOf which observation error covariance matrix R_kjThe maximum value component of the components is selected.
[0088]
In step ST12, the S / N ratio minimum value residual covariance matrix calculating means 12 selects the selected observation error covariance matrix R._kjThe observation error covariance matrix R in the north reference orthogonal coordinates independent of the sampling time k, the number of detected data j, and the number of wakes, using the maximum component of the component_minIs calculated using equation (48).
[0089]
In step ST13, the S / N ratio minimum residual covariance matrix calculating means 12 detects the detection data D output from the observation means 1._kjAnd detection data D_kjS / N ratio SNR associated with_kj, And the prediction error covariance matrix P of the current time output from the delay means 8_pk(T) and prediction vector VX at the current time_pk(T) is input, and the calculated observation error covariance matrix R_minAnd the input prediction error covariance matrix P_pk(T) and a simple residual covariance matrix S ′ that does not depend on the number of detected data j_kmin(T) is calculated, and the calculated simple residual covariance matrix S ′_kmin(T) and input detection data D_kjDetecting data D_kjS / N ratio SNR associated with_kj, Prediction error covariance matrix P at the current time_pk(T) and prediction vector VX at the current time_pk(T) is output to the rough gate inside / outside determination means 13.
[0090]
In step ST14, the rough gate inside / outside determination means 13 inputs the input simple residual covariance matrix S '._kmin(T) and prediction vector VX at the current time_pkFrom (t), a rough gate which is a target existence expectation area is calculated, and the input detection data D_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetected data D in the coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetection data D in the determined coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kj, And the input prediction error covariance matrix P_pk(T) and prediction vector VX at the current time_pk(T) is output to each S / N ratio residual covariance matrix calculating means 4.
[0091]
The processing from step ST4 to ST9 is the same as that in FIG. 2 of the first embodiment. Therefore, according to the above steps, the calculation load of the computer when using the above equation (50) is higher than the calculation load of the computer when using the above equation (44), the sampling time k, the number of detected data j, the track Number-independent measurement error covariance matrix R_minBecause it is used, it becomes even lighter.
[0092]
As described above, according to the second embodiment, the S / N ratio minimum value residual covariance matrix calculation means 12 performs the observation error covariance matrix R independent of the sampling time k, the number of detected data j, and the number of tracks._minAnd the calculated observation error covariance matrix R_minAnd a simple residual covariance matrix S ′ that does not depend on the number of detected data j_kminSince (t) is calculated, compared with the first embodiment, coarse gate inside / outside determination means 13, each S / N ratio residual covariance matrix calculation means 4, fine gate inside / outside determination means 5, data update means 6, and prediction means 7 can reduce the calculation load in each process.
[0093]
Embodiment 3 FIG.
FIG. 5 is a block diagram showing the configuration of the tracking device according to Embodiment 3 of the present invention. In the figure, reference numeral 22 denotes detection data D output by the observation means 1._kjAnd detection data D_kjS / N ratio SNR associated with_kj, And the prediction error covariance matrix P of the current time output from the delay means 8_pk(T) and prediction vector VX at the current time_pk(T) is input, and the maximum detection distance RANGE of the radar_maxThe observation error covariance matrix R_kjThe maximum component of the component is selected, and the selected observation error covariance matrix R_kjThe observation error covariance matrix R in the north reference orthogonal coordinates independent of the sampling time k, the number of detected data j, and the number of tracks due to the maximum component of the component_minAnd the calculated observation error covariance matrix R_minAnd the input prediction error covariance matrix P_pk(T) and the residual covariance matrix E independent of the number of detected data j and the number of tracks_kAnd the calculated residual covariance matrix E_k, And input detection data D_kjDetecting data D_kjS / N ratio SNR associated with_kj, Prediction error covariance matrix P at the current time_pk(T) and prediction vector VX at the current time_pkS / N ratio minimum value residual covariance matrix calculating means for outputting (t).
[0094]
In FIG. 5, reference numeral 23 denotes a residual covariance matrix E from the S / N ratio minimum value residual covariance matrix calculation means 22._kDetecting data D_kjDetecting data D_kjS / N ratio SNR associated with_kj, Prediction error covariance matrix P at the current time_pk(T) and prediction vector VX at the current time_pk(T) is input, and the input residual covariance matrix E_kAnd the prediction vector VX of the current time_pkFrom (t), a rough gate which is a target existence expectation area is calculated, and the input detection data D_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetected data D in the coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetection data D in the determined coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kj, And the input prediction error covariance matrix P_pk(T) and prediction vector VX at the current time_pkThis is a rough gate inside / outside determination means for outputting (t), and other configurations are the same as those in FIG.
[0095]
Next, the S / N ratio minimum value residual covariance matrix calculating means 22 will be described in detail.
First, from the definition of the Kalman filter, the smoothing error covariance matrix P_sk(T) is a state vector VX representing the true value of the target_k(T), smooth vector VX_skUsing (t), the following equation (52) is obtained.
[Formula 38]

In the above equation (52), E [] is a symbol representing an average.
[0096]
Where the smoothing vector VX_sk(T) and the smoothing error covariance matrix P_skAs a simple estimate of (t), detection data D_kjSmooth vector VVX calculated directly from_skj(T) and detection data D_kjSmooth error covariance matrix PP calculated directly from_sk(T) is defined as the following equations (53) and (54), respectively.
[39]

In the above equation (53), VZ_kjIndicates target detection data excluding clutter detection data in the coarse gate. T_kIndicates the sampling time k.
[0097]
Here, according to the theory of the Kalman filter, the motion model is expressed as in equation (55).
[Formula 40]

In the above equation (55), Φ_k-1Is the time t_k-1To t_kIs a known transition matrix of the state vector, and is expressed by the following equation (56).
[Expression 41]

In the following equation (56), I is a 3 × 3 unit matrix.
[0098]
In the above equation (55), W_kIs the mean zero vector, error covariance matrix Q_kA driving noise vector according to the trivariate normal distribution and corresponds to an acceleration vector. Δ_k-1Is a known transformation matrix of the drive noise vector, and is represented by the following equation (57).
[Expression 42]

[0099]
State vector VX in equation (55) above_k(T) is a vector composed of a target position vector and a velocity vector, and is represented by the following equation (58) in the case of a three-dimensional space.
[Equation 43]

In the above equation (58), x_k, Y_k, Z_kIndicates an arbitrary position in the north reference Cartesian coordinates, x_k, Y_k, Z_kThe symbol “•” at the top of the symbol indicates first-order time differentiation, and the symbol “•” at the top is attached to x_k, Y_k, Z_kIndicates the speed at that position.
[0100]
Also, from the Kalman filter theory, target detection data VZ excluding clutter detection data in the coarse gate_kjIs expressed by the following equation (59).
(44)

In the above equation (59), H is a known observation matrix, and is represented by the following equation (60).
[Equation 45]

[0101]
In the above equation (59), U_kjIs the mean zero vector and the covariance matrix Λ_kjT according to the trivariate white normal distribution of_kIs the observation noise vector in polar coordinates at_kjIs the observed noise vector U_kjIs a known transformation matrix. Here, Γ in the above equation (6)_kj(T), but the observed noise vector U_kjThe known transformation matrix Γ_kjThe true value is used when calculating, but since the true value of the target is not actually known, a predicted value is inserted instead of the true value as in equation (6). Therefore, Γ_kjAnd Γ_kjTreat (t) as the same.
[0102]
Therefore, the following formula (61) is obtained by applying the formula (53), formula (55), formula (59), and formula (4) to the formula (54).
[Equation 46]

Q in equation (61) above_kIs the drive noise W in the above equation (55)._kIs the covariance matrix.
[0103]
Here, from the theory of the Kalman filter, the smooth vector VX_sk(T) is the smoothing error covariance matrix P_skIn contrast to the optimal value in the sense of minimizing the sum of the diagonal components of (t), the detection data D_kjSmooth vector VVX calculated directly from_sk(T) is detection data D_kjSmooth error covariance matrix P calculated directly from_skIt is not an optimal value in the sense of minimizing the sum of the diagonal components of (t). Therefore, the relationship of the following equation (62) is established.
[Equation 47]

[0104]
From the relationship of the above equation (62), the relationship of the following equation (63) is established.
[Formula 48]

[0105]
Here, from the theory of the Kalman filter, the prediction error covariance matrix P_{pk + 1}(T) is calculated by the following equation (64).
[Formula 49]

[0106]
Detection data D_kjSmooth error covariance matrix PP calculated directly from_skDetection data D after one sampling using (t)_kjThe prediction error covariance matrix calculated directly from_{pk + 1}(T) is calculated as in the following equation (65).
[Equation 50]

[0107]
Using the above expression (65), expression (61), expression (56), expression (57), and the definition (60) of the observation matrix H, the following expression (66) is obtained.
[Formula 51]

[0108]
Here, the following equation (67) is established from the relationship between the equations (43) and (49).
[Formula 52]

[0109]
Therefore, when the above equation (66) is substituted into equation (67), the following equation (68) is obtained.
[53]

[0110]
Moreover, the following (69) Formula is obtained using the said (66) Formula, (67) Formula, (68) Formula.
[Formula 54]

[0111]
The left side of the equation (69) is the prediction error covariance matrix P on the right side of the equation (2)._pk(T) is detected data D_kjPrediction error covariance matrix PP calculated directly from_pkIt can be considered that it was replaced with (t). Here, the prediction error covariance matrix PP_pkResidual covariance matrix SS calculated from (t)_k(T) is defined as the following equation (70). Further, the right side of the above expression (69) is defined as the following expression (71).
[Expression 55]

In the above equation (71), E_kIndicates a residual covariance matrix independent of the number of detected data j and the number of tracks.
[0112]
Then, the following expression (72) is calculated from the above expressions (69), (70), and (71).
[56]

[0113]
Further, the following expression (73) is established from the above expressions (2), (63), (64), and (65).
[Equation 57]

[0114]
Therefore, the following equation (74) is established from the above equations (72) and (73).
[Formula 58]

[0115]
The above equation (74) indicates that the coarse gate defined by the following equation (75) is wider than the fine gate defined by the above equation (11). That is, detection data D in the precision gate defined by the equation (11)_kjAlways exists in the coarse gate defined by equation (75).
[Formula 59]

[0116]
Accordingly, in the S / N ratio minimum value residual covariance matrix calculating means 22, when the user determines that the calculation load is heavy, the residual covariance not dependent on the number of detected data j and the number of tracks is calculated using the above equation (71). Variance matrix E_kIs calculated.
[0117]
Here, Q in the above equation (71)_kIs a fixed value determined before operation, so the residual covariance matrix E in equation (71)_kIndicates that the number of detected data is determined by a fixed value independent of the number of detected data j and the number of tracks. That is, the calculation load of the computer when the coarse gate of the above expression (75) is used is S in the above expression (46)._kmin(T) to S '_kminCompared with the calculation load of the computer when the coarse gate replaced with (t) is used, the load is further reduced.
[0118]
Next, the operation will be described.
FIG. 6 is a flowchart showing the flow of processing of the tracking device according to Embodiment 3 of the present invention. Step ST1 is the same as that in FIG. 2 of the first embodiment.
[0119]
In step ST21, the S / N ratio minimum value residual covariance matrix calculating means 22 performs the radar maximum detection distance RANGE._maxBased on the observation error covariance matrix R using the above equation (47)_kjError variance σcr in distance direction of components² _min, Elevation error variance RANGE² _maxσce_min, Azimuthal error variance RANGE² _maxσcb_minOf which observation error covariance matrix R_kjThe maximum value component of the components is selected.
[0120]
In step ST22, the S / N ratio minimum value residual covariance matrix calculating means 22 selects the selected observation error covariance matrix R._kjThe observation error covariance matrix R in the north reference orthogonal coordinates independent of the sampling time k, the number of detected data j, and the number of wakes, using the maximum component of the component_minIs calculated using the above equation (48).
[0121]
In step ST23, the S / N ratio minimum residual covariance matrix calculation means 22 detects the detection data D output from the observation means 1._kjAnd detection data D_kjS / N ratio SNR associated with_kj, The prediction error covariance matrix P of the current time output from the delay means 8_pk(T) and prediction vector VX at the current time_pk(T) is input, and the calculated observation error covariance matrix R_minAnd the input detection data D_kjPrediction error covariance matrix PP calculated directly from_pk(T) and the residual covariance matrix E independent of the number of detected data j and the number of tracks_kAnd the calculated residual covariance matrix E_k, And input detection data D_kjDetecting data D_kjS / N ratio SNR associated with_kj, Prediction error covariance matrix P at the current time_pk(T) and prediction vector VX at the current time_pk(T) is output to the rough gate inside / outside determination means 23.
[0122]
In step ST24, the coarse gate inside / outside determination means 23 inputs the residual covariance matrix E_k, And current time prediction vector VX_pkFrom (t), a rough gate which is a target existence expectation area is calculated, and the input detection data D_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetected data D in the coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetection data D in the determined coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kj, And the input prediction error covariance matrix P_pk(T) and prediction vector VX at the current time_pk(T) is output to each S / N ratio residual covariance matrix calculating means 4.
[0123]
The processing from step ST4 to ST9 is the same as that in FIG. 2 of the first embodiment. Therefore, by the above steps, the calculation load of the computer when using the above equation (71) does not depend on the number of detected data j and the number of tracks compared to the calculation load of the computer when using the above equation (50). Residual covariance matrix E_kBecause it is used, it becomes even lighter.
[0124]
As described above, according to the third embodiment, the residual covariance matrix E independent of the number of detected data j and the number of wakes._kCompared to the second embodiment, the coarse gate inside / outside judgment means 23, each S / N ratio residual covariance matrix calculation means 4, the fine gate inside / outside judgment means 5, and the data update means 6 are compared with the second embodiment. Thus, the effect of reducing the computational load in each process of the prediction means 7 can be obtained.
[0125]
Embodiment 4 FIG.
FIG. 7 is a block diagram showing the configuration of the tracking device according to Embodiment 4 of the present invention. In FIG._kjAnd detection data D_kjS / N ratio SNR associated with_kj, And the prediction error covariance matrix P of the current time output from the delay means 8_pk(T) and prediction vector VX at the current time_pk(T) is input, and the maximum detection distance RANGE of the radar_maxThe observation error covariance matrix R_kjThe maximum component of the component is selected, and the selected observation error covariance matrix R_kjMaximum value component of component and input detection data D_kjDetecting data D_kjS / N ratio SNR associated with_kj, Prediction error covariance matrix P at the current time_pk(T) and prediction vector VX at the current time_pkThis is an observation error covariance matrix maximum value selection means for outputting (t).
[0126]
In FIG. 7, reference numeral 33 denotes an observation error covariance matrix R from the observation error covariance matrix maximum value selection means 32._kjMaximum value component, detection data D_kjDetecting data D_kjS / N ratio SNR associated with_kj, Prediction error covariance matrix P at the current time_pk(T) and prediction vector VX at the current time_pkEnter (t), fixed sampling interval T_intTo calculate a coarse gate having a radius radius 0 independent of the sampling time k, the number of detection data j, and the number of tracks, and the input detection data D_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetected data D in the coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kjDetection data D in the determined coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kj, And the input prediction error covariance matrix P_pk(T) and prediction vector VX at the current time_pkThis is a rough gate inside / outside determination means for outputting (t), and other configurations are the same as those in FIG.
[0127]
Here, a rough gate having a radius radius0 will be described.
As shown in the following equation (76), the sampling interval is a constant value T_intAnd drive noise covariance matrix Q_kAs a diagonal matrix irrelevant to the coordinate axes and time, as shown in the following equation (77).
[Expression 60]

In the above equation (77), q is a driving noise constant.
[0128]
Then, the coarse gate of the above expression (75) can be rewritten as the following expression (78) from the above expressions (48), (71), and (76).
[Equation 61]

[0129]
In the equation (78), the radius radius0 is expressed as the following equation (79).
[62]

[0130]
Here, in contrast to the coarse gate defined by the above equation (75), the coarse gate having the radius radius 0 defined by the above equation (78) does not depend on the sampling interval because the sampling interval is fixed. Therefore, the calculation load of the computer when the coarse gate of radius (radio 0) in the equation (78) is used is further reduced as compared with the computer load when the coarse gate of the equation (75) is used.
[0131]
Next, the operation will be described.
FIG. 8 is a flowchart showing the flow of processing of the tracking device according to Embodiment 4 of the present invention. Step ST1 is the same as that in FIG. 1 of the first embodiment.
[0132]
In step ST31, the observation error covariance matrix maximum value selection means 32 detects the detection data D output by the observation means 1._kjAnd detection data D_kjS / N ratio SNR associated with_kj, And the prediction error covariance matrix P of the current time output from the delay means 8_pk(T) and prediction vector VX at the current time_pk(T) is input, and the maximum detection distance RANGE of the radar_maxThe observation error covariance matrix R_kjThe maximum component of the component is selected, and the selected observation error covariance matrix R_kjMaximum value component of component and input detection data D_kjDetecting data D_kjS / N ratio SNR associated with_kj, Prediction error covariance matrix P at the current time_pk(T) and prediction vector VX at the current time_pk(T) is output to the rough gate inside / outside determination means 33.
[0133]
In step ST32, the coarse gate inside / outside determination means 33 receives the observation error covariance matrix R from the observation error covariance matrix maximum value selection means 32._kjMaximum value component, detection data D_kjDetecting data D_kjS / N ratio SNR associated with_kj, Prediction error covariance matrix P at the current time_pk(T) and prediction vector VX at the current time_pkEnter (t), fixed sampling interval T_intBy using the above equation (79), a rough gate having a radius radius 0 independent of the sampling time k, the number of detected data j, and the number of tracks is calculated.
[0134]
In step ST33, the rough gate inside / outside determination means 33 inputs the detected data D_kjAnd detection data D_kjS / N ratio SNR associated with_kjThe detection data D contained in the coarse gate having the radius radius 0 calculated using the equation (78)._kjAnd detection data D_kjS / N ratio SNR associated with_kjDetection data D in the determined coarse gate_kjAnd detection data D_kjS / N ratio SNR associated with_kj, And the input prediction error covariance matrix P_pk(T) and prediction vector VX at the current time_pk(T) is output to each S / N ratio residual covariance matrix calculating means 4.
[0135]
The processing from step ST4 to ST9 is the same as that in FIG. 2 of the first embodiment. Therefore, according to the above steps, the calculation load of the computer in the case of making a rough gate inside / outside determination using the equation (78) by the coarse gate of radius radius 0 independent of the sampling time k, the number of detected data j, and the number of tracks is Residual covariance matrix E independent of number of data j and number of tracks_kThus, the calculation load is further reduced compared to the calculation load of the computer in the case where the rough gate inside / outside determination is performed using the equation (71).
[0136]
As described above, according to the fourth embodiment, the rough gate inside / outside determination is performed by the coarse gate having the radius radius0 independent of the sampling time k, the number of detected data j, and the number of tracks, compared to the third embodiment. The calculation load in each process of the coarse gate inside / outside determination means 33, each S / N ratio residual covariance matrix calculation means 4, the fine gate inside / outside determination means 5, the data update means 6, and the prediction means 7 is reduced. An effect is obtained.
[0137]
Embodiment 5. FIG.
FIG. 9 is a block diagram showing the configuration of the tracking device according to Embodiment 5 of the present invention. In FIG. 9, reference numeral 41 denotes each S / N ratio residual covariance matrix S output from the precision gate inside / outside determination means 5._kjThe number of (t), that is, the number of wakes, and the number of detected data j at the current sampling time detected by the observation means 1 are compared, and the observation means 1 and the S / N ratio minimum value residual covariance matrix calculation means 2 use it. S / N ratio minimum value SNR_minIs the S / N ratio minimum value control means for notifying the observation means 1 and the S / N ratio minimum value residual covariance matrix calculating means 2, and other configurations are the same as those in FIG. 1 of the first embodiment. It is equivalent.
[0138]
In the signal detection in the observation means 1, it is determined that a target has been detected by a signal having a certain level of power with respect to the noise power δ, and an S / N ratio that is a threshold value for determining that this target has been detected. S / N ratio minimum value SNR_minThe observation means 1 is δ · SNR_minThe signal of the above power is detected as the detected target detection data D._kjDetecting data D_kjS / N ratio SNR associated with_kjOutput as. Here, the S / N ratio minimum value SNR_minIs also a parameter that determines the false alarm probability.
[0139]
Therefore, the S / N ratio minimum value SNR_minDetection data D obtained by increasing the value of and reducing the false alarm probability_kjCan be reduced. For this reason, each S / N ratio residual covariance matrix S output from the precision gate inside / outside determination means 5 before one sampling._kjWhen the number of detected data j at the current sampling time is significantly larger than the number of (t), that is, the number of wakes, the S / N ratio minimum value used by the observing means 1 is as shown in the following equation (80). SNR_minS / N ratio minimum value SSNR larger than the value of_minCan be used to narrow down the number of detected data j.
[Equation 63]

[0140]
That is, detection data D_kjS / N ratio SNR associated with_kjOnly when the following expression (81) is satisfied, the S / N ratio minimum value SSNR_minIs used for tracking, otherwise it is not used for tracking. This process can reduce the number j of detection data used for tracking, and lightens the computation load of the computer.
[Expression 64]

[0141]
Thus, the observation means 1 has a predetermined S / N ratio minimum value SNR._minDetecting data D using_kjHowever, the target detection data D is determined based on the determination result of the precision gate inside / outside determination means 5._kjClutter detection data D other than_kjIncreases, the S / N ratio minimum value SNR_minNeed to be larger. Conversely, target detection data D_kjWhen there is little S / N ratio SNR_minNeed to be small.
[0142]
Next, the operation will be described.
FIG. 10 is a flowchart showing the flow of processing of the tracking device according to Embodiment 5 of the present invention. The processes from step ST1 to ST9 are the same as those in FIG. 1 of the first embodiment.
[0143]
In step ST41, the S / N ratio minimum value control means 41 outputs each S / N ratio residual covariance matrix S output from the precision gate inside / outside determination means 5._kjThe number of (t), that is, the number of tracks is compared with the number j of detected data at the current sampling time detected by the observation means 1.
[0144]
In step ST42, the S / N ratio minimum value control means 41 compares each S / N ratio residual covariance matrix S._kjWhen the number of (t), that is, the number of wakes, and the number of detected data j at the current sampling time detected by the observation means 1 are significantly different, the observation means 1 and the S / N ratio minimum value residual covariance matrix calculation means 2 Used S / N ratio minimum value SNR_minAnd the observation means 1 and the S / N ratio minimum value residual covariance matrix calculation means 2 are notified.
[0145]
For example, each S / N ratio residual covariance matrix S_kjWhen the number of detected data j at the current sampling time detected by the observation means 1 is significantly larger than the number of (t), that is, the number of wakes, the observation means 1 and the S / N ratio minimum value residual covariance matrix calculation means S / N ratio minimum value SNR used by 2_minS / N ratio minimum value SSNR larger than the value of_minTo the observation unit 1 and the S / N ratio minimum value residual covariance matrix calculation unit 2.
[0146]
After changing the S / N ratio minimum value in step ST42, the process returns to step ST1 and the observation means 1 changes the changed S / N ratio minimum value SNR._min, The detection data D in the same manner as in the first embodiment._kjDetecting data D_kjS / N ratio SNR associated with_kjIn step ST2, the S / N ratio minimum value residual covariance matrix calculating means 2 outputs the changed S / N ratio minimum value SNR._minUsing the observation error covariance matrix R in the same manner as in the first embodiment._kmin(T) is calculated. The subsequent processing is the same as in FIG. 2 of the first embodiment.
[0147]
As described above, according to the fifth embodiment, each S / N ratio residual covariance matrix S output from the precision gate inside / outside determination means 5 is used._kjWhen the number of (t), that is, the number of wakes, and the number of detected data j at the current sampling time detected by the observation means 1 are significantly different, the S / N ratio minimum value control means 41 determines that the observation means 1 and the S / N ratio S / N ratio minimum value SNR used by the minimum value residual covariance matrix calculating means 2_minThe observation means 1 and the S / N ratio minimum value residual covariance matrix calculation means 2 are changed to the changed S / N ratio minimum value SNR._minFor example, when the number of detected data j is remarkably large, the number of detected data j in the coarse gate can be reduced, and the fine gate inside / outside determining means 5, the data updating means 6, the prediction The effect that the calculation load of calculation in each process of the means 7 is reduced is obtained.
[0148]
【The invention's effect】
As described above, according to the present invention, the observation error covariance matrix R at the current time that does not depend on the number of detection data calculated using the S / N ratio minimum value._kminFrom (t), the residual covariance matrix S independent of the number of detected data_kminS / N ratio minimum value residual covariance matrix calculating means for calculating (t), and residual covariance matrix S_kminThe coarse gate which is the target existence expected area is calculated from (t), the detected data and S / N ratio contained in the coarse gate are determined for the input detection data and S / N ratio, and the determined coarse gate is determined. The coarse gate inside / outside determination means for outputting the detection data and S / N ratio contained in the inside, and the output from the coarse gate inside / outside determination means, depending on the sampling time, the number of detection data, and the number of tracks, depending on the coarse gate Each S / N ratio residual covariance matrix S_kjUsing the outputs from the respective S / N ratio residual covariance matrix calculating means for calculating (t) and the respective S / N ratio residual covariance matrix calculating means, the target existence expectation region narrower than that of the coarse gate is refined. Calculate the gate, determine the detection data and S / N ratio in the fine gate for the input detection data and S / N ratio in the coarse gate, sampling time, number of detection data, number of wakes Depending on each S / N ratio residual covariance matrix S by the precision gate_kjBy including the fine gate inside / outside determination means for calculating (t), there is an effect that the calculation load of the fine gate for removing unnecessary signals for the target signal is reduced.
[0149]
According to this invention, each S / N ratio residual covariance matrix S output from the precision gate inside / outside determination means._kjAccording to the number of (t) and the number of detection data output by the observation means, the value of the S / N ratio minimum value used by the observation means and the S / N ratio minimum value residual covariance matrix calculation means is changed, S / N ratio minimum value control means for notifying the observation means and the S / N ratio minimum value residual covariance matrix calculating means, and the observation means and the S / N ratio minimum value residual covariance matrix calculating means are changed. By performing processing using the S / N ratio minimum value, for example, when the number of detection data is extremely large, the number of detection data in the coarse gate can be reduced, and the calculation load is reduced. There is.
[0150]
According to the present invention, the maximum value component of the observation error covariance matrix components is selected based on the maximum detection distance of the radar, and the sampling time and the number of detection data are selected using the maximum value component of the selected observation error covariance matrix. , Observation error covariance matrix R independent of the number of tracks_min, The output from the observation means, and the calculated observation error covariance matrix R_minAnd a simple residual covariance matrix S ′ independent of the number of detected data_kmin(T) is calculated, and the calculated simple residual covariance matrix S ′_kminS / N ratio minimum residual covariance matrix calculating means for outputting (t) and a simple residual covariance matrix S ′_kminThe coarse gate which is the target existence expected area is calculated from (t), the detected data and S / N ratio contained in the coarse gate are determined for the input detection data and S / N ratio, and the determined coarse gate is determined. The coarse gate inside / outside determination means for outputting the detection data and S / N ratio contained in the inside, and the output from the coarse gate inside / outside determination means, depending on the sampling time, the number of detection data, and the number of tracks, depending on the coarse gate Each S / N ratio residual covariance matrix S_kj(T) is calculated, and each calculated S / N ratio residual covariance matrix S_kjUsing the outputs from the respective S / N ratio residual covariance matrix calculating means for outputting (t) and the respective S / N ratio residual covariance matrix calculating means, the target existence expectation region narrower than that of the coarse gate is refined. Calculate the gate, determine the detection data and S / N ratio in the fine gate for the input detection data and S / N ratio in the coarse gate, sampling time, number of detection data, number of wakes Depending on each S / N ratio residual covariance matrix S by the precision gate_kjBy including the fine gate inside / outside determination means for calculating (t), the calculation load in each process of the coarse gate inside / outside determination means, each S / N ratio residual covariance matrix calculation means, and the precision gate inside / outside determination means Is effective.
[0151]
According to the present invention, the maximum value component of the observation error covariance matrix components is selected based on the maximum detection distance of the radar, and the sampling time and the number of detection data are selected using the maximum value component of the selected observation error covariance matrix. , Observation error covariance matrix R independent of the number of tracks_min, The output from the observation means, and the calculated observation error covariance matrix R_min, The residual covariance matrix E independent of the number of detected data and the number of tracks_kS / N ratio minimum value residual covariance matrix calculating means for calculating the residual covariance matrix E_kTo calculate the rough gate which is the target existence expected region, and for the input detection data and S / N ratio, determine the detection data and S / N ratio that are in the rough gate, and enter the determined rough gate. The coarse gate inside / outside judging means for outputting the detected data and the S / N ratio, and the output from the coarse gate inside / outside judging means, each S / N by the coarse gate depending on the sampling time, the number of detected data, and the number of wakes. N-ratio residual covariance matrix S_kjUsing the outputs from the respective S / N ratio residual covariance matrix calculating means for calculating (t) and the respective S / N ratio residual covariance matrix calculating means, the target existence expectation region narrower than that of the coarse gate is refined. Calculate the gate, determine the detection data and S / N ratio in the fine gate for the detection data and S / N ratio in the input coarse gate, sampling time, number of detection data, number of wakes Depending on each S / N ratio residual covariance matrix S by the fine gate_kjBy including the fine gate inside / outside determination means for calculating (t), the calculation load in each process of the coarse gate inside / outside determination means, each S / N ratio residual covariance matrix calculation means, and the precision gate inside / outside determination means Is effective.
[0152]
According to the present invention, the observation error covariance matrix maximum value selection means for selecting the maximum value component of the observation error covariance matrix components based on the maximum detection distance of the radar, and the observation error covariance matrix maximum value selection means from Using the output, a fixed sampling interval T_intTo calculate a coarse gate with radius radius0 independent of sampling time, number of detected data, and number of tracks, and determine the detected data and S / N ratio in the coarse gate for the input detected data and S / N ratio The coarse gate internal / external determination means and the output from the coarse gate internal / external determination means, each S / N ratio residual covariance matrix S by the coarse gate depends on the sampling time, the number of detected data, and the number of tracks._kjUsing the outputs from the respective S / N ratio residual covariance matrix calculating means for calculating (t) and the respective S / N ratio residual covariance matrix calculating means, the target existence expectation region narrower than that of the coarse gate is refined. Calculate the gate, determine the detection data and S / N ratio in the fine gate for the input detection data and S / N ratio in the coarse gate, sampling time, number of detection data, number of wakes Depending on each S / N ratio residual covariance matrix S by the precision gate_kjBy including the fine gate inside / outside determination means for calculating (t), the calculation load in each process of the coarse gate inside / outside determination means, each S / N ratio residual covariance matrix calculation means, and the precision gate inside / outside determination means Is effective.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a tracking device according to Embodiment 1 of the present invention.
FIG. 2 is a flowchart showing a process flow of the tracking device according to the first embodiment of the present invention.
FIG. 3 is a block diagram showing a configuration of a tracking device according to Embodiment 2 of the present invention.
FIG. 4 is a flowchart showing a flow of processing of a tracking device according to Embodiment 2 of the present invention.
FIG. 5 is a block diagram showing a configuration of a tracking device according to Embodiment 3 of the present invention.
FIG. 6 is a flowchart showing a process flow of the tracking device according to the third embodiment of the present invention.
FIG. 7 is a block diagram showing a configuration of a tracking device according to Embodiment 4 of the present invention.
FIG. 8 is a flowchart showing a process flow of a tracking device according to a fourth embodiment of the present invention.
FIG. 9 is a block diagram showing a configuration of a tracking device according to a fifth embodiment of the present invention.
FIG. 10 is a flowchart showing a flow of processing of a tracking device according to Embodiment 5 of the present invention.
FIG. 11 is a block diagram showing a configuration of a conventional tracking device.
FIG. 12 is a diagram for explaining a coordinate system according to conventional observation means.
[Explanation of symbols]
1 observation means, 2 S / N ratio minimum value residual covariance matrix calculation means, 3 coarse gate inside / outside determination means, 4 S / N ratio residual covariance matrix calculation means, 5 fine gate inside / outside determination means, 6 data update Means 7 predicting means 8 delay means 12 S / N ratio minimum residual covariance matrix calculating means 13 coarse gate inside / outside determining means 22 S / N ratio minimum residual covariance matrix calculating means 23 coarse gate Inside / outside determination means, 32 observation error covariance matrix maximum value selection means, 33 coarse gate inside / outside determination means, 41 S / N ratio minimum value control means.

Claims (9)

Observation means for outputting detection data detected from an internal sensor and an S / N ratio associated with the detection data, using a minimum S / N ratio value that is a threshold value for detecting a target;
By inputting the output of the observation means and using the observation error covariance matrix R _kmin (t) at the current time that does not depend on the number of detection data calculated using the minimum value of the S / N ratio, a residual that does not depend on the number of detection data The covariance matrix S _kmin (t) is calculated, and the calculated residual covariance matrix S _kmin (t), the input detection data and the S / N ratio are output, and the S / N ratio minimum value residual covariance matrix is calculated. Means,
Based on the residual covariance matrix S _kmin (t) input from the S / N ratio minimum value residual covariance matrix calculating means, a rough gate which is a target existence expectation region is calculated, and the input detection data and S / N ratio are calculated. About the detection data and S / N ratio contained in the coarse gate, and the coarse gate inside / outside determination means for outputting the detected detection data and S / N ratio contained in the coarse gate;
Using the output of the coarse gate inside / outside determination means, each S / N ratio residual covariance matrix S _kj (t) by the coarse gate, which depends on the sampling time, the number of detected data, and the number of wakes, is calculated and calculated. Each S / N ratio residual covariance matrix S _kj (t), and each S / N ratio residual covariance matrix calculating means for outputting detection data and S / N ratio contained in the input coarse gate;
Using the outputs of the respective S / N ratio residual covariance matrix calculating means, a fine gate which is a target existence expectation region narrower than the coarse gate is calculated, and the detected data and S / N contained in the inputted coarse gate For the ratio, the detection data and S / N ratio contained in the precision gate are determined, and each S / N ratio residual covariance matrix S _kj (by the precision gate depends on the sampling time, the number of detection data, and the number of tracks. a fine gate inside / outside determination means for calculating t) and outputting the detected detection data and S / N ratio contained in the determined fine gate and each calculated S / N ratio residual covariance matrix S _kj (t); A tracking device characterized by comprising: Observation means for outputting detection data detected from an internal sensor and an S / N ratio associated with the detection data, using a minimum S / N ratio value that is a threshold value for detecting a target;
Input the output of the observation means, the current time prediction error covariance matrix and the current time prediction vector output by the delay means described later, and do not depend on the number of detection data calculated using the S / N ratio minimum value. Using the observation error covariance matrix R _kmin (t) at the current time and the input prediction error covariance matrix at the current time, a residual covariance matrix S _kmin (t) independent of the number of detected data is calculated. Both the calculated residual covariance matrix S _kmin (t) and the input detection data, S / N ratio, current time prediction error covariance matrix and current time prediction vector are output. A dispersion matrix calculating means;
Based on the residual covariance matrix S _kmin (t) input from the S / N ratio minimum value residual covariance matrix calculating means and the prediction vector at the current time, a rough gate that is a target existence expected region is calculated and input detection For the data and S / N ratio, the detection data and S / N ratio in the coarse gate are determined, and the detected detection data and S / N ratio in the determined coarse gate, and the input current time prediction Coarse gate internal / external determination means for outputting an error covariance matrix and a prediction vector at the current time;
Using the output of the coarse gate inside / outside determination means, each S / N ratio residual covariance matrix S _kj (t) by the coarse gate, which depends on the sampling time, the number of detected data, and the number of wakes, is calculated and calculated. Each S / N ratio residual covariance matrix S _kj (t), input detection data contained in the coarse gate, S / N ratio contained in the coarse gate, prediction error covariance matrix at the current time, and Each S / N ratio residual covariance matrix calculating means for outputting a prediction vector at the current time;
Using the outputs of the respective S / N ratio residual covariance matrix calculation means, a fine gate that is a target existence expectation region narrower than the coarse gate is calculated, and the detection data and S / N contained in the input coarse gate For the ratio, the detection data and S / N ratio contained in the precision gate are determined, and each S / N ratio residual covariance matrix S _kj (by the precision gate depends on the sampling time, the number of detection data, and the number of tracks. t), and the detected data and S / N ratio included in the determined precision gate, the calculated S / N ratio residual covariance matrix S _kj (t), and the input prediction error at the current time Fine gate inside / outside determination means for outputting a dispersion matrix and a prediction vector at the current time;
Data update means for calculating and outputting a smoothing error covariance matrix at the current time and a smoothing vector at the current time using the output of the precision gate inside / outside determination means,
Prediction means for calculating and outputting a prediction error covariance matrix after one sampling from the current time and a prediction vector after one sampling from the current time, using the output of the data updating means;
A tracking device comprising: a prediction error covariance matrix at the current time and a delay means for outputting a prediction vector at the current time using the output of the prediction means. According to the number of each S / N ratio residual covariance matrix S _kj (t) output from the precision gate inside / outside determination means and the number of detection data output from the observation means, the observation means and the S / N ratio minimum value residual are determined. S / N ratio minimum value control for changing the value of the S / N ratio minimum value used by the covariance matrix calculating means and notifying the observation means and the S / N ratio minimum value residual covariance matrix calculating means. Means and
The said observation means and the said S / N ratio minimum value residual covariance matrix calculation means perform a process using the changed S / N ratio minimum value, The claim 1 or 2 characterized by the above-mentioned. Tracking device. Observation means for outputting detection data detected from an internal sensor and an S / N ratio associated with the detection data, using a minimum S / N ratio value that is a threshold value for detecting a target;
Select the maximum value component of the observation error covariance matrix components based on the maximum detection distance of the radar, and use the maximum value component of the selected observation error covariance matrix, and do not depend on the sampling time, number of detected data, and number of tracks An observation error covariance matrix R _min is calculated, and a simple residual covariance matrix S ′ _kmin (t that does not depend on the number of detected data is obtained using the output from the observation means and the calculated observation error covariance matrix R _min. ), The calculated simple residual covariance matrix S ′ _kmin (t), and the S / N ratio minimum value residual covariance matrix calculating means for outputting the input detection data and S / N ratio;
Based on the simple residual covariance matrix S ′ _kmin (t) input from the S / N ratio minimum value residual covariance matrix calculating means, a rough gate which is a target existence expected region is calculated, and the input detection data and S For the / N ratio, the detection data and S / N ratio in the coarse gate are determined, and the coarse gate inside / outside determination means for outputting the detected detection data and S / N ratio in the determined coarse gate;
Using the output of the coarse gate inside / outside determination means, each S / N ratio residual covariance matrix S _kj (t) by the coarse gate, which depends on the sampling time, the number of detected data, and the number of wakes, is calculated and calculated. Each S / N ratio residual covariance matrix S _kj (t), and each S / N ratio residual covariance matrix calculating means for outputting detection data and S / N ratio contained in the input coarse gate;
Using the outputs of the respective S / N ratio residual covariance matrix calculating means, a fine gate which is a target existence expectation region narrower than the coarse gate is calculated, and the detected data and S / N contained in the inputted coarse gate For the ratio, the detection data and S / N ratio contained in the precision gate are determined, and each S / N ratio residual covariance matrix S _kj (by the precision gate depends on the sampling time, the number of detection data, and the number of tracks. a fine gate inside / outside determination means for calculating t) and outputting the detected detection data and S / N ratio contained in the determined fine gate and each calculated S / N ratio residual covariance matrix S _kj (t); A tracking device characterized by comprising: Observation means for outputting detection data detected from an internal sensor and an S / N ratio associated with the detection data, using a minimum S / N ratio value that is a threshold value for detecting a target;
Selects the maximum value component of the observation error covariance matrix component based on the maximum detection distance of the radar, and uses the maximum value component of the selected observation error covariance matrix and does not depend on the sampling time, number of detection data, and number of tracks An observation error covariance matrix R _min is calculated, and an output from the observation means, a current time prediction error covariance matrix and a current time prediction vector from a delay means described later, and the calculated observation error covariance matrix R _min are used, 'calculates _kmin (t), calculated simple residual covariance matrix S' simple residual covariance matrix does not depend on the detection number data S _kmin (t), and inputs the detection data, S / An S / N ratio minimum value residual covariance matrix calculating means for outputting an N ratio, a prediction error covariance matrix at the current time, and a prediction vector at the current time;
From the simple residual covariance matrix S ′ _kmin (t) input from the S / N ratio minimum value residual covariance matrix calculating means and the current time prediction vector, a rough gate that is a target existence expected region is calculated, With respect to the input detection data and S / N ratio, the detection data and S / N ratio contained in the coarse gate are determined, and the detected detection data and S / N ratio contained in the determined coarse gate and the inputted current data are determined. Coarse gate inside / outside determination means for outputting a time prediction error covariance matrix and a current time prediction vector;
Using the output of the coarse gate inside / outside determination means, each S / N ratio residual covariance matrix S _kj (t) by the coarse gate, which depends on the sampling time, the number of detected data, and the number of wakes, is calculated and calculated. Each S / N ratio residual covariance matrix S _kj (t), input detection data contained in the coarse gate, S / N ratio contained in the coarse gate, prediction error covariance matrix at the current time, and Each S / N ratio residual covariance matrix calculating means for outputting a prediction vector at the current time;
Using the outputs of the respective S / N ratio residual covariance matrix calculation means, a fine gate that is a target existence expectation region narrower than the coarse gate is calculated, and the detection data and S / N contained in the input coarse gate For the ratio, the detection data and S / N ratio contained in the precision gate are determined, and each S / N ratio residual covariance matrix S _kj (by the precision gate depends on the sampling time, the number of detection data, and the number of tracks. t), and the detected data and S / N ratio included in the determined precision gate, the calculated S / N ratio residual covariance matrix S _kj (t), and the input prediction error at the current time Fine gate inside / outside determination means for outputting a dispersion matrix and a prediction vector at the current time;
Data update means for calculating and outputting a smoothing error covariance matrix at the current time and a smoothing vector at the current time using the output of the precision gate inside / outside determination means,
Prediction means for calculating and outputting a prediction error covariance matrix after one sampling from the current time and a prediction vector after one sampling from the current time, using the output of the data updating means;
A tracking device comprising: a prediction error covariance matrix at the current time and a delay means for outputting a prediction vector at the current time using the output of the prediction means. Observation means for outputting detection data detected from an internal sensor and an S / N ratio associated with the detection data, using a minimum S / N ratio value that is a threshold value for detecting a target;
Selects the maximum value component of the observation error covariance matrix component based on the maximum detection distance of the radar, and uses the maximum value component of the selected observation error covariance matrix and does not depend on the sampling time, number of detection data, and number of tracks calculating the observation error covariance matrix R _min, output from said observation means, and using the calculated observation error covariance matrix R _min, calculated detection data number, the residual covariance matrix E _k does not depend on the number of track A S / N ratio minimum value residual covariance matrix calculating means for outputting the calculated residual covariance matrix E _k and the input detection data and S / N ratio;
Based on the residual covariance matrix E _k input from the S / N ratio minimum value residual covariance matrix calculating means, a rough gate which is a target existence expectation region is calculated, and the input detection data and S / N ratio are roughly calculated. Coarse gate inside / outside determination means for determining detection data and S / N ratio in the gate and outputting the detected detection data and S / N ratio in the determined coarse gate;
Using the output of the coarse gate inside / outside determination means, each S / N ratio residual covariance matrix S _kj (t) by the coarse gate, which depends on the sampling time, the number of detected data, and the number of wakes, is calculated and calculated. Each S / N ratio residual covariance matrix S _kj (t), and each S / N ratio residual covariance matrix calculating means for outputting detection data and S / N ratio contained in the input coarse gate;
Using the outputs of the respective S / N ratio residual covariance matrix calculating means, a fine gate which is a target existence expectation region narrower than the coarse gate is calculated, and the detected data and S / N contained in the inputted coarse gate For the ratio, the detection data and S / N ratio contained in the precision gate are determined, and each S / N ratio residual covariance matrix S _kj (by the precision gate depends on the sampling time, the number of detection data, and the number of tracks. a fine gate inside / outside determination means for calculating t) and outputting the detected detection data and S / N ratio contained in the determined fine gate and each calculated S / N ratio residual covariance matrix S _kj (t); A tracking device characterized by comprising: Observation means for outputting detection data detected from an internal sensor and an S / N ratio associated with the detection data, using a minimum S / N ratio value that is a threshold value for detecting a target;
Select the maximum value component of the observation error covariance matrix components based on the maximum detection distance of the radar, and use the maximum value component of the selected observation error covariance matrix, and do not depend on the sampling time, number of detected data, and number of tracks An observation error covariance matrix R _min is calculated, and an output from the observation means, a current time prediction error covariance matrix and a current time prediction vector from a delay means described later, and the calculated observation error covariance matrix R _min are The residual covariance matrix E _k that does not depend on the number of detection data and the number of wakes is calculated, the calculated residual covariance matrix E _k , the input detection data, the S / N ratio, and the prediction error of the current time. S / N ratio minimum residual covariance matrix calculating means for outputting a variance matrix and a prediction vector at the current time;
Based on the residual covariance matrix E _k input from the S / N ratio minimum value residual covariance matrix calculating means and the prediction vector at the current time, a rough gate which is a target existence expected region is calculated, and the input detection data and S The detection data and S / N ratio in the coarse gate are determined for the / N ratio, the detected data and S / N ratio in the determined coarse gate, and the input prediction error covariance at the current time Coarse gate inside / outside determination means for outputting a matrix and a prediction vector at the current time;
Using the output of the coarse gate inside / outside determination means, each S / N ratio residual covariance matrix S _kj (t) by the coarse gate, which depends on the sampling time, the number of detected data, and the number of wakes, is calculated and calculated. Each S / N ratio residual covariance matrix S _kj (t), input detection data contained in the coarse gate, S / N ratio contained in the coarse gate, prediction error covariance matrix at the current time, and Each S / N ratio residual covariance matrix calculating means for outputting a prediction vector at the current time;
Using the outputs of the respective S / N ratio residual covariance matrix calculation means, a fine gate that is a target existence expectation region narrower than the coarse gate is calculated, and the detection data and S / N contained in the input coarse gate For the ratio, the detection data and S / N ratio contained in the precision gate are determined, and each S / N ratio residual covariance matrix S _kj (by the precision gate depends on the sampling time, the number of detection data, and the number of tracks. t), and the detected data and S / N ratio included in the determined precision gate, the calculated S / N ratio residual covariance matrix S _kj (t), and the input prediction error at the current time Fine gate inside / outside determination means for outputting a dispersion matrix and a prediction vector at the current time;
Data update means for calculating and outputting a smoothing error covariance matrix at the current time and a smoothing vector at the current time using the output of the precision gate inside / outside determination means,
Prediction means for calculating and outputting a prediction error covariance matrix after one sampling from the current time and a prediction vector after one sampling from the current time, using the output of the data updating means;
A tracking device comprising: a prediction error covariance matrix at the current time and a delay means for outputting a prediction vector at the current time using the output of the prediction means. Observation means for outputting detection data detected from an internal sensor and an S / N ratio associated with the detection data, using a minimum S / N ratio value that is a threshold value for detecting a target;
The maximum value component of the observation error covariance matrix component is selected based on the maximum detection distance of the radar, and the maximum value component of the selected observation error covariance matrix, the detection data and the S / N ratio input from the observation means are selected. An output observation error covariance matrix maximum value selection means;
By using the output of the observation error covariance matrix maximum value selection means, a coarse gate having a radius radius 0 that does not depend on the sampling time, the number of detection data, and the number of wakes is calculated at a fixed sampling interval T _int. About the S / N ratio, the detection data and the S / N ratio that are included in the coarse gate are determined, and the coarse gate inside / outside determination means that outputs the detected detection data and the S / N ratio that are included in the determined coarse gate;
Using the output of the coarse gate inside / outside determination means, each S / N ratio residual covariance matrix S _kj (t) by the coarse gate, which depends on the sampling time, the number of detected data, and the number of wakes, is calculated and calculated. Each S / N ratio residual covariance matrix S _kj (t), and each S / N ratio residual covariance matrix calculating means for outputting detection data and S / N ratio contained in the input coarse gate;
Using the outputs of the respective S / N ratio residual covariance matrix calculating means, a fine gate which is a target existence expectation region narrower than the coarse gate is calculated, and the detected data and S / N contained in the inputted coarse gate For the ratio, the detection data and S / N ratio contained in the precision gate are determined, and each S / N ratio residual covariance matrix S _kj (by the precision gate depends on the sampling time, the number of detection data, and the number of tracks. a fine gate inside / outside determination means for calculating t) and outputting the detected detection data and S / N ratio contained in the determined fine gate and each calculated S / N ratio residual covariance matrix S _kj (t); A tracking device characterized by comprising: Observation means for outputting detection data detected from an internal sensor and an S / N ratio associated with the detection data, using a minimum S / N ratio value that is a threshold value for detecting a target;
The maximum value component of the observation error covariance matrix component is selected based on the maximum detection distance of the radar, the maximum value component of the selected observation error covariance matrix, detection data and S / N ratio input from the observation means, and Using the output of the observation error covariance matrix maximum value selection means for outputting the prediction error covariance matrix of the current time and the prediction vector of the current time from the delay means described later, and the output of the observation error covariance matrix maximum value selection means, By using a fixed sampling interval T _int , a coarse gate with a radius radius 0 that does not depend on the sampling time, the number of detected data, and the number of tracks is calculated, and the detected data and the S / N ratio input to the detected data contained in the coarse gate and The S / N ratio is determined, the detected data and S / N ratio contained in the determined coarse gate, the input current time prediction error covariance matrix and the current time prediction vector. A rough gate inside / outside determination means for outputting a kuttle,
Using the output of the coarse gate inside / outside determination means, each S / N ratio residual covariance matrix S _kj (t) by the coarse gate, which depends on the sampling time, the number of detected data, and the number of wakes, is calculated and calculated. Each S / N ratio residual covariance matrix S _kj (t), input detection data contained in the coarse gate, S / N ratio contained in the coarse gate, prediction error covariance matrix at the current time, and Each S / N ratio residual covariance matrix calculating means for outputting a prediction vector at the current time;
Using the outputs of the respective S / N ratio residual covariance matrix calculation means, a fine gate that is a target existence expectation region narrower than the coarse gate is calculated, and the detection data and S / N contained in the input coarse gate For the ratio, the detection data and S / N ratio contained in the precision gate are determined, and each S / N ratio residual covariance matrix S _kj (by the precision gate depends on the sampling time, the number of detection data, and the number of tracks. t), and the detected data and S / N ratio included in the determined precision gate, the calculated S / N ratio residual covariance matrix S _kj (t), and the input prediction error at the current time Fine gate inside / outside determination means for outputting a dispersion matrix and a prediction vector at the current time;
Data update means for calculating and outputting a smoothing error covariance matrix at the current time and a smoothing vector at the current time using the output of the precision gate inside / outside determination means,
Prediction means for calculating and outputting a prediction error covariance matrix after one sampling from the current time and a prediction vector after one sampling from the current time, using the output of the data updating means;
A tracking device comprising: a prediction error covariance matrix at the current time and a delay means for outputting a prediction vector at the current time using the output of the prediction means.

Images