JP3629935B2 - Speed measurement method for moving body and speed measurement device using the method - Google Patents

Description

【００３４】
対応づけ処理部１１は、各ウィンドウＷ_１〜Ｗ_５について求められたウィンドウＷ_Ｕとの相違度ＤＦを比較し、相違度が最も小さくなるウィンドウをウィンドウＷ_Ｕに対応するものとして判別する。そしてそのウィンドウの中心点Ｑ_Ｌ（図示例の場合Ｌ＝２）を前記特徴点Ｐの対応点として決定する。
【００３５】
なお上記の相違度の代わりに、各ウィンドウＷ_１〜Ｗ_５毎にウィンドウＷ_Ｕとの間の正規化相互相関演算を行い、最も高い相関値が得られたウィンドウをウィンドウＷ_Ｕに対応するものと判別するようにしてもよい。
【００３６】
上記の対応づけ処理が、抽出されたすべての特徴点について行われると、３次元計測部１２は、対応する各特徴点Ｐ，Ｑ毎に、これら特徴点の２次元座標を三角測量の原理にあてはめて、対応する物点の３次元座標を算出する。
【００３７】
図７は、上記の三角測量の原理を示す。
図中Ｒは、道路上の車輌１６の所定の構成点を示すもので、各カメラ１ａ，１ｂの撮像面Ｇ_Ｕ，Ｇ_Ｌ上にこの点Ｒの物点像Ｐ，Ｑが現れている。
なお図中、Ｃ_Ｕはカメラ１ａの焦点を、Ｃ_Ｌはカメラ３ｂの焦点を、それぞれ示す。
【００３８】
上記の物点像Ｐ，Ｑが前記エッジ画像Ｅ_Ｕ，Ｅ_Ｌ間の対応する特徴点として現れるもので、各点上に対応する３次元座標は物点Ｒの空間位置に相当する。したがってこの対象物の代表的な物点について、その物点の画像上の結像位置を特定し、これら結像位置の２次元座標を用いて各物点の３次元座標を算出することにより、対象物の立体形状や空間位置を把握することができる。
【００３９】
図８（１）は、前記各カメラ１ａ，１ｂの位置関係により決定される空間座標系（以下これを「ステレオ座標系」という）を示し、図８（２）は、実際の空間における各物点の位置を表すための空間座標系と上記ステレオ座標系との位置関係を示す。
【００４０】
図８（１）中、Ｘ_ＵＹ_ＵＺ_Ｕ，Ｘ_ＬＹ_ＬＺ_Ｌで表される各空間座標系は、それぞれ各カメラ１ａ，１ｂのカメラ座標系を示し、これらカメラ座標系の原点Ｃ_Ｕ，Ｃ_Ｌ（前記焦点の位置）を結ぶ基線の中心点Ｏ′を原点としてステレオ座標系が設定される。なお前記したように、各カメラ１ａ，１ｂは光軸を平行に配置されているから、各カメラ座標系およびステレオ座標系の各軸Ｘ_ＵＹ_ＵＺ_Ｕ，Ｘ_ＬＹ_ＬＺ_Ｌ，Ｘ′Ｙ′Ｚ′は、互いに平行の位置関係にある。
【００４１】
各撮像面Ｇ_Ｕ，Ｇ_Ｌ上における２次元座標は、それぞれ対応するカメラ座標系のＺ_Ｕ，Ｚ_Ｌ軸との交点位置を原点ｏ_Ｕ，ｏ_Ｌとし、撮像面の水平方向にｘ_Ｕ，ｘ_Ｌ軸が、垂直方向にｙ_Ｕ，ｙ_Ｌ軸が定められている。また実際の空間位置を表す座標系は、図８（２）に示すように、カメラ座標系の原点Ｏ′から路面上に下ろした垂線の足を原点Ｏとし、道路の幅方向をＸ軸，長さ方向をＺ軸，高さ方向をＹ軸として、定められる。
【００４２】
このような座標系の設定において、前記対応づけられた各特徴点Ｐ，Ｑの２次元座標を（ｘ，ｙ_Ｕ），（ｘ，ｙ_Ｌ），各カメラ３ａ，３ｂの俯角をθ，各カメラ座標系の原点Ｃ_Ｕ，Ｃ_Ｌ間の距離をＢ，原点Ｏ，Ｏ′間の距離をＨとすると、各特徴点Ｐ，Ｑにより表される物点Ｒの３次元座標（Ｘ，Ｙ，Ｚ）は、つぎの（２）〜（４）式により算出される。
【００４３】
【数２】

【００４４】
【数３】

【００４５】
【数４】

【００４６】
このようにして、前記検出エリアｒ_Ｕ１，ｒ_Ｕ２，ｒ_Ｌ１，ｒ_Ｌ２内で抽出されたすべての特徴点に対応する３次元座標が算出されると、車輌認識部１３は、各検出エリアｒ_Ｕ１，ｒ_Ｕ２，ｒ_Ｌ１，ｒ_Ｌ２毎に、算出された各３次元座標をクラスタリング処理して、各座標を車輌毎にグループ化する。そしてこのグループの中から車輌の先頭位置を示す座標を抽出し、その抽出位置をもって車輌位置とする。この車輌位置として抽出された物点の３次元座標は、メモリ８内に格納され、以後、画像入力毎に、その時点での車輌の移動位置を示す３次元座標が順次蓄積されて、車輌の軌跡データが生成される。
【００４７】
なお車輌の認識処理は、クラスタリング処理に限らず、前述した特開平９−３３２３２号公報に記載のように、各３次元座標を、それぞれ道路の長さ方向（すなわちＺ軸方向）に沿う仮想垂直平面上に投影し、その投影結果を車輌の２次元モデルと照合する方法を用いてもよい。
【００４８】
つぎに既に過去の処理により認識されている車輌について、その移動位置を追跡して速度を算出する処理を説明する。
いまある時点で、車輌の初期位置または移動位置が認識されると、いずれか一方の入力画像（ここではカメラ３ａにより得られた入力画像Ｉ_Ｕとする）において、認識された車輌の位置に対応する特徴点（すなわち車輌の先頭位置を示す特徴点）の座標、およびこの特徴点を中心とする所定大きさの画像領域内の画像データ（以下これを「特徴画像Ｍ」という）が切り出され、追跡処理用のデータとしてメモリ８内に格納される。そして所定時間Ｔ経過後に、つぎの画像が入力されると、追跡処理部１４は、この時点でのカメラ１ａからの入力画像Ｉ_ＵＴ上で、前記１フレーム前に認識された車輌の移動位置を特定した後、この特定された２次元上の車輌位置を３次元座標に変換する。
【００４９】
図９は、入力画像_ＵＴ上において車輌の移動位置を特定するための具体的な方法を示す。この入力画像Ｉ_ＵＴは、前記図３（１）の入力画像Ｉ_Ｕの入力から所定時間経過後に得られた画像であって、図中の点Ａは、前記図３（１）の入力画像Ｉ_Ｕ上で車輌位置として特定された位置を示す。追跡処理部１４は、この点Ａの座標位置を基準に所定大きさの走査領域１７を設定した後、その走査領域内に前記特徴画像Ｍを走査して、各走査位置毎に前記対応づけ処理と同様の（１）式を実行する。この結果、相違度ＤＦが最も小さくなる走査位置において特徴画像Ｍの中心点に対応する点Ａ′が、車輌の移動位置として認識される。
【００５０】
入力画像Ｉ_ＵＴ上において車輌の移動位置を示す点Ａ′が特定されると、つぎに追跡処理部１４は、この点Ａ′の２次元座標（ｘ_Ｔ，ｙ_Ｔ）を（５）〜（７）式にあてはめて、前記点Ａ′に対応する物点の３次元座標、すなわち空間内における車輌の移動位置（Ｘ_Ｔ，Ｙ_Ｔ，Ｚ_Ｔ）を算出し、これをメモリ８内の軌跡データに追加する。
【００５１】
なおこの（５）〜（７）式は、車輌が移動しても、空間内における車輌の高さデータは変動しない点に着目してなされるもので、（５）（７）式のα_ｘ，α_ｙは、それぞれ、入力画像上のｘ，ｙの各軸方向における１画素分の長さを示す。また各式中のＹ_０は、該当車輌の画像が検出エリアに入った時点、すなわちこの車輌の３次元計測が認識可能となった時点で得られた車輌初期位置のＹ座標値である。この方法によれば、前記点Ａ′について、他方のカメラ１ｂからの入力画像Ｉ_Ｌ上での対応点を抽出することなく、３次元計測処理を実施することができるので、制御装置２の処理負担が削減される。したがって処理速度の遅い安価なシステム構成であっても、車輌速度の計測処理を高速化することが可能となる。
【００５２】
【数５】

【００５３】
【数６】

【００５４】
【数７】

【００５５】
以下同様にして、画像入力の都度、カメラ１ａからの入力画像を用いて、車輌の新たな移動位置が認識されるもので、速度計測部１５は、各車輌毎に、１段階前の車輌位置（Ｘ_Ｔ−１，Ｙ_Ｔ−１，Ｚ_Ｔ−１）と今回求められた車輌位置（Ｘ_Ｔ，Ｙ_Ｔ，Ｚ_Ｔ）とを、つぎの（８）式にあてはめて、車輌の速度Ｓを算出する。
【００５６】
【数８】

【００５７】
このように、入力画像上の車輌位置を時間軸上で正しく対応づけすることにより、対応づけられた各点の示す３次元座標をもって実際の空間における車輌の動きを正しくとらえることができるので、各車輌の速度を精度良く算出することができる。
【００５８】
なおここでは説明を簡単にするために、車輌位置として、車輌の先頭位置を示す１特徴点を抽出し、この先頭位置の変化により速度を算出しているが、認識精度を向上させるには、車輌のフロント部など所定の特徴形状を表す複数個の特徴点を認識対象とした方が良い。この場合、一方のカメラ１ａからの毎時の入力画像を用いて認識対象の各特徴点の位置を時間軸上で追跡し、その移動位置を３次元座標に変換して各特徴点毎の速度を算出した後、算出された各速度の平均値などをもって車輌の移動速度とすることになる。
【００５９】
また上記の方法では、まず３次元計測が可能な領域内に入った直後の車輌を認識して、その車輌の空間内の車輌位置を特定した後、順次同一車輌の移動位置を追跡するようにしているが、これに代えて、空間内を同じ速度で移動する物点を同一車輌として認識し、その共通の速度をもって車輌の移動速度とすることも可能である。
【００６０】
すなわち画像入力毎に、各入力画像Ｉ_Ｕ，Ｉ_Ｌ上のすべての特徴点について、前記した画像間の対応づけ処理を行った後、対応する特徴点の組毎に３次元座標値を求めておく。ついで一方の入力画像Ｉ_Ｕ上で、１フレーム前の入力画像上の各特徴点と今回の入力画像上の特徴点とを対応づけし、これら時間軸上で対応する特徴点の組毎に、それぞれの３次元座標の算出値を前記（８）式にあてはめて、空間内を移動する各物点の速度を算出する。この後、各物点を、それぞれの３次元座標と移動速度とを用いて、所定距離範囲内にありかつ所定の誤差範囲内の速度で移動する物点毎にグループ分けすることにより、各車輌の物点を切り分けて認識できるのである。
【００６１】
なおこのように画像上の各物点の速度を求める場合にも、各入力画像上に新たに出現した物点についてのみ、各入力画像間で特徴点を対応づけした後、その結果を前記（２）〜（４）にあてはめて３次元座標を算出し、その後は、一方の画像上で特徴点の位置を追跡して、その追跡結果を前記（５）〜（７）式にあてはめることにより３次元座標を算出するようにすれば、速度計測処理を高速化することができる。
【００６２】
また上記の移動速度に代えて、各物点毎に、その３次元座標の推移から空間内における移動ベクトルを特定するようにしてもよい。この場合、各物点は、各移動ベクトルの方向および長さを比較することにより、所定距離範囲内にあり、かつ空間内を同じ方向に同じ速度で移動する物点毎に切り分けて認識される。これにより交差点など、車輌の移動方向が複数方向に分岐する地点でも、各車輌の進行方向と移動速度とを、精度良く計測することができる。
なお前記したように、車輌上の各物点の高さは、移動に関わらず常に一定であるから、前記移動ベクトルの比較は、Ｘ軸方向とＺ軸方向においてのみ行えばよい。
【００６３】
つぎに、車輌速度計測装置の他の例を説明する。
この速度計測装置は一車輌に取り付けられて、その車輌や前方車輌の速度を算出するためのものである。
【００６４】
図１０は、速度計測装置の設置例を示すもので、車輌１９の前面位置に、２台のカメラ２０ａ，２０ｂが収容されたカメラボックス２１が取り付けられるとともに、車体内部にこれらカメラからの画像を処理して速度計測を実施するための制御装置２２が配備される。なお図中、２５は、計測結果などを送信するためのアンテナである。
【００６５】
各カメラ２０ａ，２０ｂは、前記実施例と同様、光軸を平行かつ撮像面をカメラボックス２１外の同一平面上に位置させた状態で、カメラボックス２１内に固定配備されており、車輌走行中に同一タイミングで動作して、車輌１９の前方位置の画像データを生成する。
【００６６】
図１１は、いずれか一方のカメラからの入力画像を示すもので、車輌の前方位置にある静止物体２４や、前方車輌の画像２３が現れている。
制御装置２２は、このような画像上で、静止物体２４や車輌２３を表す特徴点を抽出し、これら特徴点の座標により前記した実施例と同様の３次元計測処理や速度計測処理を実施して、空間内の各物点毎に自車輌から見た移動速度を算出する。
【００６７】
上記のように、道路上を走行する車輌から前方位置を撮像して、その画像上のエッジを抽出すると、図１２に示すように、前記静止物体２４や道路の境界ラインなど、空間内の位置が不動の物点にかかる特徴点が、多数、抽出される。また前記車輌よりこれら静止物体の特徴点を観測して得られた移動速度は、自車輌の移動速度に相当するものと考えられる。
【００６８】
制御装置２２は、上記原理に基づき、前記抽出された各物点をその速度に応じてクラスタリングし、最も構成点の多いクラスタの速度を、静止物体の移動速度、すなわち自車輌の速度として認識する。またその他のクラスタに分類された特徴点は、空間内の位置データに基づいてさらにグループ化され、各グループ毎の速度により、自車輌の前方にある車輌の各速度が切り分けて認識される。なおこの前方車輌の速度は、自車輌に対する相対速度として算出されるものであって、実際の速度が自車輌より遅い場合は負の値をとり、自車輌より速い速度の場合は正の値をとる。
【００６９】
このような方法により、前方車輌の速度を精度良く計測することが可能となるので、車間距離や衝突の危険度の認識を正確に行うことができ、高速道路における車輌の走行速度の自動制御などに有効な速度計測装置を提供することができる。
【００７０】
なお上記の各実施例は、いずれも車輌を計測対象としたが、この発明は、これに限らず、セキュリティシステムにおける人間の移動速度や、工場の製造ラインや物流システムにおける物体の移動速度を計測する際にも、適用することができる。
【００７１】
【発明の効果】
請求項１，４の発明では、３次元座標の算出値を時間軸上で対応づけて物点毎の移動速度を求めた後、これら物点を、所定の距離範囲内の空間を同じ速度で移動するもの毎にグループ分けすることにより、通過経路上の移動体の速度を認識するので、通過経路上に複数の移動体が存在しても、各移動体を精度良く切り分けて、その移動速度を正確に計測することができる。
【００７２】
請求項２および５の発明では、各物点に対し、前記移動速度に代えて空間内における移動ベクトルを特定した後、各物点を、所定の距離範囲内の空間を同じ方向に同じ速度で移動する物点毎にグループ分けし、各移動体の移動速度および進行方向を認識するようにしたから、通過経路が複数方向に分岐するような地点においても、各移動体を精度良く切り分けて、それぞれの移動速度を認識することが可能となる。
【００７３】
請求項３および６の発明では、移動体側に配備された複数の撮像手段により移動体の外側の所定方向を撮像し、３次元計測処理や時間軸上での３次元座標の対応づけ処理を経て物点毎の移動速度を求めた後、移動速度に基づくグループ分けにより静止物体に対応する物点のグループを抽出して、そのグループの移動速度を移動体の速度として認識するから、移動体の速度を精度良く計測することができる。
【図面の簡単な説明】
【図１】この発明の一実施例にかかる車輌速度計測装置の設置例を示す斜視図である。
【図２】図１の車輌速度計測装置の構成を示すブロック図である。
【図３】入力画像の一例を示す説明図である。
【図４】図３の各入力画像より生成されたエッジ画像を示す説明図である。
【図５】検出エリアの設定例を示す説明図である。
【図６】各画像間における特徴点の対応づけ処理の具体例を示す説明図である。
【図７】三角測量の原理を示す説明図である。
【図８】３次元計測処理にかかる各座標系の位置関係を示す説明図である。
【図９】入力画像上で車輌の移動位置を抽出する方法を示す説明図である。
【図１０】第２の車輌速度計測装置の設置状態を示す斜視図である。
【図１１】図１０の車輌速度計測装置における入力画像の例を示す説明図である。
【図１２】図１１の入力画像から生成されたエッジ画像を示す説明図である。
【符号の説明】
１ａ，１ｂ，２０ａ，２０ｂ カメラ
２，２２ 制御装置
１０ エッジ抽出部
１１ 対応づけ処理部
１２ ３次元計測部
１３ 車輌認識部
１４ 追跡処理部
１５ 速度算出部[0001]
[Industrial application fields]
The present invention relates to a technique for measuring the speed of a moving body that passes a predetermined passage route by using an image processing technique, and in particular, the three-dimensional measurement processing using a plurality of imaging means. Track changes in spatial position and measure speedTechnologyAbout.
[0002]
[Prior art]
As a device for measuring the traffic flow on the road, the applicant has recently recognized the position of the vehicle on the road using the three-dimensional recognition and pattern matching techniques, and the speed of the vehicle is A method for recognizing the number of passing vehicles was proposed (Japanese Patent Laid-Open No. 9-33232). In this method, the observation position on the road is imaged by a plurality of television cameras, the vehicle feature points extracted on the images from each camera are associated between the images, and the three-dimensional measurement process is performed. A three-dimensional measurement result is projected on a virtual plane along the road, and a matching process using a plurality of types of two-dimensional models is performed on the projected portion to recognize the position of each vehicle. Furthermore, the movement of each vehicle is tracked and the speed is measured by associating on the time axis the vehicle position recognized at each image capture time.
[0003]
[Problems to be solved by the invention]
Specifically, the correspondence of the vehicle positions on the time axis is the closest position among the detected positions obtained one step before the vehicle position detected by the matching process, and the distance. Is associated with a detection position within a predetermined threshold. However, in this method, since the association is performed using only the distance between each vehicle detection position on the time axis without considering the recognition result of the vehicle type by the matching process, there is a possibility that the association is erroneous. There is a problem that the accuracy of the measured value of the speed is deteriorated.
[0004]
The present invention has been made paying attention to the above problems, and after associating feature points representing the same object points between images obtained at predetermined time intervals, the corresponding feature points on these time axes. It is a technical subject to accurately measure the moving speed by correctly capturing the movement of the moving body using the three-dimensional coordinates indicated by
[0005]
[Means for Solving the Problems]
Claims 1, 2Each of the inventions relates to a method of measuring a speed of a moving body on the passage path by arranging a plurality of imaging sections toward the passage path of the moving body, capturing image data from each imaging section.
[0006]
The speed measurement method according to the first aspect of the present invention relates to the feature points representing the same object point of the moving body between the images captured at the same timing from the respective imaging means, and obtains the three-dimensional coordinates of the object point. In addition to calculating, at least one of the imaging means associates a feature point representing the same object point of the moving object between images captured at a predetermined time interval.AfterUsing these three-dimensional coordinates indicated by the corresponding feature points on the time axis, the speed of the moving object is determined.measure. Further, using the three-dimensional coordinates of each object point and the moving speed, each object point is grouped into objects that move in the space within a predetermined distance range at the same speed, and each movement on the passage route is performed. Recognize body speed.
[0007]
In the speed measurement method according to the invention of claim 2, after performing the same three-dimensional calculation process as that of the invention of claim 1 and the process of associating feature points on the time axis,Instead of the moving speed, a moving vector of each object point in the space is specified. And using the three-dimensional coordinates of each object point and the specified movement vector,Each object point is grouped for each object moving in the same direction at the same speed in a space within a predetermined distance range, and the speed and traveling direction of each moving object on the passage route are recognized.
[0008]
Claim 3According to the present invention, a plurality of imaging means are arranged on a moving body that passes through a predetermined passage, respectively, in a predetermined direction outside the moving body, and image data from each imaging means is captured and the movement is performed. A method for measuring the speed of the body between images captured at the same timing from each imaging means.Corresponding feature points representing the same object point,After calculating the three-dimensional coordinates of each object point and associating feature points representing the same object point between images captured at a predetermined time interval for at least one of the imaging means, these times Of the corresponding feature point on the axisEach three-dimensional coordinateUsingMeasure the speed of each object point. Furthermore, using the three-dimensional coordinates of each object point and the moving speed, each object point is grouped into those that move at the same speed, and the speed of the group with the most constituent points is recognized as the speed of the moving object. To do.
[0009]
Claim 4The speed measurement apparatus according to the invention is an apparatus for carrying out the method of claim 1 and inputs a plurality of imaging means arranged toward the passage of the moving body and images from each imaging means. Image input means for each time, and each time an image is input from the image input means, a feature point extraction means for extracting a feature point representing a moving body on each input image, and between each image input at the same timing, Extracted by the feature point extraction meansThe same object point from each feature pointAssociation means for associating feature points to be represented;By the association meansA predetermined time interval is set for at least one of the three-dimensional coordinate calculating means for calculating the three-dimensional coordinates of the object point represented by the feature points and the at least one imaging means using the two-dimensional coordinates of each feature point associated with each other. Second associating means for associating feature points representing the same object points between the captured images, and the three-dimensional coordinates of the feature points associated on the time axis by the second associating means Speed calculating means for calculating the speed of the moving body using the calculated value;By using the three-dimensional coordinates of each object point and the calculated value of the moving speed, the object points are grouped for each object moving at the same speed in a space within a predetermined distance range, and on the passing route. Speed recognition means for recognizing the speed of each moving body.
[0010]
The device of the invention of claim 5 is the device of claim 4.Instead of the speed calculation means, using the calculated values of the respective three-dimensional coordinates of the feature points associated on the time axis by the second association means,For each object point, the object point in spaceA movement vector specifying means for specifying a movement vector is provided. AndSpeed recognition meansEach object pointGroup the space within a predetermined distance range into the ones that move in the same direction at the same speed,It is comprised so that the speed of each moving body on a passage route may be recognized.
[0011]
Claim 6In the invention ofClaim 3As an apparatus for carrying out the above method, the apparatus is attached to a moving body that moves on a predetermined passage route in a predetermined direction outside the moving body.MultipleEach time an image is input from the image input means, an image input means for inputting an image from each image pickup means, and the image input means,Feature point extraction means for extracting feature points on each input image;Correspondence means for associating feature points representing the same object point of the object among the feature points extracted by the feature point extraction means between the images input at the same timing;By the association meansA predetermined time interval is set for at least one of the three-dimensional coordinate calculating means for calculating the three-dimensional coordinates of the object point represented by the feature points and the at least one imaging means using the two-dimensional coordinates of each feature point associated with each other. Second associating means for associating feature points representing the same object points between the captured images, and the three-dimensional coordinates of the feature points associated on the time axis by the second associating means Calculate the moving speed of each object point using the calculated valueUsing the speed calculation means, the three-dimensional coordinates of each object point and the calculated value of the moving speed, each object point is grouped for each object moving at the same speed, and the speed of the group with the most constituent points is Speed recognition means for recognizing the speed of a moving objectThe apparatus which comprises is comprised.
[0012]
[Action]
In the inventions of claims 1 and 4,Each time an image is captured from each imaging means, the three-dimensional coordinates of the object points that appear in common on these images are calculated. At the same time, feature points representing the same object point are associated with each other on the image obtained at the same imaging means at a predetermined time interval, and each object point is represented by the three-dimensional coordinates indicated by the associated feature points. Is calculated. And these things,Space within a predetermined distance rangeMove at the same speedthingBy grouping each group, moving objects on the passage routeCan recognize speed.
Object points on the same moving objectA space within a given distance range at the same speedSince it should move, according to this invention, even if there are a plurality of moving bodies on the passage route, each moving body can be accurately separated and its moving speed can be accurately measured.
[0013]
Claims 2 and 5In the present invention, using the three-dimensional coordinates indicated by the feature points associated on the time axis,For each object point, the object point in spaceAfter identifying the movement vector, each object point isSpace within distance rangeMove in the same direction and at the same speedthingGroup by group and move speed of each moving objectAnd direction of travelTherefore, even at points where the passage route branches in a plurality of directions, each moving body can be accurately separated and the moving speed can be recognized.
[0014]
Claims 3 and 6In the invention,Deployed on mobile sidepluralPredetermined direction outside the moving body by the imaging meansEach image is considered to include a large number of feature points related to an object point whose position in the space does not move, such as a stationary object in the vicinity of the movement path and a boundary line of the movement path. Also, since the object points applied to these stationary objects move at the same speed as viewed from the moving object, when each object point is grouped based on the moving speed, the group corresponding to the stationary object has the most constituent points. Can be thought of as a group.
[0015]
Therefore, the movement speed of each object point is calculated through the calculation process of three-dimensional coordinates using each image and the object point correspondence process on the time axis, and each object point is moved at the same speed. By grouping each group and recognizing the speed of the group having the most constituent points as the speed of the moving body, the speed of the moving body can be measured with high accuracy.
[0016]
【Example】
FIG. 1 shows an installation example of a vehicle speed measuring device according to an embodiment of the present invention.
This vehicle speed measuring device is configured by arranging an F-shaped support column 3 in the vicinity of a road RD and attaching two cameras 1a, 1b and a control device 2 to the support column 2. The image obtained by imaging the road RD from above with 1a, 1b is taken into the control device 2, the speed of the passing vehicle is measured for each roadway of the road RD, and the measurement result is sent to the outside such as a management center. Output.
[0017]
The said support | pillar 3 is arrange | positioned by the horizontal rail projecting on the road. Each camera 1a, 1b has a lens having the same focal length, and the optical axis is tilted downward at a predetermined position above the road 1 by a support member 5 fixedly disposed between the cross rails of the column 2. It is installed vertically in the direction of In this embodiment, the optical axes of the cameras 1a and 1b are parallel to the direction of the road and the imaging surfaces are on the same plane in order to easily perform the association processing and the three-dimensional measurement processing described later. Each attachment position is adjusted so that it may be located in.
On the other hand, the control device 2 is attached in the vicinity of the base portion of the support column 2 for maintenance and inspection.
[0018]
In this embodiment, the road RD is imaged by the two cameras 1a and 1b. However, the present invention is not limited to this, and three or more cameras may be used. Further, the arrangement of the cameras 1a and 1b is not limited to the vertical arrangement, but may be the horizontal arrangement. Further, the column for attaching the cameras 1a and 1b is not limited to the F-shaped column 3 described above, and an existing telegraph pole or illumination column may be modified and used.
[0019]
FIG. 2 shows the configuration of the vehicle speed measuring device.
The control device 2 includes two A / D conversion units 6a and 6b, an image memory 7, a memory 8 and a CPU 9 for controlling the entire device, an edge extraction unit 10, an association processing unit 11, and a three-dimensional measurement unit 12. The vehicle recognition unit 13, the tracking processing unit 14, and the speed measurement unit 15 are connected to each other via an image / data bus 16.
[0020]
The A / D converters 6a and 6b convert the analog image signals from the cameras 1a and 1b into digital image signals. The edge extraction unit 10 extracts edge composing points indicating the outline of an object such as a vehicle on the input image after the digital conversion, and generates a binary edge image having each edge composing point as a black pixel. These input images and edge images are stored in individual storage areas in the image memory 7, respectively.
[0021]
The association processing unit 11 associates feature points representing the same object point in the space on each image with respect to edge composing points (hereinafter referred to as “feature points”) on the respective edge images. The three-dimensional measurement unit 12 calculates the three-dimensional coordinates of the object point represented by each feature point by applying each two-dimensional coordinate to the triangulation principle for each pair of feature points associated with each other.
[0022]
The feature point association process and the three-dimensional coordinate measurement process are performed on a feature point indicating a vehicle that has newly appeared in the observation region, as will be described later. The vehicle recognition unit 13 individually recognizes the position of each vehicle using these three-dimensional coordinates, and the tracking processing unit 14 tracks the change in the recognized vehicle position every time an image is captured thereafter. The recognition result of the vehicle position at each stage is represented by three-dimensional coordinates of a predetermined representative point such as an object point representing the head of the vehicle, and the speed measurement unit 15 uses the change in the vehicle position for each vehicle. And the calculation result is output to the outside via a transmission unit (not shown).
[0023]
Next, a detailed processing procedure in the control device 2 will be described step by step.
The cameras 1a and 1b are controlled so that the imaging operations are synchronized by a timing control unit (not shown), and image data for each frame is digitally converted by the A / D conversion units 6a and 6b, Stored in the memory 7.
[0024]
The edge extraction unit 10 scans an input image from each of the cameras 1a and 1b with an edge extraction filter such as a Laplacian filter to extract a zero crossing point. Further, the edge extraction unit 10 calculates the coordinates of the zero crossing point, and generates a binary edge image in which the pixel at the coordinate position is a black pixel and the other pixels are white pixels.
[0025]
3 (1) and 3 (2) are specific examples of images input from the cameras 1a and 1b, and image data including images such as lanes on the road, vehicles, and vehicle shadows are generated. .
[0026]
4 (1) and 4 (2) show the input images I shown in FIGS. 3 (1) and 3 (2)._U, I_LEdge images E obtained by performing edge extraction processing_U, E_LThe edge components corresponding to the contour of the vehicle and the shadow of the vehicle are extracted.
[0027]
When the above edge extraction processing is completed, the current input image I_U, I_LProcessing for recognizing the initial position of a vehicle that has newly appeared on the top and processing for recognizing the movement position of a vehicle that has already been recognized by past processing are performed.
[0028]
First, a new vehicle recognition process will be described. The feature points to be subjected to the recognition process are determined based on the traveling direction of the vehicle in each lane. The vehicle recognizing unit 13 reads each image I_U, I_LOn the other hand, as shown in FIG. 5, a detection area r of a predetermined size is located at the position where the vehicle immediately after entering the region where the visual fields of the cameras 1a and 1b overlap each other appears._U1, R_U2, R_L1, R_L2Each detection area r_U1, R_U2, R_L1, R_L2For each feature point extracted in each area, feature points representing the same object point in the space are specified between the images, and the two are associated with each other.
[0029]
Specifically, this association processing is performed by any one of the edge images (here, the edge image E corresponding to the upper camera 1a)._UThe detection area r above_U1, R_U2The other edge image E with respect to a predetermined feature point inside_LThe corresponding points in the above are extracted, and FIGS. 6 (1) to 6 (4) show a specific method of the correspondence processing.
[0030]
The illustrated example shows the detection area r_U1The feature point Q corresponding to the feature point P in the image is specified._LAn epipolar line L for the feature point P is set above, and a detection area r_L1Each feature point Q located on this epipolar line L within₁~ Q₅Are extracted as corresponding candidate points of the feature point P.
In this case, since the cameras 1a and 1b are arranged vertically as described above, the epipolar line L is perpendicular to the x-axis, and the corresponding candidate points can be easily extracted.
[0031]
Next, the association processing unit 11 performs the edge image E._UInput image I corresponding to_UThe position of the upper point P (x, y_U), A window W having a predetermined size centered on this point P._USet. The edge image E_LThe second input image I corresponding to_LThe corresponding candidate point Q of₁~ Q₅Each point has its point Q₁~ Q₅And the window W_UWindow W having the same size as₁~ W₅Set.
[0032]
When each window is set, the association processing unit 11 inputs the input image I._LEach window W above₁~ W₅The following equation (1) is executed for each window and each window and the window W_UThe difference DF of the image data is calculated.
In the following formula, g_U(X, y) is the window W_UThe luminance value of a predetermined pixel in_L(X, y) is the window W_LThe luminance value of a predetermined pixel in (L = 1 to 5), SZ indicates the size of the window. Further, i and j are variables for specifying the pixel of interest in each window by varying within the range of 0 to SZ.
[0033]
[Expression 1]

[0034]
The association processing unit 11 displays each window W₁~ W₅Window W asked for_UThe difference DF is compared with the window W, and the window with the smallest difference is the window W._UIt is determined that it corresponds to. And the center point Q of the window_L(L = 2 in the illustrated example) is determined as the corresponding point of the feature point P.
[0035]
Each window W instead of the above difference₁~ W₅Every window W_UThe window where the highest correlation value is obtained is obtained by performing the normalized cross-correlation operation between_UYou may make it discriminate | determine from what corresponds.
[0036]
When the above association processing is performed for all extracted feature points, the three-dimensional measurement unit 12 uses the two-dimensional coordinates of these feature points as the principle of triangulation for each of the corresponding feature points P and Q. By fitting, the three-dimensional coordinates of the corresponding object point are calculated.
[0037]
FIG. 7 shows the principle of triangulation described above.
In the figure, R indicates a predetermined component point of the vehicle 16 on the road, and the imaging plane G of each camera 1a, 1b._U, G_LThe object point images P and Q of this point R appear on the top.
In the figure, C_UIs the focus of the camera 1a, C_LIndicates the focus of the camera 3b.
[0038]
The object image P, Q is the edge image E_U, E_LThe three-dimensional coordinates corresponding to each point correspond to the spatial position of the object point R. Therefore, for the representative object points of this object, the image formation position on the image of the object point is specified, and by calculating the 3D coordinates of each object point using the 2D coordinates of these image formation positions, The three-dimensional shape and spatial position of the object can be grasped.
[0039]
FIG. 8 (1) shows a spatial coordinate system (hereinafter referred to as “stereo coordinate system”) determined by the positional relationship between the cameras 1a and 1b, and FIG. 8 (2) shows each object in the actual space. The positional relationship between the spatial coordinate system for representing the position of the point and the stereo coordinate system is shown.
[0040]
In FIG. 8 (1), X_UY_UZ_U, X_LY_LZ_LEach spatial coordinate system represented by represents the camera coordinate system of each camera 1a, 1b, and the origin C of these camera coordinate systems._U, C_LA stereo coordinate system is set with the center point O ′ of the base line connecting (the focal position) as the origin. As described above, since the cameras 1a and 1b have the optical axes arranged in parallel, the axes X of the camera coordinate system and the stereo coordinate system are used._UY_UZ_U, X_LY_LZ_L, X′Y′Z ′ are parallel to each other.
[0041]
Each imaging surface G_U, G_LThe two-dimensional coordinates above are Z in the corresponding camera coordinate system._U, Z_LThe intersection point with the axis is the origin o_U, O_LX in the horizontal direction of the imaging surface_U, X_LThe axis is y in the vertical direction_U, Y_LAn axis is defined. As shown in FIG. 8 (2), the coordinate system representing the actual spatial position is defined by the origin O as the perpendicular foot drawn from the origin O 'of the camera coordinate system onto the road surface, and the width direction of the road as the X axis, The length direction is determined as the Z axis, and the height direction as the Y axis.
[0042]
In such a coordinate system setting, the two-dimensional coordinates of the associated feature points P and Q are expressed as (x, y_U), (X, y_L), The angle of depression of each camera 3a, 3b is θ, and the origin C of each camera coordinate system_U, C_LAssuming that the distance between the points B and the distance between the origins O and O ′ is H, the three-dimensional coordinates (X, Y, Z) of the object point R represented by the feature points P and Q are given by (2) It is calculated by the formula (4).
[0043]
[Expression 2]

[0044]
[Equation 3]

[0045]
[Expression 4]

[0046]
In this way, the detection area r_U1, R_U2, R_L1, R_L2When the three-dimensional coordinates corresponding to all the feature points extracted in the vehicle are calculated, the vehicle recognition unit 13 detects each detection area r._U1, R_U2, R_L1, R_L2Each time, the calculated three-dimensional coordinates are clustered to group the coordinates for each vehicle. And the coordinate which shows the head position of a vehicle is extracted from this group, and let the extracted position be a vehicle position. The three-dimensional coordinates of the object point extracted as the vehicle position are stored in the memory 8, and thereafter, each time an image is input, the three-dimensional coordinates indicating the moving position of the vehicle at that time are sequentially accumulated. Trajectory data is generated.
[0047]
The vehicle recognition process is not limited to the clustering process, and as described in Japanese Patent Laid-Open No. 9-33332 described above, each three-dimensional coordinate is assumed to be a virtual vertical along the road length direction (that is, the Z-axis direction). A method of projecting on a plane and collating the projection result with a two-dimensional model of the vehicle may be used.
[0048]
Next, a process for calculating the speed by tracking the moving position of a vehicle that has already been recognized by the past process will be described.
If the initial position or the moving position of the vehicle is recognized at a certain point in time, either one of the input images (here, the input image I obtained by the camera 3a)_UAnd the coordinates of the feature point corresponding to the recognized vehicle position (that is, the feature point indicating the head position of the vehicle), and image data (hereinafter referred to as image data) in an image area having a predetermined size centered on the feature point. This is referred to as “feature image M”) and is stored in the memory 8 as data for tracking processing. When the next image is input after the elapse of the predetermined time T, the tracking processing unit 14 reads the input image I from the camera 1a at this time._UTIn the above, after specifying the moving position of the vehicle recognized one frame before, the specified two-dimensional vehicle position is converted into three-dimensional coordinates.
[0049]
Figure 9 shows the input image_UTA specific method for specifying the moving position of the vehicle is shown above. This input image I_UTIs the input image I in FIG._UThe point A in the figure is obtained after a predetermined time has elapsed from the input of the input image I in FIG._UThe position specified as the vehicle position above is shown. The tracking processing unit 14 sets a scanning area 17 having a predetermined size based on the coordinate position of the point A, and then scans the feature image M in the scanning area, and performs the association processing for each scanning position. The same equation (1) is executed. As a result, the point A ′ corresponding to the center point of the feature image M at the scanning position where the dissimilarity DF is the smallest is recognized as the moving position of the vehicle.
[0050]
Input image I_UTWhen the point A ′ indicating the moving position of the vehicle is specified above, the tracking processing unit 14 then determines the two-dimensional coordinates (x_T, Y_T) Is applied to the equations (5) to (7), and the three-dimensional coordinates of the object point corresponding to the point A ′, that is, the moving position (X_T, Y_T, Z_T) Is calculated and added to the trajectory data in the memory 8.
[0051]
The equations (5) to (7) are made by paying attention to the fact that the height data of the vehicle in the space does not fluctuate even if the vehicle moves, and α in the equations (5) and (7)_x, Α_yIndicates the length of one pixel in the x and y axis directions on the input image. Y in each formula₀Is the Y coordinate value of the vehicle initial position obtained when the image of the vehicle enters the detection area, that is, when the three-dimensional measurement of the vehicle can be recognized. According to this method, the input image I from the other camera 1b is obtained for the point A ′._LSince the three-dimensional measurement process can be performed without extracting the corresponding points above, the processing load on the control device 2 is reduced. Therefore, even with an inexpensive system configuration with a low processing speed, it is possible to speed up the vehicle speed measurement process.
[0052]
[Equation 5]

[0053]
[Formula 6]

[0054]
[Expression 7]

[0055]
In the same manner, each time an image is input, a new moving position of the vehicle is recognized using the input image from the camera 1a, and the speed measuring unit 15 determines the position of the vehicle one step ahead for each vehicle. (X_T-1, Y_T-1, Z_T-1) And the vehicle position (X_T, Y_T, Z_T) Is applied to the following equation (8) to calculate the vehicle speed S:
[0056]
[Equation 8]

[0057]
Thus, by correctly associating the vehicle position on the input image on the time axis, it is possible to correctly grasp the movement of the vehicle in the actual space with the three-dimensional coordinates indicated by the associated points. The speed of the vehicle can be calculated with high accuracy.
[0058]
In order to simplify the explanation, one feature point indicating the head position of the vehicle is extracted as the vehicle position, and the speed is calculated based on the change in the head position. To improve the recognition accuracy, It is better to use a plurality of feature points representing a predetermined feature shape such as the front part of the vehicle as recognition targets. In this case, the position of each feature point to be recognized is tracked on the time axis using the hourly input image from one camera 1a, the moving position is converted into three-dimensional coordinates, and the speed for each feature point is obtained. After the calculation, the average value of the calculated speeds is used as the moving speed of the vehicle.
[0059]
In the above method, the vehicle immediately after entering the region where the three-dimensional measurement can be performed is recognized, the vehicle position in the space of the vehicle is specified, and then the movement position of the same vehicle is sequentially tracked. However, instead of this, it is also possible to recognize object points that move in the space at the same speed as the same vehicle, and use the common speed as the moving speed of the vehicle.
[0060]
That is, for each image input, each input image I_U, I_LAfter all the above feature points have been subjected to the above-described correlation process between images, a three-dimensional coordinate value is obtained for each set of corresponding feature points. Then one input image I_UIn the above, each feature point on the input image one frame before is associated with the feature point on the current input image, and each three-dimensional coordinate is calculated for each set of corresponding feature points on the time axis. The value is applied to the equation (8) to calculate the speed of each object point moving in the space. Thereafter, each object point is grouped into object points that move within a predetermined distance range and at a speed within a predetermined error range by using the respective three-dimensional coordinates and moving speeds. It is possible to recognize the object points by separating them.
[0061]
In addition, also when calculating | requiring the speed of each object point on an image in this way, after associating the feature point between each input image only about the object point which newly appeared on each input image, the result is said (( By applying to 2) to (4) to calculate the three-dimensional coordinates, and then tracking the position of the feature point on one image and applying the tracking result to the above formulas (5) to (7) If the three-dimensional coordinates are calculated, the speed measurement process can be speeded up.
[0062]
Further, instead of the above moving speed, a moving vector in the space may be specified for each object point from the transition of the three-dimensional coordinates. In this case, by comparing the direction and length of each movement vector, each object point is recognized by being separated for each object point that is within a predetermined distance range and moves in the same direction at the same speed in the space. . As a result, the traveling direction and the moving speed of each vehicle can be accurately measured even at a point such as an intersection where the moving direction of the vehicle branches in a plurality of directions.
As described above, since the height of each object point on the vehicle is always constant regardless of the movement, the movement vectors may be compared only in the X-axis direction and the Z-axis direction.
[0063]
Next, another example of the vehicle speed measuring device will be described.
This speed measuring device is attached to one vehicle and is used for calculating the speed of the vehicle and the preceding vehicle.
[0064]
FIG. 10 shows an installation example of the speed measuring device. A camera box 21 containing two cameras 20a and 20b is attached to the front position of the vehicle 19, and images from these cameras are displayed inside the vehicle body. A control device 22 is provided for processing and performing speed measurement. In the figure, reference numeral 25 denotes an antenna for transmitting measurement results and the like.
[0065]
Each camera 20a, 20b is fixedly arranged in the camera box 21 with the optical axis parallel and the imaging surface located on the same plane outside the camera box 21 as in the above-described embodiment, and the vehicle is running. The image data of the front position of the vehicle 19 is generated at the same timing.
[0066]
FIG. 11 shows an input image from one of the cameras, and a stationary object 24 at the front position of the vehicle and an image 23 of the front vehicle appear.
The control device 22 extracts feature points representing the stationary object 24 and the vehicle 23 on such an image, and performs the same three-dimensional measurement process and speed measurement process as the above-described embodiment based on the coordinates of these feature points. Thus, the moving speed viewed from the own vehicle is calculated for each object point in the space.
[0067]
As described above, when a front position is imaged from a vehicle traveling on a road and an edge on the image is extracted, a position in space such as the stationary object 24 or a road boundary line as shown in FIG. A large number of feature points related to object points that are not moving are extracted. Further, the moving speed obtained by observing the feature points of these stationary objects from the vehicle is considered to correspond to the moving speed of the own vehicle.
[0068]
Based on the above principle, the control device 22 clusters the extracted object points according to the speed, and recognizes the speed of the cluster with the most component points as the moving speed of the stationary object, that is, the speed of the own vehicle. . Further, the feature points classified into other clusters are further grouped based on the position data in the space, and the respective speeds of the vehicles ahead of the host vehicle are recognized by the speed of each group. The speed of the preceding vehicle is calculated as a relative speed with respect to the own vehicle, and takes a negative value when the actual speed is slower than the own vehicle, and takes a positive value when the actual speed is faster than the own vehicle. Take.
[0069]
Such a method makes it possible to accurately measure the speed of the vehicle ahead, so it is possible to accurately recognize the distance between vehicles and the risk of collision, and to automatically control the running speed of the vehicle on the highway. It is possible to provide an effective speed measuring device.
[0070]
In each of the above embodiments, the vehicle is the measurement target. However, the present invention is not limited to this, and the movement speed of a person in a security system and the movement speed of an object in a factory production line or a distribution system are measured. This can also be applied.
[0071]
【The invention's effect】
According to the first and fourth aspects of the present invention, after calculating the moving speed for each object point by associating the calculated values of the three-dimensional coordinates on the time axis, the object point is moved at the same speed in a space within a predetermined distance range. By grouping each moving object, the speed of the moving object on the passing route is recognized.Even if there are a plurality of moving bodies on the passage route, each moving body can be accurately separated and its moving speed can be accurately measured.
[0072]
Claims 2 and 5In this invention, after specifying a movement vector in space instead of the moving speed for each object point, each object point is moved for each object point moving in the same direction at the same speed in a space within a predetermined distance range. Divided into groups, and the moving speed of each moving objectAnd the direction of travelSince it did in this way, also at the point where a passage route branches in a plurality of directions, it becomes possible to distinguish each moving body with high accuracy and to recognize each moving speed.
[0073]
Claims 3 and 6In this invention, the predetermined direction outside the moving body is imaged by a plurality of imaging means arranged on the moving body side,After obtaining the moving speed of each object point through 3D measurement processing and 3D coordinate mapping processing on the time axis, a group of object points corresponding to a stationary object is extracted by grouping based on the moving speed. Because the movement speed of the group is recognized as the speed of the moving body,Speed can be measured with high accuracy.
[Brief description of the drawings]
FIG. 1 is a perspective view showing an installation example of a vehicle speed measuring device according to an embodiment of the present invention.
2 is a block diagram showing a configuration of the vehicle speed measuring device of FIG. 1;
FIG. 3 is an explanatory diagram illustrating an example of an input image.
4 is an explanatory diagram showing edge images generated from each input image of FIG. 3; FIG.
FIG. 5 is an explanatory diagram illustrating an example of setting a detection area.
FIG. 6 is an explanatory diagram illustrating a specific example of a process for associating feature points between images.
FIG. 7 is an explanatory diagram showing the principle of triangulation.
FIG. 8 is an explanatory diagram showing a positional relationship of each coordinate system related to a three-dimensional measurement process.
FIG. 9 is an explanatory diagram showing a method for extracting a moving position of a vehicle on an input image.
FIG. 10 is a perspective view showing an installation state of the second vehicle speed measuring device.
11 is an explanatory diagram showing an example of an input image in the vehicle speed measuring device of FIG.
12 is an explanatory diagram showing an edge image generated from the input image of FIG. 11. FIG.
[Explanation of symbols]
1a, 1b, 20a, 20b camera
2,22 Control device
10 Edge extractor
11 Association processing section
12 3D measurement unit
13 Vehicle recognition unit
14 Tracking processor
15 Speed calculator

Claims (6)

A method of deploying a plurality of imaging means toward a passing path of a moving body, capturing image data from each imaging means, and measuring a speed of the moving body on the passing path,
While associating feature points representing the same object point between the images captured at the same timing from each imaging means, calculating the three-dimensional coordinates of each object point,
After associating feature points representing the same object point between images captured at a predetermined time interval for at least one of the imaging means, each three-dimensional coordinate of the corresponding feature point on the time axis Measure the moving speed of each object point using
By using the three-dimensional coordinates of each object point and the moving speed, the object points are grouped into objects that move in the space within a predetermined distance range at the same speed, and each moving object on the passage route is grouped. A speed measurement method for a moving object, characterized by recognizing speed . A method of deploying a plurality of imaging means toward a passing path of a moving body, capturing image data from each imaging means, and measuring a speed of the moving body on the passing path,
While associating feature points representing the same object point between the images captured at the same timing from each imaging means, calculating the three-dimensional coordinates of each object point,
After associating feature points representing the same object point between images captured at a predetermined time interval for at least one of the imaging means, each three-dimensional coordinate of the corresponding feature point on the time axis For each object point , specify the movement vector of that object point in space ,
Using the three-dimensional coordinates of each object point and the specified movement vector, each object point is grouped into items that move in the same direction at the same speed in a space within a predetermined distance range, and the passage route A speed measurement method for a moving body, characterized by recognizing a speed and a traveling direction of each of the above moving bodies. A plurality of imaging means are arranged on a moving body that passes through a predetermined passing path in a predetermined direction outside the moving body, and image data from each imaging means is captured, and the speed of the moving body is set. A method of measuring,
While associating feature points representing the same object point between the images captured at the same timing from each imaging means, calculating the three-dimensional coordinates of each object point,
After associating feature points representing the same object point between images captured at a predetermined time interval for at least one of the imaging means, each three-dimensional coordinate of the corresponding feature point on the time axis Measure the moving speed of each object point using
Using the three-dimensional coordinates of each object point and the moving speed, group each object point according to what moves at the same speed, and recognize the speed of the group with the most constituent points as the speed of the moving object. A method for measuring the speed of a moving object characterized by the above . A plurality of imaging means arranged toward the passage of the moving body;
Image input means for inputting an image from each imaging means;
Feature point extracting means for extracting feature points on each input image each time an image is input from the image input means;
Correspondence means for associating feature points representing the same object points from among the feature points extracted by the feature point extraction means between the images input at the same timing;
Using the two-dimensional coordinates of each feature point associated with the association means, and a three-dimensional coordinate calculation means for calculating three-dimensional coordinates of the object point representing of these feature points,
A second associating means for associating feature points representing the same object points between images captured at a predetermined time interval with respect to at least one of the imaging means;
Speed calculating means for calculating the moving speed of each object point using the calculated value of each three-dimensional coordinate of the feature point correlated on the time axis by the second correlating means;
By using the three-dimensional coordinates of each object point and the calculated value of the moving speed, the object points are grouped for each object moving at the same speed in a space within a predetermined distance range, and on the passing route. velocity measuring device comprising; and a speed recognition means for recognizing the speed of each moving object. A plurality of imaging means arranged toward the passage of the moving body;
Image input means for inputting an image from each imaging means;
Feature point extracting means for extracting feature points on each input image each time an image is input from the image input means;
Correspondence means for associating feature points representing the same object points from among the feature points extracted by the feature point extraction means between the images input at the same timing;
Using the two-dimensional coordinates of each feature point associated with the association means, and a three-dimensional coordinate calculation means for calculating three-dimensional coordinates of the object point representing of these feature points,
A second associating means for associating feature points representing the same object points between images captured at a predetermined time interval with respect to at least one of the imaging means;
Movement vector specifying means for specifying, for each object point, a movement vector of the object point in space using the calculated values of the three-dimensional coordinates of the feature points associated on the time axis by the second association means When,
Using the three-dimensional coordinates of each object point and the specified movement vector, each object point is grouped into items that move in the same direction at the same speed in a space within a predetermined distance range, and the passage route velocity measuring device comprising; and a speed recognition means for recognizing the speed of each moving object above. A plurality of imaging means each attached to a moving body that moves on a predetermined passing path toward a predetermined direction outside the moving body;
Image input means for inputting an image from each imaging means;
Feature point extracting means for extracting feature points on each input image each time an image is input from the image input means;
Correspondence means for associating feature points representing the same object point of the object among the feature points extracted by the feature point extraction means between the images input at the same timing;
Using the two-dimensional coordinates of each feature point associated with the association means, and a three-dimensional coordinate calculation means for calculating three-dimensional coordinates of the object point representing of these feature points,
A second associating means for associating feature points representing the same object points between images captured at a predetermined time interval with respect to at least one of the imaging means;
Speed calculating means for calculating the moving speed of each object point using the calculated value of each three-dimensional coordinate of the feature point correlated on the time axis by the second correlating means;
Using the three-dimensional coordinates of each object point and the calculated value of the moving speed, each object point is grouped for each object that moves at the same speed, and the speed of the group with the most constituent points is set as the speed of the moving object. A speed measuring device comprising speed recognition means for recognizing .

Images