JP4337256B2 - Weather detection device, weather detection system, and weather detection method - Google Patents

Description

但し、ｆ（ｘ、ｙ）は計測領域中の画素の輝度値、Ｍは計測領域の横方向における画素数、Ｎは計測領域の縦方向における画素数である。ｕ、ｖは、−Ｍ／２≦ｕ≦Ｍ／２、−Ｎ／２≦ｖ≦Ｎ／２にそれぞれ設定される。
【００３４】
次いで、Ｓ１４に移行し、式（１）のパワースペクトルＦ（ｕ、ｖ）に基づいて、次の式（２）に示すように、パワースペクトルの強さ（振幅スペクトル）Ｐ（ｕ、ｖ）が算出される。
Ｐ（ｕ、ｖ）＝（Ｒ2＋Ｉ2）1/2 ‥‥（２）
但し、ＲはＦ（ｕ、ｖ）の実部であり、ＩはＦ（ｕ、ｖ）の虚部である。
【００３５】
そして、パワースペクトルの強さＰ（ｕ、ｖ）に基づいて、次の式（３）に示すように、計測領域におけるパワースペクトルの総和ＰＭの算出が行われる。
【数２】

但し、ＰＭは、Ｐ（０、０）を除いた値とする。
【００３６】
次いで、Ｓ１６に移行し、降雪状態の判定が行われる。降雪状態の判定は、算出されたパワースペクトルの総和ＰＭに基づいて行われる。
【００３７】
例えば、降雪モードを降雪強度の順に「強」、「中」、「弱」、「なし」として設定し、降雪モードの「強」と「中」のしきい値をＡ１、降雪モードの「中」と「弱」のしきい値をＡ２、降雪モードの「弱」と「なし」のしきい値をＡ３と設定する。この場合、しきい値Ａ１、Ａ２及びＡ３は、Ａ１＞Ａ２＞Ａ３と設定される。
【００３８】
そして、算出して得られたパワースペクトルの総和ＰＭがしきい値Ａ１以上であれば降雪モードが「強」と判定され、算出して得られたパワースペクトルの総和ＰＭがしきい値Ａ１より小さくしきい値Ａ２以上であれば降雪モードが「中」と判定され、算出して得られたパワースペクトルの総和ＰＭがしきい値Ａ２より小さくしきい値Ａ３以上であれば降雪モードが「弱」と判定され、算出して得られたパワースペクトルの総和ＰＭがしきい値Ａ３より小さいときには降雪モードが「なし」と判定される。
【００３９】
そして、Ｓ１８に移行し、判定された降雪モードの情報が外部出力部３４から外部装置４へ出力され、制御処理を終了する。
【００４０】
次に、本実施形態に係る気象検出装置、気象検出システム及び気象検出方法を用いて行った実際の気象検出結果について説明する。
【００４１】
図３〜図６に監視カメラ２の撮影映像に基づいて得られた画像データを示す。図３は、降雪状態が強い時の画像データである。図４は、降雪状態が中程度の時の画像データである。図５は、降雪状態が弱い時の画像データである。図６は、降雪していない時の画像データである。なお、図３〜図６の画像データは、横ｍ×縦ｎが３２０×２４０（７６８００点）のｂｍｐ画像である。また、図３〜図６において、黒い部分は、黒色の検出板２１である。
【００４２】
これらの画像データの検出板上の画素について、上述した図２のＳ１２〜Ｓ１４までの画像処理、即ちフーリエ変換処理を行った。その結果、図７〜図１４のフーリエ変換データを得た。
【００４３】
図７は、降雪状態が強い時の画像データをフーリエ変換して得られたものであって、それを立体的に示したものである。図７において、ｕ、ｖの横方向は周波数成分を示し、ＰＭの縦方向は各周波数成分における強度を示している。図８は、図７を平面的に示したものであり、降雪状態が強い時のフーリエ変換における周波数分布を示すものである。
【００４４】
図９は、降雪状態が中程度の時の画像データをフーリエ変換して得られたものであり、それを立体的に示したものである。また、図１０は、図９を平面的に示したものであり、降雪状態が中程度の時のフーリエ変換における周波数分布を示すものである。
【００４５】
図１１は、降雪状態が弱い時の画像データをフーリエ変換して得られたものであり、それを立体的に示したものである。また、図１２は、図１１を平面的に示したものであり、降雪状態が弱い時のフーリエ変換における周波数分布を示すものである。
【００４６】
図１３は、降雪なしの時の画像データをフーリエ変換して得られたものであり、それを立体的に示したものである。また、図１４は、図１３を平面的に示したものであり、降雪なしの時のフーリエ変換における周波数分布を示すものである。
【００４７】
なお、図７〜図１４は、ｕ、ｖの各周波数成分において、低周波成分を中心部分に分布させ高周波成分を４隅の部分に分布させて、シャッフリングしたものである。
【００４８】
図７及び図８の降雪状態が強い場合のフーリエ変換を見ると、低周波成分を中心としてパワースペクトル（ＰＭ）が大きくなっており、縦方向（ｖ）の周期性が強く示されている。
【００４９】
図９及び図１０の降雪状態が中程度の場合のフーリエ変換を見ると、低周波成分を中心としてパワースペクトル（ＰＭ）が大きくなっているが、縦方向（ｖ）の周期性がやや強く示されている。
【００５０】
図１１及び図１２の降雪状態が弱い場合のフーリエ変換を見ると、全体としてパワースペクトル（ＰＭ）が小さくなっている。
【００５１】
図１３及び図１４の降雪なしの場合のフーリエ変換を見ると、全体としてパワースペクトル（ＰＭ）が微少なものとなっている。
【００５２】
そして、各降雪状態におけるパワースペクトルの総和を算出すると、図１５に示すように、降雪状態が強い場合は１１８１２９１．１であり、降雪状態が中程度の場合は１０１５７２０．０であり、降雪状態が弱い場合は８８５１２５．７であり、降雪なしの場合は３９９３１６．７であった。
【００５３】
このように各降雪状態をパワースペクトルの総和の大小により判別することが可能である。この場合、降雪状態を判別するしきい値としては、例えば、降雪状態の「強」と「中」のしきい値Ａ１を１０５００００、降雪状態の「中」と「弱」のしきい値Ａ２を９５００００、降雪状態の「弱」と「なし」のしきい値Ａ３を４５００００とすれば、降雪状態をパワースペクトルの総和に基づいて判別することができる。
【００５４】
以上のように、本実施形態に係る気象検出システム、気象検出装置及び気象検出方法によれば、気象状況を検出すべき場所を撮影した画像データを画像処理することにより、降雪状況を検出することができる。このため、多数の場所の降雪状況を迅速に検出することができる。
【００５５】
また、画像処理としてフーリエ変換を行うことにより、画像データにおいて降雪による画像の周期性を検出し、その周期性の程度に基づいて降雪状態を判別することができる。従って、監視領域の周囲の明るさによる画像の明暗などに影響されることなく、降雪状態の判別が確実に行える。
【００５６】
また、フーリエ変換により得られたパワースペクトルの最大値でなく、パワースペクトルの総和に基づいて降雪状態を判別することにより、降雪状態が「強」と「中」の場合でも確実に判別することができる。
【００５７】
更に、本実施形態に係る気象検出システム、気象検出装置及び気象検出方法によれば、既存の監視カメラの映像を利用すれば、新たに監視カメラの設置工事などが必要なく、気象検出システムの導入が容易である。
【００５８】
なお、本実施形態では高速道路の降雪状況を検出する気象検出システム、気象検出装置及び気象検出方法について説明したが、本発明に係る気象検出システム、気象検出装置及び気象検出方法はこのようなものに限られるものではなく、その他の場所の降雪状況を検出するものであってもよい。
【００５９】
（第二実施形態）
次に、第二実施形態に係る気象検出装置、気象検出システム及び気象検出方法について説明する。
【００６０】
第一実施形態に係る気象検出装置、気象検出システム及び気象検出方法は監視領域の降雪状態を検出するものであったが、本実施形態に係る気象検出装置、気象検出システム及び気象検出方法は監視領域の降雪時の風力を検出するものである。
【００６１】
本実施形態に係る気象検出装置及び気象検出システムは、図１に示す第一実施形態の気象検出装置及び気象検出システムと同様なハード構成を有している。
【００６２】
図１６に気象検出装置等に係る気象検出処理のフローチャートを示す。
【００６３】
図１６のステップＳ２０にて、画像データの読み込みが行われる。画像データは、例えば、３ｓｅｃの一定周期で読み込まれる。次いで、Ｓ２２に移行し、読み込まれた画像データのフーリエ変換処理が行われる。このフーリエ変換処理は、前述した図２のＳ１２のフーリエ変換処理と同様に行われる。
【００６４】
次いで、Ｓ２４に移行し、パワースペクトルの分布の検出が行われる。例えば、パワースペクトルの分布検出は、画像データをフーリエ変換における周波数成分のパワースペクトルの分布を検出することにより行われる。
【００６５】
そして、Ｓ２６に移行し、パワースペクトルの分布に基づいて監視領域の風力の判定が行われる。例えば、図８に示すように、周波数分布の縦方向（ｖ）にパワースペクトルが分布しているときには、画像データにおいて横方向の周期性が強いことを意味するため、雪が横方向に降っており、風力が強いと判断することができる。一方、周波数分布の横方向（ｕ）にパワースペクトルが分布しているときには、画像データにおいて縦方向の周期性が強いことを意味するため、雪が縦方向に降っており、風力が弱いと判断することができる。
【００６６】
なお、画像データのフーリエ変換は、この場合もシャッフリングしたものが用いられる。
【００６７】
そして、Ｓ２８に移行し、判定された風力の情報が外部出力部３４から外部装置４へ出力され、制御処理を終了する。
【００６８】
以上のように、本実施形態に係る気象検出システム、気象検出装置及び気象検出方法によれば、気象状況を検出すべき場所を撮影した画像データを画像処理することにより、監視領域の風力を検出することができる。このため、多数の場所の風力状況を迅速に検出することができる。
【００６９】
また、画像処理としてフーリエ変換を行うことにより、画像データにおいて降雪等による画像の周期性を検出し、その周期性の程度に基づいて監視領域の風力を判別することができる。従って、監視領域の周囲の明るさによる画像の明暗などに影響されることなく、風力の判別が確実に行える。
【００７０】
（第三実施形態）
次に、第三実施形態に係る気象検出システム及び気象検出装置について説明する。本実施形態が第一実施形態と異なるのは、計測処理部３３の内部構成、及び、計測処理部３３においてパワースペクトルの総和に基づいて降雪状態を判断する方法である。
【００７１】
図１７は、本実施形態の気象検出装置３に備えられた計測処理部３３を示す図である。同図に示すように、計測処理部３３には、パワースペクトルの総和ＰＭと降雪状態との関係を示す関係式が記憶された関係式記憶部３３ａが内蔵されている。以下において、この関係式を「降雪強度算出式」と称する。
【００７２】
図１８を参照して、降雪強度算出式について説明する。図１８は、降雪強度算出式を起てるために使用したグラフである。まず、第一実施形態の気象検出システムを用いてパワースペクトルの総和ＰＭを求めてこれを横軸とし、各パワースペクトルの総和を求めたときの実際の降雪強度（ｃｍ／ｈ）を縦軸としてプロットした。尚、このグラフは、昼間の測定によって得られた結果を示している。そして、プロットされた各点の分布を示す最適な曲線を求めたところ、画像出力結果ｘ（パワースペクトルの総和）と降雪強度ｙ（ｃｍ／ｈ）との相関を示す以下のような二元二次の降雪強度算出式が得られた。
ｙ＝ａｘ²＋ｂｘ＋ｃ
（但し、ａ＝３．０×１０^-17、ｂ＝−４．０×１０^-9、ｃ＝０．１１１３）
尚、この降雪強度関係式のピアソンの積率相関係数Ｒは０．８４６２となり、パワースペクトルの総和ｘと降雪強度ｙとは強い相関があることが分かった。
【００７３】
そして、本実施形態の気象検出システム１によって実際に降雪状態を検出するには、まず、第一実施形態と同様に、図２のＳ１０〜Ｓ２０の動作を行う。そして、Ｓ１４においてパワースペクトルの総和ＰＭを求めた後に、計測処理部３３は、関係式記憶部３３ａに記憶された上記の降雪強度算出式にパワースペクトルの総和ＰＭを代入することで、降雪強度（ｃｍ／ｈ）を判定する（Ｓ１６）。このようにして得られた降雪強度は、降雪状態を定量的に数値で示すものであり、第一実施形態で検出された結果よりもさらに信頼度の高いものとなる。
【００７４】
図１９及び図２０は、時刻と降雪強度との関係を示すグラフであり、本実施形態の気象検出システム１で検出された降雪強度を実線で表し、手作業で計測した降雪強度（以下、真値という）を破線で表している。図１９は、９時４０分から１０時２５分までの測定結果を示し、図２０は、１２時１０分から１４時０５分までの測定結果を示している。各グラフに示すように、気象検出システム１で検出された降雪強度の傾向は、真値に近似しており、本実施形態の気相検出装置及び気象検出システムの気象検出精度が極めて優れていることが分かる。このため、多数の場所の気象状況を迅速且つ正確に検出することができる。
【００７５】
図２１は、夜間用の降雪強度算出式を起てるために使用したグラフである。まず、昼間用の降雪強度算出式を求める場合と同様に、第一実施形態の気象検出システムを用いてパワースペクトルの総和ＰＭを求めてこれを横軸とし、各パワースペクトルの総和ＰＭを求めたときの実際の降雪強度（ｃｍ／ｈ）を縦軸としてプロットした。そして、プロットされた各点の分布を示す最適な曲線を求めたところ、パワースペクトルの総和ｘと降雪強度ｙ（ｃｍ／ｈ）との相関を示す以下のような二元二次の降雪強度算出式が得られた。
ｙ＝ａｘ²＋ｂｘ＋ｃ
（但し、ａ＝２．３９×１０^-16、ｂ＝−７．３２×１０^-8、ｃ＝５．６０１）
尚、この降雪強度関係式のピアソンの積率相関係数Ｒは０．８５９０となり、パワースペクトルの総和ｘと降雪強度ｙとは強い相関があることが分かった。
【００７６】
図２２は、時刻と降雪強度との関係を示すグラフであり、本実施形態の気象検出システム１で検出された降雪強度を実線で表し、真値を破線で表している。尚、測定時刻は、夜間の１８時０５分から２０時２０分までである。このグラフに示すように、気象検出システム１で検出された降雪強度の傾向は真値に近似しており、夜間においても、本実施形態に係る気相検出装置及び気象検出システムの気象検出精度は優れていることが分かる。
【００７７】
【発明の効果】
以上説明したように本発明によれば、気象状況を検出すべき場所を撮影した画像データを画像処理することにより、その気象状況を検出することができる。このため、多数の場所の気象状況を迅速に検出することができる。
【００７８】
また、画像データを二次元離散フーリエ変換することにより、画像データにおける周期性を通じて気象状況を検出することができる。また、この検出結果は、撮影時の周囲の明るさなどの影響を受けにくく、正確な気象状況の検出が可能となる。
【図面の簡単な説明】
【図１】第一実施形態に係る気象検出システムの説明図である。
【図２】気象検出装置の動作を示すフローチャートである。
【図３】画像データの説明図である。
【図４】画像データの説明図である。
【図５】画像データの説明図である。
【図６】画像データの説明図である。
【図７】フーリエ変換後の画像データの説明図である。
【図８】フーリエ変換後の画像データの説明図である。
【図９】フーリエ変換後の画像データの説明図である。
【図１０】フーリエ変換後の画像データの説明図である。
【図１１】フーリエ変換後の画像データの説明図である。
【図１２】フーリエ変換後の画像データの説明図である。
【図１３】フーリエ変換後の画像データの説明図である。
【図１４】フーリエ変換後の画像データの説明図である。
【図１５】画像データにおけるパワースペクトル総和の説明図である。
【図１６】第二実施形態に係る気象検出装置等の説明図である。
【図１７】第三実施形態に係る気象検出システムの計測処理部を示す図である。
【図１８】昼間用の降雪強度算出式を求めるためのグラフである。
【図１９】第三実施形態の気象検出システムで得られた結果と真値との関係を示すグラフである。
【図２０】第三実施形態の気象検出システムで得られた結果と真値との関係を示すグラフである。
【図２１】夜間用の降雪強度算出式を求めるためのグラフである。
【図２２】第三実施形態の気象検出システムで得られた結果と真値との関係を示すグラフである。
【符号の説明】
１…気象検出システム、２…監視カメラ（撮影手段）、３…気象検出装置、２１…検出板、３３…計測処理部、３３ａ…関係式記憶部。[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a weather detection device, a weather detection system, and a weather detection method for detecting a weather situation such as snowfall using an image processing technique.
[0002]
[Prior art]
On highways and the like, it is necessary to remove snow on the road in winter due to weather conditions. This snow removal work is performed by driving a snowplow according to weather conditions such as the amount of snowfall per unit time and removing snow on the road with the snowplow. At that time, it is necessary to know the weather condition accurately in order to perform snow removal work quickly and accurately.
[0003]
Conventionally, the weather situation is 1m per hour by a worker using a snow gauge. ² It was measured by directly measuring how many grams of snow fell on the ground.
[0004]
[Problems to be solved by the invention]
However, such a weather detection method cannot quickly grasp the weather situation. In addition, when there are a large number of measurement points for weather conditions, it is necessary to deploy workers accordingly, which requires a lot of personnel. For this reason, there is an urgent need for technological development such as a device that can quickly detect weather conditions in many places.
[0005]
Accordingly, the present invention has been made to solve the above-described problems, and it is an object of the present invention to provide a weather detection device, a weather detection system, and a weather detection method that can quickly detect weather conditions at many locations. And
[0006]
[Means for Solving the Problems]
That is, the meteorological detection apparatus according to the present invention inputs image data obtained by photographing a snowfall monitoring region, and obtains the Fourier transform means for performing two-dimensional discrete Fourier transform on the image data, and two-dimensional discrete Fourier transform. Calculation means for calculating the total sum of the power spectra, and snowfall judgment means for judging the snowfall state of the snowfall monitoring area based on the value of the sum of the power spectra.
[0007]
In addition, the weather detection system according to the present invention inputs an image capturing unit that captures a monitoring area where a snowfall situation should be monitored, and a video signal output from the image capturing unit, and snowfall in the monitoring area is based on image data of the video signal. And the above-described weather detection device for detecting the state.
[0008]
In addition, the weather detection method according to the present invention is obtained by inputting the image data obtained by photographing the snowfall monitoring region and performing a two-dimensional discrete Fourier transform on the image data, and the two-dimensional discrete Fourier transform. A calculation step for calculating the sum of the power spectra, and a snowfall determination step for determining the snowfall state of the snowfall monitoring region based on the value of the sum of the power spectra.
[0009]
According to these inventions, it is possible to detect a snowfall state through periodicity in image data by performing two-dimensional discrete Fourier transform on the image data. In addition, this detection result is not easily affected by the brightness of the surroundings at the time of shooting, and it is possible to accurately detect snowfall.
[0010]
In the meteorological detection apparatus according to the present invention, it is preferable that the snowfall determining means determine the snowfall state using a relational expression indicating the relationship between the sum of the power spectra and the snowfall state. When such a configuration is adopted, it is possible to quantitatively indicate the snowfall state by substituting the sum of the power spectrum into the relational expression.
[0011]
In addition, the weather detection apparatus according to the present invention receives image data obtained by photographing a monitoring region, and obtains Fourier transform means for performing two-dimensional discrete Fourier transform on the image data, and two-dimensional discrete Fourier transform. A wind force judging means for judging the wind force in the monitoring area based on the distribution state of the power spectrum.
[0012]
In addition, the weather detection system according to the present invention includes an imaging unit that captures a monitoring area, and a video signal output from the imaging unit, and detects the wind force in the monitoring area based on image data of the video signal. And a detection device.
[0013]
The meteorological detection method according to the present invention is obtained by inputting the image data obtained by photographing the monitoring area and performing a two-dimensional discrete Fourier transform on the image data, and the two-dimensional discrete Fourier transform. A wind force determination step of determining the wind force in the monitoring area based on the distribution state of the power spectrum.
[0014]
According to these inventions, by performing two-dimensional discrete Fourier transform on image data, it is possible to detect wind force in the monitoring region through periodicity in the image data. Further, this detection result is not easily affected by ambient brightness at the time of shooting, and accurate wind power detection is possible.
[0015]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In addition, in each figure, the same code | symbol is attached | subjected to the same element and description is abbreviate | omitted. Further, the dimensional ratios in the drawings do not necessarily match those described.
[0016]
(First embodiment)
FIG. 1 is a configuration diagram of a weather detection system and a weather detection device according to the present embodiment.
[0017]
As shown in FIG. 1, the weather detection system 1 is a system that detects a snowfall situation on an expressway where a vehicle travels, and includes a monitoring camera 2 and a weather detection device 3.
[0018]
The monitoring camera 2 is a photographing means for photographing a highway or a general national road, and for example, a CCD camera is used. Note that the surveillance camera 2 is not limited to capturing a moving image, and may be a camera that captures a still image. The surveillance camera 2 may be a movable camera whose photographing position can be changed. In this case, you may image | photograph the detection plate 21 mentioned later regularly at a fixed time interval.
[0019]
Further, when a monitoring camera used in a road traffic monitoring system is installed, the traffic monitoring camera may be used as the monitoring camera 2 of the weather detection system 1. In this case, a weather signal can be detected based on the video signal obtained using an existing surveillance camera. Therefore, it is not necessary to separately install the monitoring camera 2 for the weather detection system 1, and the installation cost of the weather detection system 1 can be greatly reduced.
[0020]
A detection plate 21 is provided in front of the monitoring camera 2. The detection plate 21 functions as a background for detecting falling snow, and is formed of a black plate. By making the detection plate 21 black, the presence of white snow becomes clear and it becomes easy to image snow. Further, it is desirable that the detection plate 21 has a waterproof coating on the surface. In this case, it is possible to prevent snow from adhering to the surface of the detection plate 21, and the function as the background when photographing snow is not impaired. Furthermore, it is desirable to install illumination that illuminates the detection plate 21 so that a snowfall state can be detected even at night.
[0021]
Although only one monitoring camera 2 is illustrated in FIG. 1, the weather detection system 1 includes a plurality of monitoring cameras 2, and the monitoring cameras 2 are installed at various locations on the road.
[0022]
Each monitoring camera 2 is electrically connected to the weather detection device 3 and outputs a video signal to the weather detection device 3. The weather detection device 3 receives a video signal from the monitoring camera 2 and performs image processing based on the video signal to detect a weather condition in the monitoring area, and functions as a weather detection means in the weather detection system 1. .
[0023]
The weather detection device 3 includes a video input unit 31, an image processing unit 32, a measurement processing unit 33, an external device output unit 34, and a human interface 35.
[0024]
The video input unit 31 inputs a video signal output from the surveillance camera 2, A / D converts a video signal that is an analog signal into a digital signal, and converts the A / D converted video signal at regular intervals. Input to the image memory as image data. The input of the image data is performed at a constant cycle, for example, 3 sec. Done every time. The input image data is, for example, a set of pixels having gradation values for each coordinate point on the x-axis and y-axis, and the gradation of the pixels is, for example, 8-bit 256 gradations.
[0025]
The image processing unit 32 performs processing such as Fourier transform processing of image data, power spectrum calculation processing, snowfall mode determination, and the like, and includes, for example, an image processing processor. The measurement processing unit 33 performs control processing such as data storage processing and communication processing with an external device, and is configured by, for example, a general-purpose microprocessor.
[0026]
The external output unit 34 outputs a weather condition in the monitoring area as a detection signal, and is connected to the external device 4. The external device 4 displays weather information and the like, and is, for example, a weather database or warning display board installed in the control center, a display panel installed near the expressway, or the like.
[0027]
The human interface 35 is for setting system parameters, outputting processing results, and the like, and is installed between the image processing unit 32, the measurement processing unit 33, the external output unit 34, and the monitor / maintenance console 5. The monitor / maintenance console 5 is installed in a control center (not shown), and performs settings of the weather detection device 3 and the like. In the control center, the video signal transmitted from the monitoring camera 2 is input to the monitor via the weather detection device 3, and the monitoring area can be visually monitored by the monitoring person.
[0028]
Next, a weather detection method in the weather detection system and the weather detection apparatus according to the present embodiment will be described.
[0029]
In FIG. 1, a surveillance area on a highway is photographed by the surveillance camera 2. The video signal of the monitoring camera 2 is input to the video input unit 31 of the weather detection device 3, A / D converted, and stored as image data. Based on this image data, the weather detection device 3 detects the weather condition.
[0030]
FIG. 2 shows a flowchart of the weather detection process in the weather detection device 3.
[0031]
In step S10 of FIG. 2 (hereinafter simply referred to as “S10”, the same applies to other steps), image data is read. The image data is read at a constant cycle of 3 sec, for example. Next, the process proceeds to S12, where a Fourier transform process is performed for two-dimensional discrete Fourier transform on the read image data.
[0032]
This Fourier transform process is performed in a part of the image data, that is, in the measurement region. As the measurement area, an area where the number of pixels is a factorial of 2 is set, for example, an area of 64 × 64 is set.
[0033]
Then, the power spectrum F (u, v) in the measurement region is calculated by the following equation (1).
[Expression 1]

Here, f (x, y) is the luminance value of the pixel in the measurement region, M is the number of pixels in the horizontal direction of the measurement region, and N is the number of pixels in the vertical direction of the measurement region. u and v are set to −M / 2 ≦ u ≦ M / 2 and −N / 2 ≦ v ≦ N / 2, respectively.
[0034]
Next, the process proceeds to S14, and based on the power spectrum F (u, v) of the formula (1), as shown in the following formula (2), the power spectrum strength (amplitude spectrum) P (u, v) Is calculated.
P (u, v) = (R2 + I2) 1/2 (2)
Where R is the real part of F (u, v) and I is the imaginary part of F (u, v).
[0035]
Based on the power spectrum strength P (u, v), as shown in the following equation (3), the sum PM of the power spectrum in the measurement region is calculated.
[Expression 2]

However, PM is a value excluding P (0, 0).
[0036]
Next, the process proceeds to S16, and a snowfall state is determined. The determination of the snowfall state is performed based on the calculated sum PM of power spectra.
[0037]
For example, the snowfall mode is set as “strong”, “medium”, “weak”, “none” in the order of snowfall intensity, the threshold values of “strong” and “medium” in the snowfall mode are A1, and “medium” in the snowfall mode. ”And“ weak ”threshold values are set to A2, and the“ weak ”and“ none ”threshold values in the snowfall mode are set to A3. In this case, the threshold values A1, A2, and A3 are set as A1>A2> A3.
[0038]
If the sum PM of the calculated power spectrum is equal to or greater than the threshold value A1, the snowfall mode is determined to be “strong”, and the calculated sum PM of the power spectrum obtained is smaller than the threshold value A1. If it is equal to or greater than the threshold A2, the snowfall mode is determined to be “medium”. If the sum PM of the power spectra obtained by calculation is smaller than the threshold A2 and greater than or equal to the threshold A3, the snowfall mode is “weak”. When the total PM of the power spectrum obtained by calculation is smaller than the threshold value A3, the snowfall mode is determined as “none”.
[0039]
And it transfers to S18, the information of the determined snowfall mode is output to the external device 4 from the external output part 34, and a control process is complete | finished.
[0040]
Next, actual weather detection results performed using the weather detection device, the weather detection system, and the weather detection method according to the present embodiment will be described.
[0041]
3 to 6 show image data obtained based on the captured video of the monitoring camera 2. FIG. 3 shows image data when the snowfall is strong. FIG. 4 shows image data when the snowfall state is medium. FIG. 5 shows image data when the snowfall state is weak. FIG. 6 shows image data when there is no snow. Note that the image data in FIGS. 3 to 6 are bmp images having a horizontal m × vertical n of 320 × 240 (76800 points). 3 to 6, the black portion is the black detection plate 21.
[0042]
The above-described image processing from S12 to S14 in FIG. 2, that is, Fourier transform processing was performed on the pixels on the image data detection plate. As a result, Fourier transform data shown in FIGS.
[0043]
FIG. 7 is obtained by Fourier transforming image data when the snowfall is strong, and shows it three-dimensionally. In FIG. 7, the horizontal directions u and v indicate frequency components, and the vertical direction of PM indicates the intensity at each frequency component. FIG. 8 is a plan view of FIG. 7 and shows the frequency distribution in the Fourier transform when the snowfall state is strong.
[0044]
FIG. 9 is obtained by performing Fourier transform on the image data when the snowfall state is medium, and shows it three-dimensionally. FIG. 10 is a plan view of FIG. 9 and shows the frequency distribution in the Fourier transform when the snowfall state is medium.
[0045]
FIG. 11 is obtained by Fourier transforming image data when the snowfall state is weak, and shows it three-dimensionally. FIG. 12 is a plan view of FIG. 11 and shows the frequency distribution in the Fourier transform when the snowfall state is weak.
[0046]
FIG. 13 is obtained by Fourier transforming image data when there is no snowfall, and shows it three-dimensionally. FIG. 14 is a plan view of FIG. 13 and shows the frequency distribution in the Fourier transform when there is no snowfall.
[0047]
7 to 14 are shuffled by distributing the low frequency component in the central portion and the high frequency component in the four corner portions in each frequency component of u and v.
[0048]
7 and 8, when the Fourier transform is strong when the snowfall state is strong, the power spectrum (PM) is large around the low frequency component, and the periodicity in the vertical direction (v) is strongly shown.
[0049]
9 and FIG. 10, the Fourier transform in the case where the snowfall state is moderate shows that the power spectrum (PM) is large centering on the low frequency component, but the periodicity in the vertical direction (v) is slightly stronger. Has been.
[0050]
Looking at the Fourier transform when the snowfall state is weak in FIGS. 11 and 12, the power spectrum (PM) is small as a whole.
[0051]
Looking at the Fourier transform in the case of no snowfall in FIGS. 13 and 14, the power spectrum (PM) is very small as a whole.
[0052]
Then, when the sum of the power spectrum in each snowfall state is calculated, as shown in FIG. 15, when the snowfall state is strong, it is 1181291.1, and when the snowfall state is medium, it is 1015720.0, and the snowfall state is It was 885125.7 when it was weak and 39316.7 when there was no snowfall.
[0053]
In this way, it is possible to determine each snowfall state based on the sum of the power spectra. In this case, as threshold values for determining the snowfall state, for example, a threshold value A1 of “strong” and “medium” in the snowfall state is 1050000, and a threshold value A2 of “medium” and “weak” in the snowfall state is set. If the threshold value A3 of “weak” and “none” in the snowfall state is 450,000, the snowfall state can be determined based on the sum of the power spectra.
[0054]
As described above, according to the weather detection system, the weather detection device, and the weather detection method according to the present embodiment, it is possible to detect a snowfall situation by performing image processing on image data of a place where a weather situation should be detected. Can do. For this reason, the snowfall situation of many places can be detected rapidly.
[0055]
Further, by performing Fourier transform as image processing, it is possible to detect the periodicity of the image due to snowfall in the image data and determine the snowfall state based on the degree of the periodicity. Therefore, it is possible to reliably determine the snowfall state without being affected by the brightness of the image due to the brightness around the monitoring area.
[0056]
In addition, by determining the snowfall state based on the sum of the power spectrum, not the maximum power spectrum obtained by Fourier transform, it is possible to reliably determine whether the snowfall state is “strong” or “medium”. it can.
[0057]
Furthermore, according to the weather detection system, the weather detection device, and the weather detection method according to the present embodiment, if the video of the existing monitoring camera is used, there is no need to newly install the monitoring camera and the introduction of the weather detection system. Is easy.
[0058]
In the present embodiment, the weather detection system, the weather detection device, and the weather detection method for detecting the snowfall situation on the expressway have been described. However, the weather detection system, the weather detection device, and the weather detection method according to the present invention are as described above. However, the present invention is not limited to this, and it may be one that detects snowfall conditions in other places.
[0059]
(Second embodiment)
Next, a weather detection device, a weather detection system, and a weather detection method according to the second embodiment will be described.
[0060]
The meteorological detection device, the meteorological detection system, and the meteorological detection method according to the first embodiment detect a snowfall state in the monitoring area, but the meteorological detection device, the meteorological detection system, and the meteorological detection method according to the present embodiment monitor. This is to detect wind force during snowfall in the area.
[0061]
The weather detection device and the weather detection system according to the present embodiment have the same hardware configuration as the weather detection device and the weather detection system of the first embodiment shown in FIG.
[0062]
FIG. 16 shows a flowchart of a weather detection process according to the weather detection device and the like.
[0063]
In step S20 in FIG. 16, image data is read. The image data is read at a constant cycle of 3 sec, for example. Next, the process proceeds to S22, and Fourier transform processing of the read image data is performed. This Fourier transform process is performed in the same manner as the Fourier transform process in S12 of FIG.
[0064]
Next, the process proceeds to S24, and the distribution of the power spectrum is detected. For example, the power spectrum distribution detection is performed by detecting the power spectrum distribution of the frequency component in the Fourier transform of the image data.
[0065]
And it transfers to S26 and the determination of the wind force of a monitoring area | region is performed based on distribution of a power spectrum. For example, as shown in FIG. 8, when the power spectrum is distributed in the vertical direction (v) of the frequency distribution, it means that the horizontal periodicity is strong in the image data, so that snow is falling in the horizontal direction. It can be judged that wind power is strong. On the other hand, when the power spectrum is distributed in the horizontal direction (u) of the frequency distribution, it means that the periodicity in the vertical direction is strong in the image data. Therefore, it is determined that the snow is falling in the vertical direction and the wind force is weak. be able to.
[0066]
The Fourier transform of the image data is also shuffled in this case.
[0067]
And it transfers to S28, the information of the determined wind force is output to the external device 4 from the external output part 34, and a control process is complete | finished.
[0068]
As described above, according to the weather detection system, the weather detection device, and the weather detection method according to the present embodiment, the wind power in the monitoring region is detected by performing image processing on the image data obtained by photographing the place where the weather situation should be detected. can do. For this reason, it is possible to quickly detect wind power conditions in a large number of places.
[0069]
Further, by performing Fourier transform as image processing, it is possible to detect the periodicity of the image due to snowfall or the like in the image data and determine the wind force in the monitoring region based on the degree of the periodicity. Therefore, it is possible to reliably determine the wind force without being affected by the brightness of the image due to the brightness around the monitoring area.
[0070]
(Third embodiment)
Next, a weather detection system and a weather detection apparatus according to the third embodiment will be described. This embodiment is different from the first embodiment in the internal configuration of the measurement processing unit 33 and the method for determining the snowfall state based on the sum of the power spectrum in the measurement processing unit 33.
[0071]
FIG. 17 is a diagram illustrating the measurement processing unit 33 provided in the weather detection device 3 of the present embodiment. As shown in the figure, the measurement processing unit 33 incorporates a relational expression storage unit 33a in which a relational expression indicating the relation between the sum PM of power spectra and the snowfall state is stored. Hereinafter, this relational expression is referred to as “snowfall intensity calculation expression”.
[0072]
The snowfall intensity calculation formula will be described with reference to FIG. FIG. 18 is a graph used to generate the snowfall intensity calculation formula. First, the sum PM of power spectra is obtained using the weather detection system of the first embodiment, and this is taken as the horizontal axis, and the actual snowfall intensity (cm / h) when the sum of each power spectrum is found is taken as the vertical axis. Plotted. This graph shows the results obtained by daytime measurements. Then, when an optimum curve indicating the distribution of each plotted point is obtained, the following two-way binary indicating the correlation between the image output result x (total power spectrum) and the snowfall intensity y (cm / h) is shown. The following snowfall strength calculation formula was obtained.
y = ax ² + Bx + c
(However, a = 3.0 × 10 ^-17 B = −4.0 × 10 ^-9 C = 0.1113)
Note that the Pearson product moment correlation coefficient R in this snowfall intensity relational expression is 0.8462, which indicates that the sum x of the power spectrum and the snowfall intensity y have a strong correlation.
[0073]
And in order to actually detect a snowfall state by the weather detection system 1 of this embodiment, operation | movement of S10-S20 of FIG. 2 is first performed similarly to 1st embodiment. And after calculating | requiring total PM of a power spectrum in S14, the measurement process part 33 substitutes the total PM of a power spectrum for said snowfall intensity | strength calculation formula memorize | stored in the relational expression memory | storage part 33a, and snowfall intensity | strength ( cm / h) is determined (S16). The snowfall intensity obtained in this manner quantitatively indicates the snowfall state as a numerical value, and has a higher reliability than the result detected in the first embodiment.
[0074]
19 and 20 are graphs showing the relationship between time and snowfall intensity. The snowfall intensity detected by the weather detection system 1 of the present embodiment is indicated by a solid line, and the snowfall intensity measured manually (hereinafter, true) is shown. Value)) is represented by a broken line. FIG. 19 shows the measurement results from 9:40 to 10:25, and FIG. 20 shows the measurement results from 12:10 to 14:05. As shown in each graph, the tendency of the snowfall intensity detected by the weather detection system 1 approximates the true value, and the weather detection accuracy of the gas phase detection device and the weather detection system of the present embodiment is extremely excellent. I understand that. For this reason, the weather conditions of many places can be detected quickly and accurately.
[0075]
FIG. 21 is a graph used to generate the snowfall intensity calculation formula for nighttime. First, as in the case of obtaining the snowfall intensity calculation formula for daytime, the sum PM of the power spectra was obtained using the weather detection system of the first embodiment, and this was taken as the horizontal axis, and the sum PM of each power spectrum was obtained. The actual snowfall intensity (cm / h) was plotted as the vertical axis. Then, when the optimum curve showing the distribution of each plotted point was obtained, the following binary secondary snowfall intensity calculation showing the correlation between the sum x of the power spectrum and the snowfall intensity y (cm / h) The formula was obtained.
y = ax ² + Bx + c
(However, a = 2.39 × 10 ^-16 , B = −7.32 × 10 ^-8 C = 5.601)
Note that the Pearson product moment correlation coefficient R in this snowfall intensity relational expression is 0.8590, which indicates that the sum x of the power spectrum and the snowfall intensity y have a strong correlation.
[0076]
FIG. 22 is a graph showing the relationship between time and snowfall intensity. The snowfall intensity detected by the weather detection system 1 of the present embodiment is indicated by a solid line, and the true value is indicated by a broken line. The measurement time is from 18:05 to 20:20 at night. As shown in this graph, the tendency of the snowfall intensity detected by the weather detection system 1 is close to a true value, and the weather detection accuracy of the gas phase detection device and the weather detection system according to the present embodiment is even at night. It turns out that it is excellent.
[0077]
【The invention's effect】
As described above, according to the present invention, the weather condition can be detected by performing image processing on the image data obtained by photographing the place where the weather condition should be detected. For this reason, the weather conditions of many places can be detected quickly.
[0078]
In addition, weather conditions can be detected through periodicity in image data by performing two-dimensional discrete Fourier transform on the image data. Further, this detection result is not easily affected by ambient brightness at the time of shooting, and an accurate weather condition can be detected.
[Brief description of the drawings]
FIG. 1 is an explanatory diagram of a weather detection system according to a first embodiment.
FIG. 2 is a flowchart showing the operation of the weather detection apparatus.
FIG. 3 is an explanatory diagram of image data.
FIG. 4 is an explanatory diagram of image data.
FIG. 5 is an explanatory diagram of image data.
FIG. 6 is an explanatory diagram of image data.
FIG. 7 is an explanatory diagram of image data after Fourier transform.
FIG. 8 is an explanatory diagram of image data after Fourier transform.
FIG. 9 is an explanatory diagram of image data after Fourier transform.
FIG. 10 is an explanatory diagram of image data after Fourier transform.
FIG. 11 is an explanatory diagram of image data after Fourier transform.
FIG. 12 is an explanatory diagram of image data after Fourier transform.
FIG. 13 is an explanatory diagram of image data after Fourier transform.
FIG. 14 is an explanatory diagram of image data after Fourier transform.
FIG. 15 is an explanatory diagram of power spectrum summation in image data.
FIG. 16 is an explanatory diagram of a weather detection device and the like according to a second embodiment.
FIG. 17 is a diagram illustrating a measurement processing unit of a weather detection system according to a third embodiment.
FIG. 18 is a graph for obtaining a snowfall intensity calculation formula for daytime.
FIG. 19 is a graph showing a relationship between a result obtained by the weather detection system of the third embodiment and a true value.
FIG. 20 is a graph showing a relationship between a result obtained by the weather detection system of the third embodiment and a true value.
FIG. 21 is a graph for obtaining a snowfall intensity calculation formula for night use.
FIG. 22 is a graph showing a relationship between a result obtained by the weather detection system of the third embodiment and a true value.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Weather detection system, 2 ... Monitoring camera (imaging means), 3 ... Weather detection apparatus, 21 ... Detection board, 33 ... Measurement processing part, 33a ... Relational expression memory | storage part.

Claims (7)

Fourier transform means for inputting image data obtained by photographing a snowfall monitoring area and performing two-dimensional discrete Fourier transform on the image data;
Calculating means for calculating the sum of the power spectra obtained by the two-dimensional discrete Fourier transform;
Snowfall judging means for judging a snowfall state of the snowfall monitoring area based on a value of the sum of the power spectra;
A weather detection device. The weather detection apparatus according to claim 1, wherein the snowfall determination unit determines the snowfall state using a relational expression indicating a relationship between the sum of the power spectra and the snowfall state. Photographing means for photographing the monitoring area where the snowfall situation should be monitored;
The weather detection device according to claim 1 or 2, wherein a video signal output from the photographing unit is input, and a snowfall state of the monitoring area is detected based on image data of the video signal;
Weather detection system with. A Fourier transform step of inputting image data obtained by photographing the snowfall monitoring area and performing two-dimensional discrete Fourier transform on the image data;
A calculation step of calculating a sum of power spectra obtained by the two-dimensional discrete Fourier transform;
A snowfall determining step of determining a snowfall state of the snowfall monitoring region based on a value of the sum of the power spectra;
A weather detection method. Fourier transform means for inputting image data obtained by photographing the monitoring area, and performing two-dimensional discrete Fourier transform on the image data;
Wind power judgment means for judging the wind power of the monitoring region based on the distribution state of the power spectrum obtained by the two-dimensional discrete Fourier transform,
A weather detection device. Photographing means for photographing the surveillance area;
The weather detection device according to claim 5, wherein a video signal output from the imaging unit is input, and wind force in the monitoring area is detected based on image data of the video signal;
Weather detection system with. A Fourier transform step of inputting image data obtained by photographing the monitoring region and performing two-dimensional discrete Fourier transform on the image data;
A wind force determination step of determining the wind force of the monitoring region based on the distribution state of the power spectrum obtained by the two-dimensional discrete Fourier transform;
A weather detection device.

Images