JP4580117B2 - Parking vehicle detection device - Google Patents

Description

【０００５】
この誤判定の原因は次のように考えられる。
得られた輝度偏差とエッジ量のパラメータを駐車と空車に分類し、そのヒストグラムを算出した分布図を図６に示す。図５（ｂ）のように黒色車両は輝度の変化が乏しいため得られるエッジ量は小さい。これに対し、重なり車両は部分的にしか駐車マス領域を覆い隠していないが、その車両が白色車両の場合はエッジ量が黒色車両を上回る場合が発生する。
【０００６】
これが、図６における駐車と空車のオーバーラップ部分となる。このため、黒色車両を検出するためにしきい値を低く設定すると、重なり車両を誤検出することになる。逆に、重なり車両による誤判定を回避するためしきい値を高く設定すると、黒色車両を未検出することになる。したがって、しきい値処理により駐車と空車を判定する場合、このような駐車と空車のオーバーラップ部分で誤判定が発生することになる。以上のことから、エッジで駐車と空車を正しく判定できない原因は、黒色車両のエッジを明確に抽出できないという点にある。
【０００７】
本発明は前記した問題点を解決せんとするもので、その目的とするところは、黒色車両を白色車両と同等のレベルで扱うことができ、かつ、駐車と空車の分離度が非常に高いパラメータを採用することにより、駐車車両の検出を高速、かつ、容易に行うことができる駐車車両検出装置を提供せんとするにある。
【０００８】
【課題を解決するための手段】
本発明の駐車車両検出装置は前記した目的を達成せんとするもので、その手段は、車両が駐車する駐車場などの監視領域を俯瞰撮影する撮影手段と、前記撮影手段から送られてくる輝度信号をデジタル信号に変換するＡＤ変換部と、ＡＤ変換された監視領域内のデジタル輝度信号から車両のエッジを抽出するエッジ抽出部と、抽出されたエッジを元の原画像輝度で割る演算を行うエッジ強調処理部と、該エッジ強調処理部において得られたエッジ画像から駐車車両の有無を検出する駐車車両検出部とを具備したことを特徴とする。
【００１０】
【発明の実施の形態】
以下、本発明に係る駐車車両検出装置の実施の形態を図１と共に説明する。
１は計測対象とする駐車場などを俯瞰撮影できるように駐車場の高所に設置されたテレビカメラ等の撮影手段、２は該撮影手段１によって得られた画像のサンプリング時の折り返し歪みや高域ノイズを除去するためのローパスフィルタ、３はＡＤ変換器、４は車両、重なり車両、影、駐車マスなどを含む原画像に輪郭抽出フィルタの一種であるソーベルフィルタをかけて求めたエッジ抽出部、５は該エッジ抽出部４で得られたエッジを元の原画像輝度で割る演算を行うエッジ強調処理部、６は該エッジ強調処理部５において得られたエッジ画像から駐車車両の有無を検出する駐車車両検出部である。
【００１１】
次に、前記した構成に基づいて動作を説明するに、駐車場等における所定の監視領域を撮影手段１によって撮影すると、該撮影手段１よりの撮影画像はローパスフィルタ２を通じてＡＤ変換器３に送られる。該ＡＤ変換器３はテレビカメラから送られてくるフレーム単位の撮影画像信号（輝度信号）を所定の周期でサンプリングしてデジタル信号に変換しエッジ抽出部４に送る。
【００１２】
該エッジ抽出部４はフレーム中の画像のエッジ画像を抽出するが、黒色車両を白色車両と同等のエッジ量で検出しないと従来と同様に誤検出となる。そこで、本発明にあっては、黒色車両のエッジ量を強調するために、エッジ強調処理部５において前記エッジ抽出部４で得られたエッジを元の原画像輝度で割算を行う。
このエッジ強調処理部５により得られたエッジ強調画像を図２に示すが、図２からも分かるように、黒色車両は白色車両と同等なエッジ量となる。
【００１３】
このエッジ強調画像を用いて駐車車両の検出を駐車車両検出部６によって行うことにより、白色車両は勿論のこと、黒色車両の計数が行える。その結果を表２に示し、また、この時の駐車と空車の分布図を図３に示す。この結果から明らかなように、駐車と空車の分布が完全に分離できており、判定の正解率を１００％とすることができる。
【００１４】
【表２】

【００１５】
さらに，従来と本発明の違いを明確にするために、駐車と空車の分離度をそれぞれ算出した結果を図４に示す。このように、従来に比べ本発明では３倍以上の分離度が得られていることから、エッジ強調画像による効果が十分確認できる。
【００１６】
なお、前記したエッジ抽出部４のエッジ画像として、ソーベル画像を代表として用いたが、ラプラシアンやグラーディエント等の他のエッジ画像を用いてもよい。
【００１７】
【発明の効果】
本発明は前記したように、駐車場等を俯瞰撮影して得られた画像から車両等のエッジ画像を得、該エッジ画像を元の原画像輝度で割算して黒色車両のエッジ量を白色車両のエッジ量と同等にした状態で駐車車両を検出するようにしたので、黒色車両を白色車両と同等のレベルで扱うことができ、従って、駐車車両の検出を高速かつ容易に行え、その結果、安定した台数計測が可能である等の効果を有するものである。
【図面の簡単な説明】
【図１】本発明の車両検出装置のブロック図である。
【図２】エッジ強調前の画像と強調後の画像を示す図である。
【図３】駐車と空車の分布図である。
【図４】駐車と空車の分離度を示すグラフである。
【図５】原画像とエッジ画像を示す図である。
【図６】同上の原画像とエッジ画像の分布図である。
【符号の説明】
１ 撮影手段
２ ローパスフィルタ
３ ＡＤ変換器
４ エッジ抽出部
５ エッジ強調処理部
６ 駐車車両検出部[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a parked vehicle detection method and apparatus used when, for example, a parking lot is photographed by a photographing means such as a television camera to detect the parking situation of the vehicle, and in particular, detection of a black vehicle is ensured. about the parked vehicle detection DeSo location it eliminates the determination performed by erroneous.
[0002]
[Prior art]
Conventionally, when detecting whether or not a vehicle is parked in a parking area such as a parking lot, first, as shown in FIG. 5, a determination area is provided for each parking square (white square frame). The determination is made using the luminance deviation of the original image and the edge amount of the edge image in the region.
[0003]
[Problems to be solved by the invention]
By the way, when the luminance deviation in the above-described conventional example is used, an error occurs in all the factors of the black vehicle, the overlapping vehicle, the shadow, and the road surface state, so the accuracy rate is 91.9% (see Table 1). It ’s bad. In addition, in the edge amount, the erroneous determination is reduced and the correct answer rate is improved to 96.3% (see Table 1) compared with the original image, but the black vehicle undetected and the overlapping vehicle erroneous detection cannot be solved. Therefore, the determination based on the luminance deviation and the edge amount cannot perform stable and accurate detection.
[0004]
[Table 1]

[0005]
The cause of this erroneous determination is considered as follows.
FIG. 6 shows a distribution chart in which the obtained luminance deviation and edge amount parameters are classified into parking and empty cars, and the histograms are calculated. As shown in FIG. 5B, the black vehicle has a small change in luminance, so that the obtained edge amount is small. On the other hand, although the overlapping vehicle partially covers and hides the parking mass area, when the vehicle is a white vehicle, the edge amount may exceed the black vehicle.
[0006]
This is an overlap portion between the parking and the empty vehicle in FIG. For this reason, if the threshold value is set low in order to detect a black vehicle, an overlapping vehicle will be erroneously detected. On the other hand, if the threshold value is set high to avoid erroneous determination due to overlapping vehicles, a black vehicle will not be detected. Therefore, when parking and an empty vehicle are determined by threshold processing, an erroneous determination occurs at an overlap portion between the parking and the empty vehicle. From the above, the reason why parking and empty vehicles cannot be correctly determined by the edge is that the edge of the black vehicle cannot be clearly extracted.
[0007]
The present invention is intended to solve the above-mentioned problems, and the object of the present invention is a parameter that can handle a black vehicle at a level equivalent to that of a white vehicle and has a very high degree of separation between parking and empty vehicles. by adopting, certain detection of the parked vehicle fast, and, to St. provide a parked vehicle detection DeSo location that can be easily performed.
[0008]
[Means for Solving the Problems]
The parked vehicle detection device according to the present invention is intended to achieve the above-described object, and includes means for photographing a monitoring area such as a parking lot where the vehicle is parked and a luminance sent from the photographing means. An AD conversion unit that converts the signal into a digital signal, an edge extraction unit that extracts the edge of the vehicle from the digital luminance signal in the monitoring area subjected to the AD conversion, and an operation that divides the extracted edge by the original original image luminance An edge enhancement processing unit and a parked vehicle detection unit that detects the presence / absence of a parked vehicle from an edge image obtained by the edge enhancement processing unit are provided .
[0010]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, an embodiment of a parked vehicle detection device according to the present invention will be described with reference to FIG.
Reference numeral 1 denotes a photographing means such as a TV camera installed at a high place in a parking lot so that a parking lot or the like to be measured can be seen from above, and reference numeral 2 denotes aliasing distortion or high during sampling of an image obtained by the photographing means 1. Low-pass filter for removing area noise, 3 for AD converter, 4 for edge extraction obtained by applying Sobel filter which is a kind of contour extraction filter to original image including vehicles, overlapping vehicles, shadows, parking mass, etc. , 5 is an edge enhancement processing unit that performs an operation of dividing the edge obtained by the edge extraction unit 4 by the original original image luminance, and 6 is an information indicating whether or not there is a parked vehicle from the edge image obtained by the edge enhancement processing unit 5. It is a parked vehicle detection part to detect.
[0011]
Next, to explain the operation based on the above-described configuration, when a predetermined monitoring area in a parking lot or the like is photographed by the photographing means 1, the photographed image from the photographing means 1 is sent to the AD converter 3 through the low-pass filter 2. It is done. The AD converter 3 samples a frame-by-frame captured image signal (luminance signal) sent from the television camera, converts it into a digital signal, and sends it to the edge extraction unit 4.
[0012]
The edge extraction unit 4 extracts an edge image of the image in the frame, but if a black vehicle is not detected with an edge amount equivalent to that of a white vehicle, erroneous detection will occur as in the conventional case. Therefore, in the present invention, in order to emphasize the edge amount of the black vehicle, the edge obtained by the edge extraction unit 4 is divided by the original original image luminance in the edge enhancement processing unit 5.
FIG. 2 shows an edge-enhanced image obtained by the edge enhancement processing unit 5. As can be seen from FIG. 2, the black vehicle has the same edge amount as the white vehicle.
[0013]
By detecting the parked vehicle using the edge-enhanced image, the parked vehicle detection unit 6 can count not only white vehicles but also black vehicles. The results are shown in Table 2, and the distribution of parking and empty cars at this time is shown in FIG. As is clear from this result, the distribution of parking and empty cars can be completely separated, and the accuracy rate of determination can be made 100%.
[0014]
[Table 2]

[0015]
Furthermore, in order to clarify the difference between the prior art and the present invention, FIG. 4 shows the results of calculating the degree of separation between parking and empty vehicles. As described above, since the present invention has a degree of separation of 3 times or more as compared with the prior art, the effect of the edge-enhanced image can be sufficiently confirmed.
[0016]
Although the Sobel image is used as a representative edge image of the edge extraction unit 4, other edge images such as Laplacian and gradient may be used.
[0017]
【The invention's effect】
As described above, the present invention obtains an edge image of a vehicle or the like from an image obtained by taking a bird's-eye view of a parking lot or the like, and divides the edge image by the original image luminance to obtain the edge amount of the black vehicle as white. Since the parked vehicle is detected in a state equivalent to the edge amount of the vehicle, the black vehicle can be handled at the same level as the white vehicle, and therefore, the parked vehicle can be detected quickly and easily. It has the effect that stable number measurement is possible.
[Brief description of the drawings]
FIG. 1 is a block diagram of a vehicle detection device of the present invention.
FIG. 2 is a diagram illustrating an image before edge enhancement and an image after enhancement.
FIG. 3 is a distribution map of parking and empty vehicles.
FIG. 4 is a graph showing the degree of separation between parking and empty cars.
FIG. 5 is a diagram illustrating an original image and an edge image.
FIG. 6 is a distribution diagram of the original image and the edge image.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 Image | photographing means 2 Low pass filter 3 AD converter 4 Edge extraction part 5 Edge emphasis processing part 6 Parked vehicle detection part

Claims (1)

An imaging means for taking a bird's-eye view of a monitoring area such as a parking lot where the vehicle is parked;
An AD converter that converts the luminance signal sent from the photographing means into a digital signal;
An edge extraction unit for extracting the edge of the vehicle from the digital luminance signal in the monitoring area that has been AD converted;
An edge enhancement processing unit that performs an operation of dividing the extracted edge by the original original image luminance;
A parked vehicle detection unit that detects the presence or absence of a parked vehicle from the edge image obtained in the edge enhancement processing unit;
The parking vehicle detection apparatus characterized by comprising.

Images