CN103617410A

Movatterモバイル変換

Info

Publication number: CN103617410A
Application number: CN201310388244.7A
Authority: CN
Inventors: 赵敏; 孙棣华; 刘卫宁; 郑林江; 廖孝勇; 唐毅; 姚明明
Original assignee: Chongqing University
Current assignee: Chongqing University
Priority date: 2013-08-30
Filing date: 2013-08-30
Publication date: 2014-03-05

Abstract

本发明涉及交通检测技术领域，尤其涉及一种基于视频检测技术的高速公路隧道停车检测方法，包括如下步骤：1）从高速公路隧道摄像头获取的视频中按帧抽取图片；2）利用步骤1）中抽取的图片，提取前景目标；3）获取前景目标轮廓面积、质心位置和外接矩形宽高比；4）采用多特征匹配方法对车辆进行跟踪；5）得出车辆目标的跟踪结果J，只有轮廓面积、外接矩形宽高比和质心位置都满足要求时，J为1，否则为0；6）重复步骤1-5），并根据车辆在场景中被跟踪的视频帧数，并结合跟踪结果的统计情况，来判断是否出现停车事件。

The present invention relates to the technical field of traffic detection, in particular to a parking detection method for expressway tunnels based on video detection technology, comprising the following steps: 1) extracting pictures frame by frame from the video obtained by the expressway tunnel camera; 2) using step 1) 3) Obtain the contour area, centroid position, and aspect ratio of the circumscribed rectangle of the foreground object; 4) Use the multi-feature matching method to track the vehicle; 5) Obtain the tracking result J of the vehicle object, only When the outline area, the aspect ratio of the circumscribed rectangle and the position of the centroid all meet the requirements, J is 1, otherwise it is 0; 6) Repeat steps 1-5), and according to the number of video frames of the vehicle being tracked in the scene, combined with the tracking results statistics to determine whether there is a parking incident.

Description

Translated fromChinese

基于视频检测技术的高速公路隧道停车检测方法Parking Detection Method in Expressway Tunnel Based on Video Detection Technology

技术领域technical field

本发明涉及交通检测技术领域，尤其涉及一种基于视频检测技术的高速公路隧道停车检测方法。 The invention relates to the technical field of traffic detection, in particular to a method for detecting parking in expressway tunnels based on video detection technology. the

背景技术Background technique

随着我国高速公路建设的快速发展，高速公路隧道大量投入运营。高速公路隧道使交通高效联通的同时，由于其特殊的行车环境，又成为高速公路的交通瓶颈。由于高速公路隧道内车速较快且环境封闭，发生停车事件后，如不及时处理，极易导致二次交通事故的发生，引起车辆碰撞或引发交通拥堵，严重影响高速公路的正常运营。 With the rapid development of expressway construction in our country, a large number of expressway tunnels have been put into operation. Expressway tunnels not only make the traffic more efficient, but also become the traffic bottleneck of the expressway due to its special driving environment. Due to the fast speed and closed environment in the expressway tunnel, if the parking incident occurs, if it is not dealt with in time, it will easily lead to secondary traffic accidents, cause vehicle collisions or cause traffic congestion, and seriously affect the normal operation of the expressway. the

通过查找专利和论文，发现基于视频的停车检测方法主要有两类，即基于网格模型的方法和基于目标跟踪的方法。基于网格模型进行停车检测的方法主要感知图像区域的变化，从而判断是否出现停车。长安大学申请的“一种基于块累积的高速公路车辆停车检测方法”（CN:102110366）,通过将视频图像分割成多个块区域，并采用二值化方法将目标与背景图像进行分割，然后通过统计相邻块区域的个数确定是不是停车事件。这种方法在视频条件较好的场景下能够快速准确地判断出停车事件，但在高速公路隧道中，由于车辆灯光和系统照明的干扰，有时会在图像中形成局部光斑，导致二值化分割的效果不理想，容易引起误检。基于目标跟踪进行停车检测的方法，即通过车辆跟踪获取车辆运动信息，依据车辆停止时位置固定这一特点来判断停车事件。该方法目前大部分学者都选在高速公路露天路段且天气晴朗、光照基本保持不变或者缓慢变化的情况下进行的研究，这种外部环境稳定干扰因素很少的情况确实能够获得较好的检测准确率，但是在高速公路隧道中，会出现影响车辆跟踪的灯光干扰，并且车辆停止时会经常出现车灯闪烁的现象，导致检测出的车辆位置发生跳变，所以直接采用这种方法无法正确地检测出停车事件。 By searching patents and papers, it is found that there are two main types of video-based parking detection methods, namely grid model-based methods and target tracking-based methods. The method of parking detection based on the grid model mainly perceives the changes of the image area, so as to judge whether there is parking. "A block accumulation-based highway vehicle parking detection method" applied by Chang'an University (CN:102110366), by dividing the video image into multiple block areas, and using the binarization method to segment the target and background images, and then Determine whether it is a parking event by counting the number of adjacent block areas. This method can quickly and accurately judge parking events in scenes with good video conditions, but in highway tunnels, due to the interference of vehicle lights and system lighting, sometimes local light spots will be formed in the image, resulting in binary segmentation The effect is not ideal, and it is easy to cause false detection. The method of parking detection based on target tracking is to obtain vehicle motion information through vehicle tracking, and judge the parking event according to the feature that the vehicle stops at a fixed position. At present, most scholars choose to conduct research on the open-air section of the expressway and the weather is clear, and the light basically remains unchanged or changes slowly. This kind of external environment with few interference factors can indeed obtain better detection. Accuracy, but in expressway tunnels, there will be light interference that affects vehicle tracking, and the lights will often flicker when the vehicle stops, causing the detected vehicle position to jump, so this method cannot be used directly. detect parking events. the

正因如此，需要一种能够准确、实时、有效的停车检测方法，为高速公路隧道的交通安全提供有用的数据信息。 Because of this, an accurate, real-time and effective parking detection method is needed to provide useful data information for traffic safety in expressway tunnels. the

发明内容Contents of the invention

有鉴于此，本发明提供一种基于视频检测技术的高速公路隧道停车检测方法，运算开销小、实时性强。 In view of this, the present invention provides a highway tunnel parking detection method based on video detection technology, which has low computing overhead and strong real-time performance. the

本发明通过以下技术手段解决上述技术问题： The present invention solves the above technical problems by the following technical means:

基于视频检测技术的高速公路隧道停车检测方法，包括如下步骤： A parking detection method for expressway tunnels based on video detection technology, including the following steps:

1）从高速公路隧道摄像头获取的视频中按帧抽取图片； 1) Extract pictures frame by frame from the video captured by the highway tunnel camera;

2）利用步骤1）中抽取的图片，提取前景目标； 2) Use the picture extracted in step 1) to extract the foreground target;

3）获取前景目标轮廓面积、质心位置和外接矩形宽高比； 3) Obtain the area of the foreground target outline, the position of the centroid and the aspect ratio of the circumscribed rectangle;

4）采用多特征匹配方法对车辆进行跟踪； 4) Use multi-feature matching method to track the vehicle;

5）得出车辆目标的跟踪结果J，只有轮廓面积、外接矩形宽高比和质心位置都满足要求时，J为1，否则为0； 5) Get the tracking result J of the vehicle target. Only when the outline area, the aspect ratio of the circumscribed rectangle and the position of the centroid meet the requirements, J is 1, otherwise it is 0;

6）重复步骤1-5），并根据步骤5）获得的车辆跟踪结果，最终判断是否出现停车事件： 6) Repeat steps 1-5), and according to the vehicle tracking results obtained in step 5), finally determine whether there is a parking event:

统计车辆目标累积成功跟踪的帧数T；当跟踪结果J为1时，T加1，J为0时，T不变； Count the cumulative number of frames T of successful tracking of the vehicle target; when the tracking result J is 1, add 1 to T, and when J is 0, T remains unchanged;

统计车辆目标连续跟踪失败的帧数F；当跟踪结果J为0时，F加1，J为1时，F置0； Count the number of frames F of continuous vehicle target tracking failures; when the tracking result J is 0, add 1 to F, and when J is 1, set F to 0;

当车辆目标累积成功跟踪的帧数T≥Alert时，判断出现停车；其中Alert>0，为停车检测报警阈值。 When the accumulative number of successfully tracked frames of the vehicle target T≥Alert, it is judged that parking occurs; where Alert>0 is the parking detection alarm threshold. the

进一步，所述步骤1）中，还根据隧道特征确定图片中的感兴趣区域；步骤2-6）为对感兴趣区域进行的处理。 Further, in the step 1), the region of interest in the picture is also determined according to the characteristics of the tunnel; step 2-6) is the processing of the region of interest. the

进一步，所述步骤2）中，具体包括如下步骤： Further, the step 2) specifically includes the following steps:

21）建立背景模型； 21) Establish a background model;

22）对背景模型实时更新； 22) Update the background model in real time;

23）将当前帧图像与背景图像作差； 23) Make a difference between the current frame image and the background image;

24）将背景差分的图像进行二值化； 24) Binarize the background difference image;

25）将二值化的图像进行形态学处理。 25) Perform morphological processing on the binarized image. the

进一步，所述步骤21）中，采用高斯法建立背景模型；所述步骤22）中，采用基于像素变化率的背景更新方法；所述步骤24）中二值化阈值的选取采用最大类间方差法；所述步骤25）中，选择开运算法进行形态学处理。 Further, in the step 21), the Gaussian method is used to establish the background model; in the step 22), the background update method based on the pixel change rate is adopted; in the step 24), the selection of the binarization threshold adopts the maximum inter-class variance method; in the step 25), the open operation algorithm is selected for morphological processing. the

进一步，所述步骤3）中，具体包括如下步骤： Further, the step 3) specifically includes the following steps:

31）处理前景目标，获得目标的轮廓面积Area； 31) Process the foreground target and obtain the outline area Area of the target;

32）处理前景目标，获得目标的质心横坐标Cen_X和纵坐标Cen_Y； 32) Process the foreground target and obtain the abscissa Cen_X and ordinate Cen_Y of the target's centroid;

33）获取前景目标的外接矩形，计算目标的外接矩形的宽高比W_H；采用公式 $W_H = \frac{Wide}{Height};$ 33) Obtain the bounding rectangle of the foreground target, and calculate the aspect ratio W_H of the bounding rectangle of the target; use the formula $W_h = \frac{wide}{Height};$

其中Wide表示目标外接矩形的宽度，Height表示目标外接矩形的高度。 Among them, Wide represents the width of the bounding rectangle of the target, and Height represents the height of the bounding rectangle of the target. the

进一步，所述步骤4）具体包括如下步骤： Further, the step 4) specifically includes the following steps:

41）选择初始跟踪目标。初始跟踪目标的纵坐标Cen_Y需满足：|Cen_Y-Bot|≤Dis_i，其中Bot表示步骤一当中所选择的车辆检测区域的最小纵坐标，Dis_i表示初始目标选择的阈值。 41) Select the initial tracking target. The ordinate Cen_Y of the initial tracking target needs to satisfy: |Cen_Y-Bot|≤Dis_i, where Bot represents the minimum ordinate of the vehicle detection area selected in step 1, and Dis_i represents the threshold for initial target selection. the

42）判断目标轮廓面积是否匹配。 42) Determine whether the target contour area matches. the

计算当前目标和跟踪目标的轮廓面积之比

其中Area_tar表示当前目标的轮廓面积，Area_ini表示跟踪目标的轮廓面积。 Calculate the ratio of the contour area of the current target and the tracked target

Among them, Area_tar represents the contour area of the current target, and Area_ini represents the contour area of the tracking target.

判断当前目标和跟踪目标的轮廓面积之比Area_div是否满足：Area_min≤Area_div≤Area_max，其中Area_min表示轮廓面积之比的下限，Area_max表示轮廓面积之比的上限，且0<Area_min<Area_max。 Determine whether the ratio Area_div of the contour area of the current target and the tracking target satisfies: Area_min≤Area_div≤Area_max, where Area_min represents the lower limit of the ratio of the contour area, Area_max represents the upper limit of the ratio of the contour area, and 0<Area_min<Area_max. the

43）判断目标外接矩形的宽高比是否匹配。 43) Determine whether the aspect ratio of the bounding rectangle of the target matches. the

计算当前目标和跟踪目标的外接矩形宽高比之比

其中W_H_tar表示当前目标的宽高比，W_H_ini表示跟踪目标的宽高比。 Calculate the ratio of the bounding rectangle aspect ratio of the current target and the tracked target

Among them, W_H_tar represents the aspect ratio of the current target, and W_H_ini represents the aspect ratio of the tracking target.

判断当前目标和跟踪目标的外接矩形宽高比之比W_H_div是否满足：W_H_min≤W_H_div≤W_H_max，其中W_H_min表示宽高比之比的下限，W_H_max表示宽高比之比的上限，且0<W_H_min<W_H_max。 Determine whether the aspect ratio W_H_div of the circumscribed rectangle of the current target and the tracking target satisfies: W_H_min≤W_H_div≤W_H_max, where W_H_min represents the lower limit of the aspect ratio, W_H_max represents the upper limit of the aspect ratio, and 0<W_H_min< W_H_max. the

44）判断目标质心位置是否匹配。 44) Determine whether the target centroid position matches. the

计算当前目标和初始目标的质心位置距离 $Cen_dis = \sqrt{{(Cen_X_tar - Cen_X_ini)}^{2} + {(Cen_Y_tar - Cen_Y_ini)}^{2}},$ 其中Cen_X_tar、Cen_Y_tar表示当前目标的质心坐标，Cen_X_ini,Cen_Y_ini表示跟踪目标的质心坐标。 Calculate the distance between the current target and the initial target's centroid position $Cen_dis = \sqrt{{(Cen_x_tar - Cen_x_ini)}^{2} + {(Cen_Y_tar - Cen_Y_ini)}^{2}},$ Among them, Cen_X_tar and Cen_Y_tar represent the coordinates of the center of mass of the current target, and Cen_X_ini and Cen_Y_ini represent the coordinates of the center of mass of the tracking target.

判断当前目标和初始目标的质心位置距离Cen_dis是否满足：Cen_dis_min≤Cen_dis≤Cen_dis_max，其中Cen_dis_min表示质心位置距离的下限，Cen_dis_max表示质心位置距离的上限，且0<Cen_dis_min<Cen_dis_max。 Determine whether the centroid position distance between the current target and the initial target is Cen_dis: Cen_dis_min≤Cen_dis≤Cen_dis_max, where Cen_dis_min represents the lower limit of the centroid position distance, Cen_dis_max represents the upper limit of the centroid position distance, and 0<Cen_dis_min<Cen_dis_max. the

45）得出车辆目标的跟踪结果J。只有轮廓面积、外接矩形宽高比和质心位置都满足要求时，J为1，否则为0。 45) Obtain the tracking result J of the vehicle target. Only when the contour area, the aspect ratio of the circumscribed rectangle and the position of the centroid all meet the requirements, J is 1, otherwise it is 0. the

本发明的基于视频检测技术的高速公路隧道停车检测方法，本发明针对高速公路隧道场景，选择目标的轮廓面积、质心位置和外接矩形宽高比，采用基于多特征融合的车辆跟踪方法，根据车辆在场景中被跟踪的视频帧数，并结合跟踪结果的统计情况，来判断是否出现停车事件。运算开销小、实时性强可准确、高效地解决高速公路隧道内的停车事件的自动检测问题，提升高速公路隧道的交通运行安全水平。 The highway tunnel parking detection method based on video detection technology of the present invention, the present invention is aimed at the highway tunnel scene, selects the contour area, centroid position and circumscribed rectangle aspect ratio of the target, adopts a vehicle tracking method based on multi-feature fusion, according to the vehicle The number of video frames being tracked in the scene, combined with the statistics of the tracking results, to determine whether there is a parking incident. With low computing overhead and strong real-time performance, it can accurately and efficiently solve the problem of automatic detection of parking events in expressway tunnels, and improve the traffic safety level of expressway tunnels. the

具体实施方式Detailed ways

图1示出了基于视频检测技术的高速公路隧道停车检测方法的流程示意图。 Fig. 1 shows a schematic flow chart of a parking detection method for expressway tunnels based on video detection technology. the

具体实施方式 Detailed ways

以下将结合附图，对本发明进行详细说明。 The present invention will be described in detail below in conjunction with the accompanying drawings. the

参见图1，本实施例的基于视频检测技术的高速公路隧道停车检测方法，包括如下步骤“ Referring to Fig. 1, the highway tunnel parking detection method based on video detection technology of the present embodiment comprises the following steps "

1）从高速公路隧道摄像头获取的视频中按帧抽取图片；还根据隧道特征确定图片中的感兴趣区域，即标定停车检测区域；步骤2-6）为对感兴趣区域进行的处理。标定感兴趣区域能够使处理的图像范围变小，从而减少处理的数据量，提高算法的高效性。此外标定感兴趣区域还可以避免一些干扰因素对检测结果的影响。感兴趣区域一般选择高速公路隧道内车辆可到达区域，包括正常的行车道和紧急停车道。 1) Extract pictures frame by frame from the video captured by the highway tunnel camera; also determine the region of interest in the picture according to the characteristics of the tunnel, that is, calibrate the parking detection area; steps 2-6) are the processing of the region of interest. Marking the region of interest can reduce the range of processed images, thereby reducing the amount of processed data and improving the efficiency of the algorithm. In addition, calibrating the region of interest can also avoid the influence of some interference factors on the detection results. The area of interest is generally selected as the accessible area of vehicles in the expressway tunnel, including normal driving lanes and emergency parking lanes. the

具体包括如下步骤： Specifically include the following steps:

21）建立背景模型。采用高斯法建立背景模型。 21) Build a background model. The background model was established by Gaussian method. the

22）对背景模型实时更新。采用基于像素变化率的背景更新方法，即连续三帧图像中变化率小于某一阈值的像素点以一定的更新率实时更新。 22) Update the background model in real time. The background update method based on the pixel change rate is adopted, that is, the pixels whose change rate is less than a certain threshold in three consecutive frames of images are updated in real time at a certain update rate. the

23）将当前帧图像与背景图像作差。 23) Make a difference between the current frame image and the background image. the

24）将背景差分的图像进行二值化。二值化阈值的选取采用ostu算法（最大类间方差法，被公认为图像分割阈值选择的最佳方法）。 24) Binarize the background difference image. The selection of the binarization threshold adopts the ostu algorithm (the method of maximum variance between classes, which is recognized as the best method for image segmentation threshold selection). the

25）将二值化的图像进行形态学处理。形态学处理方法选择开运算，目的是去除较小的噪声并且能够填充一些空隙。 25) Perform morphological processing on the binarized image. The morphological processing method chooses the open operation, the purpose is to remove small noise and fill some gaps. the

3）获取前景目标轮廓面积、质心位置和外接矩形宽高比；具体包括如下步骤： 3) Obtain the area of the foreground target outline, the position of the centroid and the aspect ratio of the circumscribed rectangle; the specific steps are as follows:

4）采用多特征匹配方法对车辆进行跟踪；具体包括如下步骤： 4) Use the multi-feature matching method to track the vehicle; specifically include the following steps:

计算当前目标和跟踪目标的轮廓面积之比

计算当前目标和跟踪目标的外接矩形宽高比之比

计算当前目标和初始目标的质心位置距离 $Cen_dis = \sqrt{{(Cen_X_tar - Cen_X_ini)}^{2} + {(Cen_Y_tar - Cen_Y_ini)}^{2}},$ 其中 Cen_X_tar、Cen_Y_tar表示当前目标的质心坐标，Cen_X_ini,Cen_Y_ini表示跟踪目标的质心坐标。 Calculate the distance between the current target and the initial target's centroid position $Cen_dis = \sqrt{{(Cen_x_tar - Cen_x_ini)}^{2} + {(Cen_Y_tar - Cen_Y_ini)}^{2}},$ Among them, Cen_X_tar and Cen_Y_tar represent the coordinates of the center of mass of the current target, and Cen_X_ini and Cen_Y_ini represent the coordinates of the center of mass of the tracking target.

45）得出车辆目标的跟踪结果J，只有轮廓面积、外接矩形宽高比和质心位置都满足要求时，J为1，否则为0。 45) Get the tracking result J of the vehicle target. Only when the outline area, the aspect ratio of the circumscribed rectangle and the position of the centroid meet the requirements, J is 1, otherwise it is 0. the

实验结果表明，由于隧道内灯光照明的影响，有时会在图像中形成局部光斑。这种局部光斑是图像中的敏感区域，亮度容易发生变化，经常被处理成前景，通过背景重建和背景更新也很难消除。此类光斑刚出现时，如果不加以区分，会被当作初始跟踪目标。由于监控视频中车辆均从图像底部进入监控区域，而车辆目标在单帧图像中运动的距离有限，所以初始跟踪目标距离图像底部的距离较小。 The experimental results show that, due to the influence of the lighting in the tunnel, sometimes local light spots will be formed in the image. This kind of local light spot is a sensitive area in the image, and its brightness is easy to change. It is often processed as a foreground, and it is difficult to eliminate it through background reconstruction and background update. When this kind of light spot first appears, if it is not distinguished, it will be regarded as the initial tracking target. Since the vehicles in the surveillance video enter the monitoring area from the bottom of the image, and the moving distance of the vehicle target in a single frame image is limited, the distance between the initial tracking target and the bottom of the image is relatively small. the

基于车辆跟踪的停车检测方法通过检测车辆质心位置是否固定来得到最后的判别结果，但是在高速公路隧道中，由于车灯闪烁的影响，车辆跟踪过程中会出现间断的情况，另外车辆停止时车辆质心位置会发生跳变，采用这种方法无法得到正确的判断结果。对于正常行驶的车辆而言，由于监控场景的大小有限，在经过一定时间后车辆会驶出监控区域，从而被成功跟踪的帧数较少。而停止的车辆，成功跟踪的帧数就会较多。 The parking detection method based on vehicle tracking obtains the final judgment result by detecting whether the position of the vehicle's center of mass is fixed. However, in expressway tunnels, due to the influence of flickering lights, there will be interruptions in the vehicle tracking process. In addition, when the vehicle stops, the vehicle The position of the center of mass will jump, and the correct judgment result cannot be obtained by using this method. For a normal driving vehicle, due to the limited size of the monitoring scene, the vehicle will drive out of the monitoring area after a certain period of time, so the number of successfully tracked frames is small. For a stopped vehicle, more frames are successfully tracked. the

由于车灯闪烁的影响，车辆跟踪失败时，车辆很可能没有驶出监控区域，此时不能立即删除目标，而应该保留目标等待下一帧的跟踪。只有当连续跟踪失败的帧数较多时，车辆才驶出监控区域，此时才应该删除目标。因此，从车辆进入监控场景开始，统计车辆目标累积跟踪成功的帧数和连续跟踪失败的帧数。再依据统计的结果提出停车事件的判断方法。 Due to the impact of flickering lights, when the vehicle tracking fails, the vehicle may not have driven out of the monitoring area. At this time, the target cannot be deleted immediately, but the target should be kept for the next frame of tracking. Only when the number of consecutive tracking failure frames is large, the vehicle will leave the monitoring area, and the target should be deleted at this time. Therefore, from the moment the vehicle enters the monitoring scene, the cumulative number of frames that successfully track the vehicle target and the number of frames that fail to track continuously are counted. Then according to the statistical results, the judgment method of the parking event is put forward. the

最后说明的是，以上实施例仅用以说明本发明的技术方案而非限制，尽管参照较佳实施例对本发明进行了详细说明，本领域的普通技术人员应当理解，可以对本发明的技术方案进行修改或者等同替换，而不脱离本发明技术方案的宗旨和范围，其均应涵盖在本发明的权利要求范围当中。 Finally, it is noted that the above embodiments are only used to illustrate the technical solutions of the present invention without limitation. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be carried out Modifications or equivalent replacements without departing from the spirit and scope of the technical solution of the present invention shall be covered by the claims of the present invention. the

Claims

1. the freeway tunnel parking detection method based on video detection technology, is characterized in that: comprise the steps:

1) the video obtaining from freeway tunnel camera, extract frame by frame picture;

2) utilize the picture extracting in step 1), extract foreground target;

3) obtain foreground target contour area, centroid position and boundary rectangle the ratio of width to height;

4) adopt many feature matching methods to follow the tracks of vehicle;

5) draw the tracking results J of vehicle target, while only having contour area, boundary rectangle the ratio of width to height and centroid position all to meet the demands, J is 1, otherwise is 0;

6) repeating step 1-5), and the vehicle tracking result obtaining according to step 5), finally judge whether to occur Parking:

The frame number T that calculating vehicle target accumulated is successfully followed the tracks of; When tracking results J is 1, it is 0 o'clock that T adds 1, J, and T is constant;

The frame number F of calculating vehicle target Continuous Tracking failure; When tracking results J is 0, it is 1 o'clock that F adds 1, J, and F sets to 0;

When vehicle target is accumulated the frame number T >=Alert successfully following the tracks of, stopping appears in judgement; Wherein Alert>0, is parking detection alarm threshold value.

2. the freeway tunnel parking detection method based on video detection technology as claimed in claim 1, is characterized in that: in described step 1), also according to tunnel feature, determine the area-of-interest in picture; Step 2-6) processing for area-of-interest is carried out.

3. the freeway tunnel parking detection method based on video detection technology as claimed in claim 1, is characterized in that: described step 2), specifically comprise the steps:

21) set up background model;

22) to background model real-time update;

23) current frame image and background image is poor;

24) image of background difference is carried out to binaryzation;

25) image of binaryzation is carried out to morphology processing.

4. the freeway tunnel parking detection method based on video detection technology as claimed in claim 3, is characterized in that: described step 21), adopt Gaussian processes to set up background model; Described step 22), in, adopt the background update method based on change ratio of pixel; Described step 24) in, binary-state threshold chooses employing maximum variance between clusters; Described step 25), in, select opening operation method to carry out morphology processing.

5. the freeway tunnel parking detection method based on video detection technology as claimed in claim 1, is characterized in that: in described step 3), specifically comprise the steps:

31) process foreground target, obtain the contour area Area of target;

32) process foreground target, obtain barycenter horizontal ordinate Cen_X and the ordinate Cen_Y of target;

33) obtain the boundary rectangle of foreground target, calculate the ratio of width to height W_H of the boundary rectangle of target; Adopt formula

Wherein Wide represents the width of target boundary rectangle, and Height represents the height of target boundary rectangle.

6. the freeway tunnel parking detection method based on video detection technology as claimed in claim 5, is characterized in that: described step 4) specifically comprises the steps:

41) select initial tracking target.The ordinate Cen_Y of initial tracking target need meet: | Cen_Y-Bot|≤Dis_i, wherein Bot represents the minimum ordinate in the central selected vehicle detection of step 1 region, Dis_i represents the threshold value that initial target is selected.

42) judge whether objective contour area mates.

Calculate the ratio of the contour area of current goal and tracking target

wherein Area_tar represents the contour area of current goal, and Area_ini represents the contour area of tracking target.

Whether the ratio Area_div that judges the contour area of current goal and tracking target meets: Area_min≤Area_div≤Area_max, wherein Area_min represents the lower limit of the ratio of contour area, Area_max represents the upper limit of the ratio of contour area, and 0<Area_min<Area_max.

43) whether the ratio of width to height that judges target boundary rectangle mates.

Calculate the ratio of boundary rectangle the ratio of width to height of current goal and tracking target

wherein W_H_tar represents the ratio of width to height of current goal, and W_H_ini represents the ratio of width to height of tracking target.

Whether the ratio W_H_div that judges boundary rectangle the ratio of width to height of current goal and tracking target meets: W_H_min≤W_H_div≤W_H_max, wherein W_H_min represents the lower limit of the ratio of the ratio of width to height, W_H_max represents the upper limit of the ratio of the ratio of width to height, and 0<W_H_min<W_H_max.

44) judge whether target centroid position mates.

Calculate the centroid position distance of current goal and initial targetwherein Cen_X_tar, Cen_Y_tar represent the center-of-mass coordinate of current goal, Cen_X_ini, and Cen_Y_ini represents the center-of-mass coordinate of tracking target.

Whether the centroid position distance C en_dis that judges current goal and initial target meets: Cen_dis_min≤Cen_dis≤Cen_dis_max, wherein Cen_dis_min represents the lower limit of centroid position distance, Cen_dis_max represents the upper limit of centroid position distance, and 0<Cen_dis_min<Cen_dis_max.

45) draw the tracking results J of vehicle target, while only having contour area, boundary rectangle the ratio of width to height and centroid position all to meet the demands, J is 1, otherwise is 0.