CN102963294B

CN102963294B - Method for judging opening and closing states of high beam of vehicle driving at night

Info

Publication number: CN102963294B
Application number: CN201210431055.9A
Authority: CN
Inventors: 朱虹; 潘行杰; 王栋; 苟荣涛; 何振; 张云星; 杨静; 张晓蕾
Original assignee: Xian University of Technology
Current assignee: Zhejiang Harmony Intelligent Technology Co Ltd
Priority date: 2012-11-02
Filing date: 2012-11-02
Publication date: 2014-12-10
Anticipated expiration: 2032-11-02
Also published as: CN102963294A

Abstract

The invention discloses a method for judging the turned-on state of high beam lights of vehicles driving at night. The placement position of the road is delineated to delineate the effective monitoring area of the road; the image after the extraction and processing of the car light is processed to open and calculate, and the labeling process is completed; the car light of the same vehicle is determined; whether the car light of the same vehicle is in the high beam state, when When condition 1 and condition 2 are satisfied at the same time, it indicates that the high beam of the vehicle is in the on state, otherwise, it is judged to be in the off state; repeat the above steps until all the judgments of all vehicles are completed. The method of the present invention can adapt to different lighting environments of street lights at night, suppress non-light areas by means of polarized light filtering, and then extract features of high beam lights to accurately determine vehicles with high beam lights turned on.

Description

A method for judging whether the high beam headlights of vehicles driving at night are turned on

技术领域technical field

本发明属智能交通监控技术领域，涉及一种夜间行驶车辆远光灯开启状态的判别方法。The invention belongs to the technical field of intelligent traffic monitoring, and relates to a method for judging the on state of high beam lights of vehicles running at night.

背景技术Background technique

视频监控模式在构建智能交通系统中，作为获取并分析交通状况的一种手段，目前已经被广泛应用。对于夜间行驶的车辆，打开车灯照亮路面是安全行驶的基本，然而，远光灯的滥用，会导致会车时对方车辆驾驶员完全看不清前方路面而容易发生车祸，因此，限制远光灯的使用，是保证交通安全的一个必要手段。Video surveillance mode has been widely used as a means of obtaining and analyzing traffic conditions in building intelligent transportation systems. For vehicles driving at night, turning on the headlights to illuminate the road is the basis for safe driving. However, the abuse of high beams will cause the driver of the other vehicle to completely lose sight of the road ahead when passing the car and is prone to car accidents. The use of lights is a necessary means to ensure traffic safety.

采用视频帧图像，分析行驶车辆是否打开远光灯，不失为一种有效的方法。然而，由于夜间车辆行驶过程中，周围环境有很大的不同，例如，有些路段的路灯比较密集，光线很亮，有些路段路灯比较稀疏，光线很暗，这时，拍摄到的车辆远近光灯的形态有比较大的差异，很难进行准确的识别。It is an effective method to analyze whether the driving vehicle turns on the high beam by using the video frame image. However, because the surrounding environment is very different when the vehicle is driving at night, for example, some street lights are dense and the light is very bright, while some street lights are sparse and the light is very dark. There are relatively large differences in the shape of the species, and it is difficult to accurately identify them.

发明内容Contents of the invention

本发明的目的是提供一种夜间行驶车辆远光灯开启状态的判别方法，解决了现有技术受周围环境影响，导致拍摄到的车辆远近光灯的形态有比较大的差异，很难进行准确识别的问题。The purpose of the present invention is to provide a method for judging the state of the high beam lights of vehicles driving at night, which solves the problem that the existing technology is affected by the surrounding environment, resulting in relatively large differences in the form of the high and low beam lights of the photographed vehicles, and it is difficult to accurately determine problem of identification.

本发明所采用的技术方案是，一种夜间行驶车辆远光灯开启状态的判别方法，按照以下步骤实施：The technical solution adopted in the present invention is a method for discriminating the open state of the high beam of a vehicle running at night, which is implemented according to the following steps:

步骤1、拍照得到行驶车辆的车灯偏光滤波图像Step 1. Take a photo to obtain the polarized filter image of the headlights of the driving vehicle

首先，在监控摄像头的镜头前设置水平偏振片和垂直偏振片，通过调节两个偏振片之间的夹角，控制摄像头的进光量，直到捕获的画面仅保留行驶车辆的车灯部分的亮光后，固定调整好这两个偏振片，实现对行驶车辆的偏光滤波拍照；First, install a horizontal polarizer and a vertical polarizer in front of the lens of the surveillance camera. By adjusting the angle between the two polarizers, the amount of light entering the camera is controlled until the captured picture only retains the bright light of the headlights of the driving vehicle. , fixedly adjust the two polarizers, and realize the polarized filter photography of the driving vehicle;

步骤2、在上步处理过的偏光滤波图像中提取车灯区域Step 2. Extract the headlight area from the polarized filter image processed in the previous step

2.1)去除转向灯区域2.1) Remove the turn signal area

对读入的当前时刻，即t时刻的经过偏光滤波后的视频帧彩色图像的红、绿、蓝三个分量为[r_t(x,y)]_m×n、[g_t(x,y)]_m×n、[b_t(x,y)]_m×n，其中，m×n表示帧图像的大小为m行n列，(x,y)表示像素点的坐标，[·]表示整个图像；For the current moment read in, that is, the red, green and blue components of the color image of the video frame after polarization filtering at time t are [r _t (x,y)] _m×n , [g _t (x,y )] _m×n , [b _t (x,y)] _m×n , where m×n means that the size of the frame image is m rows and n columns, (x, y) means the coordinates of the pixel point, and [ ] means the whole image;

设转向灯判断矩阵为[L_t(x,y)]_m×n，则判断出转向灯区域的计算公式如下：Assuming that the turn signal judgment matrix is [L _t (x,y)] _m×n , the calculation formula for judging the turn signal area is as follows:

其中，x＝1,2,...,m，y＝1,2,...,n，δ为调整量，Th_r是判别车灯光源的红色分量阈值；Wherein, x=1,2,...,m, y=1,2,...,n, δ is the adjustment amount, Th _r is the red component threshold value for judging the light source of the vehicle light;

2.2)提取图像中所有车辆的远近光车灯区域2.2) Extract the high and low beam headlight areas of all vehicles in the image

将亮度值设为[f_t(x,y)]_m×n，计算公式如下：Set the brightness value as [ft _t (x,y)] _m×n , the calculation formula is as follows:

f_t(x,y)＝0.299·r_t(x,y)+0.587·g_t(x,y)+0.114·b_t(x,y)， (2)f _t (x, y) = 0.299 r _t (x, y) + 0.587 g _t (x, y) + 0.114 b _t (x, y), (2)

对该亮度值进行二值化处理，提取出远近光车灯区域，即二值化图像[g_t(x,y)]_m×n计算公式为：Perform binarization on the luminance value to extract the far and near beam headlight area, that is, the binarized image [g _t (x, y)] _m×n is calculated as follows:

其中，阈值Th的选择范围为Th∈[ε+Δε,ξ-Δξ]，ε为较暗的非车灯区域的亮度，Δε为一增量，ξ为较亮的车灯区域的亮度，Δξ为一增量；Among them, the selection range of the threshold Th is Th∈[ε+Δε,ξ-Δξ], ε is the brightness of the darker non-car light area, Δε is an increment, ξ is the brightness of the brighter car light area, Δξ is an increment;

步骤3、根据摄像头的安放位置，划定道路的有效监控区域Step 3. According to the location of the camera, delineate the effective monitoring area of the road

设定有效监控区域，仅对落入有效监控区域部分的连通域进行分析，即进行以下处理：Set the effective monitoring area, and only analyze the connected domains that fall into the effective monitoring area, that is, perform the following processing:

其中，Ω为所划定的有效监控区域；Among them, Ω is the delineated effective monitoring area;

步骤4、对步骤3处理后的图像进行开运算处理Step 4, performing an opening operation on the image processed in step 3

4.1)对上步得到的进行开运算处理，开运算时，选择的结构元素为N×N，N为奇数，原点为结构元素中心位置，设对进行开运算处理后的结果图为 4.1) For the obtained in the previous step Carry out opening operation processing, during opening operation, the selected structural element is N×N, N is an odd number, the origin is the center position of the structural element, set the pair The result graph after the opening operation is

4.2)对开运算处理后的结果图进行连通域合并处理4.2) The result map after the split operation Perform connected domain merging

为了在车辆远光灯打开时，将同一侧同时点亮的远光灯区域与近光灯区域合并为同一个区域，连通域合并处理的计算公式如下：In order to merge the high beam area and the low beam area that are simultaneously lit on the same side into the same area when the vehicle's high beam is turned on, the calculation formula of the connected domain merging process is as follows:

其中，[a,b]为连通域合并的距离，a＜0,b＞0均为整数；Among them, [a, b] is the distance of merging connected domains, and a<0, b>0 are all integers;

4.3)对连通域合并运算处理后的结果图进行贴标签处理；4.3) The result graph after processing the combined operation of connected domains carry out the labeling process;

步骤5、确定同一车辆的车灯Step 5. Determine the headlights of the same vehicle

按照从上到下，从左到右的顺序，首先选择左侧的连通域，求其质心(x_c,y_c)，质心的计算公式如下：According to the order from top to bottom and from left to right, first select the connected domain on the left, and find its centroid (x _c , y _c ). The formula for calculating the centroid is as follows:

${x x}_{c c} = = \frac{11}{{N N}_{{Ω Ω}_{k k}}} \cdot &Center Dot; \underset{((x x,, y the y)) &Element; &Element; {Ω Ω}_{k k}}{Σ Σ} x x,, - - - - - - ((66))$

${y the y}_{c c} = = \frac{11}{{N N}_{{Ω Ω}_{k k}}} \cdot \cdot \underset{((x x,, y the y)) &Element; &Element; {Ω Ω}_{k k}}{Σ Σ} y the y,, - - - - - - ((77))$

其中，Ω_k为找到的连通域；Among them, Ω _k is the found connected domain;

之后，按照质心的位置(x_c,y_c)画一条水平直线，即在x＝x_c上判断是否有穿越的连通域，Afterwards, draw a horizontal straight line according to the position of the centroid (x _c , y _c ), that is, judge whether there is a crossed connected domain on x=x _c ,

如果有一个，则视其为同一车辆的车灯；If there is one, it is considered to be a headlight of the same vehicle;

如果有若干个，则将最左侧的视为其同一车辆的车灯；If there are several, the leftmost one is regarded as the headlight of the same vehicle;

步骤6、判断同一车辆的车灯是否为远光灯状态Step 6. Determine whether the headlights of the same vehicle are in high beam state

对判断为属于同一车辆的一对车灯区域，设为Ω_Lk和Ω_Rk，判断其联通域的面积特征与形状特征，左侧灯面积S_L与右侧灯面积S_R的特征计算公式分别为： $S_{L} = \underset{(x, y) &Element; Ω_{Lk}}{Σ} \overset{&OverBar;}{g} (x, y) - - - (8)$ For a pair of car light areas judged to belong to the same vehicle, set Ω _Lk and Ω _Rk to determine the area and shape features of the connected domain. The characteristic calculation formulas of the left light area S _L and the right light area S _R are respectively for: $S_{L} = \underset{(x, the y) &Element; Ω_{Lk}}{Σ} \overset{&OverBar;}{g} (x, the y) - - - (8)$

${S S}_{R R} = = \underset{((x x,, y the y)) &Element; &Element; {Ω Ω}_{Rk Rk}}{Σ Σ} \overset{&OverBar; &OverBar;}{g g} ((x x,, y the y)) - - - - - - ((99))$

形状特征为外接矩形的长宽比，分别找到Ω_Lk及Ω_Rk的最左侧、最右侧、最上侧、最下侧的四个点，将该四个点作为外接矩形的左边、右边、上边、下边的起点构成外接矩形，设为Rec_Lk和Rec_Rk，其长、宽分别为L_Lk、L_Rk、H_Lk、H_Rk，形状特征的计算公式如下：The shape feature is the aspect ratio of the circumscribed rectangle. Find the four points on the leftmost, rightmost, uppermost, and lowermost sides of Ω _Lk and Ω _Rk respectively, and use these four points as the left, right, and bottom of the circumscribed rectangle. The starting point of the upper side and the lower side constitutes a circumscribed rectangle, which is set as Rec _Lk and Rec _Rk , and its length and width are L _Lk , L _Rk , H _Lk , H _Rk respectively. The calculation formula of the shape feature is as follows:

${ρ ρ}_{Lk Lk} = = \frac{{H h}_{Lk Lk}}{{L L}_{Lk Lk}} - - - - - - ((1010))$

${ρ ρ}_{Rk Rk} = = \frac{{H h}_{Rk Rk}}{{L L}_{Rk Rk}} - - - - - - ((1111))$

计算远光灯判定条件1为：ρ_Lk+ρ_Rk＜th_ρ， (12)Calculation of high beam judgment condition 1 is: ρ _Lk +ρ _Rk <th _ρ , (12)

其中，th_ρ为远光灯的判断阈值；Among them, th _ρ is the judgment threshold of the high beam;

计算远光灯判定条件2为：S_R+S_L＞2·Th_S， (13)Calculation of high beam judging condition 2 is: S _R + S _L >2 Th _S , (13)

其中，th_S为远光灯的判断阈值，为同一水平位置上所有连通域的平均面积；Among them, th _S is the judgment threshold of the high beam, which is the average area of all connected domains at the same horizontal position;

当条件1和条件2同时满足时，表明该车辆的远光灯处于开启状态，否则，判定为处于关闭状态；When condition 1 and condition 2 are satisfied at the same time, it indicates that the high beam of the vehicle is in the on state, otherwise, it is determined to be in the off state;

重复步骤5、步骤6，直到所有的车辆全部判断完成。Repeat steps 5 and 6 until all vehicles are judged.

本发明的有益效果是，能够适应夜间不同路灯光照环境，通过偏光滤波的方式，抑制非灯光区域，之后提取远光灯特征，准确地确定出打开远光灯的车辆。The beneficial effect of the present invention is that it can adapt to different street lighting environments at night, suppress non-light areas by means of polarized light filtering, and then extract features of high beams to accurately determine vehicles with high beams turned on.

附图说明Description of drawings

图1为拍到的道路上行驶车辆的偏光滤波图像；Figure 1 is a polarized filtered image of a vehicle traveling on the road;

图2为偏光滤波效果示意图，是车辆打开了近光灯状态的图像；Figure 2 is a schematic diagram of the polarization filtering effect, which is an image of the vehicle with the low beam turned on;

图3为偏光滤波效果示意图，是车辆打开了远光灯状态的图像；Figure 3 is a schematic diagram of the effect of polarization filtering, which is an image of a vehicle with high beams turned on;

图4为本发明方法中的车灯区域提取效果示意图，是对图1的二值化图像，虚线连接的是同一个车辆两侧灯；Fig. 4 is a schematic diagram of the extraction effect of the car light area in the method of the present invention, which is the binarized image of Fig. 1, and the dotted lines connect the lights on both sides of the same vehicle;

图5为本发明方法中的车灯区域提取效果示意图，是对图2的二值化图像；Fig. 5 is a schematic diagram of the extraction effect of the car light area in the method of the present invention, which is the binarized image of Fig. 2;

图6为本发明方法中的车灯区域提取效果示意图，是对图3的二值化图像；Fig. 6 is a schematic diagram of the extraction effect of the car light area in the method of the present invention, which is the binarized image of Fig. 3;

图7为本发明方法中的道路有效监控范围示意图；Fig. 7 is a schematic diagram of the effective monitoring range of the road in the method of the present invention;

图8为本发明方法中的贴标签效果示意图，是对图5进行标签处理后的图像；Fig. 8 is a schematic diagram of the labeling effect in the method of the present invention, which is an image after labeling is performed on Fig. 5;

图9为本发明方法中的贴标签效果示意图，是对图6进行标签处理后的图像。Fig. 9 is a schematic diagram of the labeling effect in the method of the present invention, which is an image after labeling processing of Fig. 6 .

具体实施方式Detailed ways

本发明的夜间行驶车辆远光灯开启状态的判别方法，按照以下步骤实施：The method for discriminating the open state of the high beam of the vehicle running at night according to the present invention is implemented according to the following steps:

首先，在监控摄像头的镜头前设置水平偏振片和垂直偏振片，通过调节两个偏振片之间的夹角，控制摄像头的进光量，直到捕获的画面仅保留行驶车辆的车灯部分的亮光后，固定调整好这两个偏振片，实现对行驶车辆的偏光滤波拍照。First, install a horizontal polarizer and a vertical polarizer in front of the lens of the surveillance camera. By adjusting the angle between the two polarizers, the amount of light entering the camera is controlled until the captured picture only retains the bright light of the headlights of the driving vehicle. , fix and adjust the two polarizers, and realize the polarized light filter photography of the driving vehicle.

如图1、图2、图3所示，是实施例中的道路上行驶多个车辆的车灯偏光滤波，以及一个车辆的两个状态下的车灯滤波效果，其中，图1是道路上行驶多个车辆的偏光滤波图像，矩形框内是转向灯，图2车辆打开了近光灯状态的图像，图3是车辆打开了远光灯状态的图像。As shown in Fig. 1, Fig. 2 and Fig. 3, it is the headlight polarization filter of multiple vehicles driving on the road in the embodiment, and the effect of the headlight filter in two states of a vehicle, wherein Fig. 1 is the Polarized filtered images of multiple vehicles, the rectangular frame is the turn signal, Figure 2 is the image of the vehicle with the low beam turned on, and Figure 3 is the image of the vehicle with the high beam turned on.

2.1)去除转向灯区域2.1) Remove the turn signal area

对读入的当前时刻，即t时刻的经过偏光滤波后的视频帧彩色图像的红绿蓝三个分量为[r_t(x,y)]_m×n、[g_t(x,y)]_m×n、[b_t(x,y)]_m×n，其中，m×n表示帧图像的大小为m行n列，(x,y)表示像素点的坐标，[·]表示整个图像；For the current moment read in, that is, the red, green and blue components of the color image of the video frame after polarization filtering at time t are [r _t (x,y)] _m×n , [g _t (x,y)] _m×n 、[b _t (x,y)] _m×n , among them, m×n means that the size of the frame image is m rows and n columns, (x, y) means the coordinates of the pixels, and [ ] means the whole image ;

由于转向灯与远近光灯不同的是，转向灯是红色的，在此，设转向灯判断矩阵为[L_t(x,y)]_m×n，则判断出转向灯区域的计算公式如下：Since the turn signal is different from the high and low beams, the turn signal is red. Here, if the turn signal judgment matrix is [L _t (x,y)] _m×n , the calculation formula for judging the turn signal area is as follows:

其中，x＝1,2,...,m，y＝1,2,...,n，δ为调整量，根据安装设备的地点取值，优选取值范围为[10,40]，Th_r是判别车灯光源的红色分量阈值，优选取值范围为[80,150]；Among them, x=1,2,...,m, y=1,2,...,n, δ is the adjustment value, the value is taken according to the location where the equipment is installed, and the preferred value range is [10,40], Th _r is the red component threshold for judging the light source of the car light, and the preferred value range is [80,150];

对该亮度值进行二值化处理，提取出远近光车灯区域，The brightness value is binarized to extract the far and low beam headlight area,

因为车灯部分与其他部分的亮度差异很大，因此，在此采用全局阈值的方法来实现，即二值化图像[g_t(x,y)]_m×n计算公式为：Because the brightness of the headlight part is very different from other parts, the method of global threshold is used here to realize it, that is, the calculation formula of the binarized image [g _t (x, y)] _m×n is:

其中，阈值Th的选择范围为Th∈[ε+Δε,ξ-Δξ]，ε为较暗的非车灯区域的亮度，Δε为一增量，优选取值范围为Δε∈[0.5ε,ε]，ξ为较亮的车灯区域的亮度，Δξ为一增量，优选取值范围为Δξ∈[0.2ξ,0.8ξ]；ε，ξ的取值，取决于偏光滤波结果。Among them, the selection range of the threshold Th is Th∈[ε+Δε,ξ-Δξ], ε is the brightness of the darker non-car light area, Δε is an increment, and the preferred value range is Δε∈[0.5ε,ε ], ξ is the brightness of the brighter headlight area, Δξ is an increment, and the preferred value range is Δξ∈[0.2ξ,0.8ξ]; the value of ε, ξ depends on the result of polarization filtering.

对于偏光滤波较为理想的情况，即如图1的情况，建议优选最简单地确定阈值为7L/8，其中L为摄像头给定的图像亮度值范围[0,L]的最高值。For the ideal situation of polarization filtering, that is, the situation shown in Figure 1, it is recommended to determine the threshold as the simplest value of 7L/8, where L is the highest value of the image brightness value range [0, L] given by the camera.

如图4、图5、图6所示，是实施例中的车灯区域提取效果，其中，图4是图1道路上行驶中所有车辆的车灯区域提取图像，其中矩形框内是转向灯区域，按照公式(1)计算处理后，被消除，图5是图2的车灯区域，图6是图3的车灯区域。As shown in Figure 4, Figure 5, and Figure 6, it is the extraction effect of the headlight area in the embodiment, wherein, Figure 4 is the image of the headlight area extraction of all vehicles driving on the road in Figure 1, and the rectangular frame is the turn signal The area is eliminated after being calculated and processed according to the formula (1). Fig. 5 is the car light area in Fig. 2, and Fig. 6 is the car light area in Fig. 3 .

考虑到夜间道路两旁可能有店铺或者建筑物上的照明，在二值化时，会被误判为车灯，因此，按照图7所示的示意图，设定有效监控区域，仅对落入有效监控区域部分的连通域进行分析，即进行以下处理：Considering that there may be lights on shops or buildings on both sides of the road at night, they will be misjudged as car lights during binarization. Therefore, according to the schematic diagram shown in Figure 7, set the effective monitoring area, which is only effective for falling The connected domain of the monitoring area is analyzed, that is, the following processing is performed:

其中，Ω为由图7所划定的有效监控区域，Among them, Ω is the effective monitoring area delineated by Fig. 7,

如图7所示，是实施例中的有效监控区域示意图，其中，梯形所框定的部分为监控有效区域，两侧的边界是监控摄像头监视画面中的道路两侧的边界，前后界限是车辆行驶过程中，摄像头能够捕获到足以用来有效判断车灯大小的范围。(如果后界太远，车灯太小，容易被当作噪声去除，如果前界太近，则车灯已不在画面中。)As shown in Figure 7, it is a schematic diagram of the effective monitoring area in the embodiment, wherein the part framed by the trapezoid is the monitoring effective area, the boundaries on both sides are the boundaries on both sides of the road in the surveillance camera monitoring picture, and the front and rear boundaries are the vehicle driving During the process, the camera can capture a range sufficient to effectively judge the size of the headlight. (If the rear boundary is too far away, the headlights are too small, which are easily removed as noise; if the front boundary is too close, the headlights are no longer in the picture.)

4.1)对上步得到的进行开运算处理，以便消除噪声的影响，开运算方法在现有数字图像处理相关专业教材上均有详细说明，这里不再详述。开运算时，选择的结构元素为N×N，N为奇数，原点为结构元素中心位置，设对进行开运算处理后的结果图为 4.1) For the obtained in the previous step The opening operation is performed in order to eliminate the influence of noise. The opening operation method has been described in detail in the existing professional textbooks related to digital image processing, and will not be described in detail here. During the opening operation, the selected structural element is N×N, N is an odd number, and the origin is the center position of the structural element. The result graph after the opening operation is

因为车辆的远光灯和近光灯处于同一水平位置上，当远光灯打开时，近光灯同时亮，这里所述的连通域合并处理，是为了在车辆远光灯打开时，将同一侧同时点亮的远光灯区域与近光灯区域合并为同一个区域，连通域合并处理的计算公式如下：Because the high beam and the low beam of the vehicle are at the same horizontal position, when the high beam is turned on, the low beam is on at the same time, the connected domain merging process described here is to combine the same The high-beam area and the low-beam area that are lit at the same time on both sides are merged into the same area, and the calculation formula for the merging of connected domains is as follows:

其中，[a,b]为连通域合并的距离，a＜0,b＞0均为整数，根据经验取值，建议优选a∈[-8,-4]，b∈[4,8]；Among them, [a, b] is the distance of the connected domain merge, a < 0, b > 0 are all integers, according to the value of experience, it is recommended to choose a∈[-8,-4], b∈[4,8];

4.3)对连通域合并运算处理后的结果图进行贴标签处理4.3) The result graph after processing the combined operation of connected domains Do labeling

贴标签方法在数字图像处理方面的专业教材上均有详细说明，这里不再详述。对图3、图4、图5进行贴标签之后，便可对不同的连通域进行标识，贴标签处理结果如图8、图9所示，其中，图8是对图5进行标签处理后的图像，图9是对图6进行标签处理后的图像，附图中的浅色表示1号标签，深色表示2号标签；The labeling method is described in detail in professional textbooks on digital image processing, and will not be described in detail here. After labeling Figure 3, Figure 4, and Figure 5, different connected domains can be identified, and the labeling processing results are shown in Figure 8 and Figure 9, where Figure 8 is the image after labeling of Figure 5, Fig. 9 is an image after label processing of Fig. 6, the light color in the accompanying drawing indicates label No. 1, and the dark color indicates label No. 2;

在汽车结构布局设计上，将车灯设置于车头两侧同一高度位置，所以，按照从上到下，从左到右的顺序，首先选择左侧的连通域，求其质心(xc,yc)，质心的计算公式如下：In the layout design of the car structure, the lights are set at the same height on both sides of the front of the car. Therefore, in the order from top to bottom and from left to right, first select the connected domain on the left and find its centroid (xc, yc) , the formula for calculating the centroid is as follows:

${y the y}_{c c} = = \frac{11}{{N N}_{{Ω Ω}_{k k}}} \cdot &Center Dot; \underset{((x x,, y the y)) &Element; &Element; {Ω Ω}_{k k}}{Σ Σ} y the y,, - - - - - - ((77))$

之后，按照质心的位置(x_c,y_c)画一条水平直线，即在x＝x_c上判断是否有穿越的连通域，如果有一个，则视其为同一车辆的车灯；如果有若干个，则将最左侧的视为其同一车辆的车灯；Afterwards, draw a horizontal line according to the position of the centroid (x _c , y _c ), that is, judge whether there is a connected domain crossed on x=x _c , if there is one, it is regarded as the headlight of the same vehicle; if there are several , the leftmost one is regarded as the headlight of the same vehicle;

如图4所示，是实施例中的同一车辆车灯检测示意图，其中的虚线给出的直线，连接的就是按照公式(6)，(7)计算出质心位置后，测得的同一车辆的两侧车灯。As shown in Figure 4, it is the same vehicle lamp detection schematic diagram in the embodiment, wherein the straight line given by the dotted line is connected with the measured position of the same vehicle after calculating the centroid position according to the formulas (6) and (7). Headlights on both sides.

${ρ ρ}_{Lk Lk} = = \frac{{H h}_{Lk Lk}}{{L L}_{Lk Lk}} - - - - - - ((1010))$

${ρ ρ}_{Rk Rk} = = \frac{{H h}_{Rk Rk}}{{L L}_{Rk Rk}} - - - - - - ((1111))$

其中，th_ρ为远光灯的判断阈值，远光灯打开时，由于一侧两个灯都点亮，为经验值，所以建议th_ρ的优选取值范围为th_ρ∈[1,1.6]；Among them, th _ρ is the judgment threshold of the high beam. When the high beam is turned on, since both lights on one side are lit, it is an empirical value. Therefore, it is recommended that the optimal value range of th _ρ be th _ρ ∈ [1,1.6] ;

当条件1和条件2同时满足时，表明该车辆的远光灯处于开启状态，否则，判定为处于关闭状态。When condition 1 and condition 2 are satisfied at the same time, it indicates that the high beam of the vehicle is in the on state, otherwise, it is determined to be in the off state.

本发明的方法通过以上的步骤，便可在行驶车辆进入有效监控区域内时，判断其远光灯是否开启，如果属于远光灯开启状态，再利用运动目标跟踪技术，再计算出其在监控区域内持续开启远光灯的时间，为智能交通系统提供一种保障安全的有效检测方法。Through the above steps, the method of the present invention can judge whether the high beam is turned on when the driving vehicle enters the effective monitoring area. The time for the high beams to be turned on continuously in the area provides an effective detection method for the intelligent transportation system to ensure safety.

Claims

1. a kind of discriminant method of the open state of vehicle high beam at night, it is characterized in that, implement according to the following steps:

Step 1. Take a photo to obtain the polarized filter image of the headlights of the driving vehicle

First, install a horizontal polarizer and a vertical polarizer in front of the lens of the surveillance camera. By adjusting the angle between the two polarizers, the amount of light entering the camera is controlled until the captured picture only retains the bright light of the headlights of the driving vehicle. , fixedly adjust the two polarizers, and realize the polarized filter photography of the driving vehicle;

Step 2. Extract the headlight area from the polarized filter image processed in the previous step

2.1) Remove the turn signal area

For the current moment read in, that is, the red, green and blue components of the color image of the video frame after polarization filtering at time t are [r _t (x,y)] _m×n , [g _t (x,y)] _m×n 、[b _t (x,y)] _m×n , among them, m×n means that the size of the frame image is m rows and n columns, (x, y) means the coordinates of the pixels, and [ ] means the whole image ;

Assuming that the turn signal judgment matrix is [L _t (x,y)] _m×n , the calculation formula for judging the turn signal area is as follows:

Wherein, x=1, 2,..., m, y=1, 2,..., n, δ is the adjustment amount, Th _r is the threshold value of the red component for judging the light source of the vehicle light;

2.2) Extract the high and low beam headlight areas of all vehicles in the image

Set the brightness value as [ft _t (x,y)] _m×n , the calculation formula is as follows:

f _t (x, y) = 0.299 r _t (x, y) + 0.587 g _t (x, y) + 0.114 b _t (x, y), (2)

Perform binarization on the luminance value to extract the far and near beam headlight area, that is, the binarized image [g _t (x, y)] _m×n is calculated as follows:

Among them, the selection range of the threshold Th is Th∈[ε+Δε,ξ-Δξ], ε is the brightness of the darker non-car light area, Δε is an increment, ξ is the brightness of the brighter car light area, Δξ is an increment;

Step 3. According to the location of the camera, delineate the effective monitoring area of the road

Set the effective monitoring area, and only analyze the connected domains that fall into the effective monitoring area, that is, perform the following processing:

Among them, Ω is the delineated effective monitoring area;

Step 4, performing an opening operation on the image processed in step 3

4.1) For the obtained in the previous step Carry out opening operation processing, during opening operation, the selected structural element is N×N, N is an odd number, the origin is the center position of the structural element, set the pair The result graph after the opening operation is

4.2) The result map after the split operation Perform connected domain merging

In order to merge the high beam area and the low beam area that are simultaneously lit on the same side into the same area when the vehicle's high beam is turned on, the calculation formula of the connected domain merging process is as follows:

Among them, [a, b] is the distance of merging connected domains, and a<0, b>0 are all integers;

4.3) The result map after the merge operation of the connected domain carry out the labeling process;

Step 5. Determine the headlights of the same vehicle

According to the order from top to bottom and from left to right, first select the connected domain on the left and find its centroid (xc, yc). The formula for calculating the centroid is as follows:

{x x}_{c c} = = \frac{11}{{N N}_{{Ω Ω}_{k k}}} \cdot &Center Dot; \underset{((x x,, y the y)) &Element; &Element; {Ω Ω}_{k k}}{Σ Σ} x x,, - - - - - - ((66))

{y the y}_{c c} = = \frac{11}{{N N}_{{Ω Ω}_{k k}}} \cdot \cdot \underset{((x x,, y the y)) &Element; &Element; {Ω Ω}_{k k}}{Σ Σ} y the y,, - - - - - - ((77))

Among them, Ω _k is the found connected domain;

Afterwards, draw a horizontal straight line according to the position of the centroid (x _c , y _c ), that is, judge whether there is a crossed connected domain on x=x _c ,

If there is one, it is considered to be a headlight of the same vehicle;

If there are several, the leftmost one is regarded as the headlight of the same vehicle;

Step 6. Determine whether the headlights of the same vehicle are in high beam state

For a pair of car light areas judged to belong to the same vehicle, set Ω _Lk and Ω _Rk to determine the area and shape features of the connected domain. The characteristic calculation formulas of the left light area S _L and the right light area S _R are respectively for:

{S S}_{L L} = = \underset{((x x,, y the y)) &Element; &Element; {Ω Ω}_{Lk Lk}}{Σ Σ} \overset{&OverBar; &OverBar;}{g g} ((x x,, y the y)) - - - - - - ((88))

{S S}_{R R} = = \underset{((x x,, y the y)) &Element; &Element; {Ω Ω}_{Rk Rk}}{Σ Σ} \overset{&OverBar; &OverBar;}{g g} ((x x,, y the y)) - - - - - - ((99))

The shape feature is the aspect ratio of the circumscribed rectangle. Find the four points on the leftmost, rightmost, uppermost, and lowermost sides of Ω _Lk and Ω _Rk respectively, and use these four points as the left, right, and bottom of the circumscribed rectangle. The starting point of the upper side and the lower side constitutes a circumscribed rectangle, which is set as ReC _Lk and ReC _Rk , and its length and width are L _Lk , L _Rk , H _Lk , and H _Rk respectively. The calculation formula of the shape feature is as follows:

{ρ ρ}_{Lk Lk} = = \frac{{H h}_{Lk Lk}}{{L L}_{Lk Lk}} - - - - - - ((1010))

{ρ ρ}_{Rk Rk} = = \frac{{H h}_{Rk Rk}}{{L L}_{Rk Rk}} - - - - - - ((1111))

Calculating high beam judgment condition 1 is: ρ _Lk +ρ _Rk <th _ρ , (12)

Among them, th _ρ is the judgment threshold of the high beam;

Calculation of high beam judgment condition 2 is: S _R + S _L >2 Th _S , (13)

Among them, th _S is the judgment threshold of the high beam, which is the average area of all connected domains at the same horizontal position;

When condition 1 and condition 2 are satisfied at the same time, it indicates that the high beam of the vehicle is in the on state, otherwise, it is determined to be in the off state;

Repeat steps 5 and 6 until all vehicles are judged.

2. The method for judging the state of the high beam headlights of vehicles driving at night according to claim 1, characterized in that in the step 2.1), the preferred value range of the adjustment amount δ is [10,40], and the red component The preferred range of threshold Th _r is [80,150].

3. The method for judging the high-beam headlights of vehicles driving at night according to claim 1, characterized in that: in step 2.2),

In general, the preferred value range of the increment Δε is Δε∈[0.5ε,ε], and the preferred value range of the increment Δξ is Δξ∈[0.2ξ, 0.8ξ],

For the ideal case of polarization filtering, the preferred threshold Th is 7L/8, where L is the highest value of the image brightness value range [0, L] given by the camera.

4. The method for judging the state of the high beam headlights of vehicles driving at night according to claim 1, characterized in that: in the step 4.2), preferably a∈[8,-4], b∈[4,8] .

5. The method for judging the state of the high beam headlights of vehicles running at night according to claim 1, characterized in that: in the step 6, the preferred value range of the judgment threshold th _ρ of the high beam headlights is th _ρ ∈ [ 1, 1.6].

6. The method for judging whether the high beam of a vehicle running at night according to any one of claims 1 to 5 is characterized in that: after the step 6, the high beam of the vehicle is calculated by using the moving target tracking technology The amount of time the lights remain on in the monitored area.