基于车载视频的占用公交车道违章检测算法研究

发布时间：2018-11-02 09:07

【摘要】：随着我国经济的快速发展,公共交通也得到了快速发展,大大方便了人们的出行,但是随着汽车保有量持续上升,社会车辆占用公交车道的现象越来越频繁,严重影响了正常的交通秩序。传统的定点监控和人工监控的方法越来越难以应对日益严重的公交车道违章现象。作为一种新的智能监控技术,占用公交车道违章检测系统通过架设在公交车上的监控摄像头获取公交车前方的路况,能够自动识别场景中的公交车道,并检测出公交车道内的违章车辆,并进行报警,相对于传统的检测方式,该系统有效降低了人工的干预,提高了检测效率。本文详细介绍了基于车载视频的占用公交车道违章检测算法的组成和实现,在车道检测和识别模块中,我们使用了改进的Hough变换算法,检测出左右行车道,然后利用普通车道(白色)与公交车道(黄色)在颜色上的不同,训练出SVM分类器,然后对车道进行判断分类,并通过状态机的方法保持车道检测的稳定性。在车牌检测与判别模块中,首先使用基于Haar特征的Adaboost级联分类器对检测区域中的车牌目标进行检测,对于可能出现的误检情况,又分析了车牌的边缘和颜色特征,提出了Sobel边缘特征验证和RGB特征验证的双重验证方法对车牌进行验证,当确定目标是车牌后,我们通过将车牌图像转换到HSV颜色空间,提取出车牌区域的H特征,利用蓝色像素比例和SVM分类器对车牌颜色进行分类,判断车牌是否属于普通车牌,并对检测结果使用了LK光流法进行了跟踪。然后判断车牌是否在公交车道内,从而检测出违章车牌。基于车载视频的占用公交车道违章检测系统有效减少了人为因素的干预,同时该系统保证了违章的检测效率和准确率。
[Abstract]:With the rapid development of economy in our country, public transportation has also been developed rapidly, which greatly facilitates the travel of people. However, with the continuous increase in the number of cars, the phenomenon of social vehicles occupying bus lanes is becoming more and more frequent. Seriously affected the normal traffic order. The traditional methods of fixed-point monitoring and manual monitoring are becoming more and more difficult to deal with the increasingly serious phenomenon of bus lane violations. As a new intelligent monitoring technology, occupying the bus lane violation detection system can automatically identify the bus lane in the scene by using the surveillance camera installed on the bus to obtain the road condition in front of the bus. The system can effectively reduce the manual intervention and improve the detection efficiency compared with the traditional detection method. This paper introduces in detail the composition and implementation of the algorithm of detecting the violation of the bus lane based on the vehicle video. In the module of lane detection and recognition, we use the improved Hough transform algorithm to detect the left and right carriageways. Then the SVM classifier is trained by using the color difference between the ordinary lane (white) and the bus lane (yellow), and then the lane is judged and classified, and the stability of lane detection is maintained by the state machine method. In the module of license plate detection and discrimination, the Adaboost concatenated classifier based on Haar feature is used to detect the license plate target in the detection area, and the edge and color features of the license plate are analyzed for the possible false detection. A double verification method of Sobel edge feature verification and RGB feature verification is proposed to validate the license plate. When the target is the license plate, we extract the H feature of the license plate region by converting the license plate image to the HSV color space. The blue pixel ratio and SVM classifier are used to classify the color of the license plate to determine whether the license plate belongs to the ordinary license plate or not. The detection results are tracked by the LK optical flow method. Then determine whether the license plate is in the bus lane, so as to detect illegal license plate. The system based on vehicle video can effectively reduce the interference of human factors and ensure the detection efficiency and accuracy.
【学位授予单位】：华中科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U495;TP391.41

【相似文献】