基于Sobel算子和CNN的车辆轨迹识别系统

Vehicle Trajectory Recognition System Based on Sobel Operator and CNN

随着机动车保有量的飞速增长,由此带来的交通安全问题以及如何有效地监控交通运输车辆和分析车辆轨迹行为成为当前社会关注的热点之一。对目前现有的车辆轨迹识别技术进行了改进。车辆的轨迹识别主要包括车辆目标识别和连续视频帧中车辆位置的定位两个部分。采用HSV颜色模型和Sobel算子相结合的技术来快速提取车辆目标的候选区域信息,将该信息输入到训练好的CNN模型中以完成车辆目标信息的提取。然后根据前后帧的关联性提取每帧视频中车辆的位置,通过以上步骤可以计算得到车辆一系列连续轨迹点,进而分析车辆在行驶过程中的状态。实验结果表明,该系统不仅能够准确地识别出视频中车辆目标,而且在连续视频帧中能够有效地对车辆轨迹进行跟踪识别。

With the increasing of motor vehicles rapidly,the resulting traffic safety and howto effectively monitor the traffic vehicles and analyze the vehicle trajectory behavior become one of the social hot spots.For this,we have improved the existing technologies about the trajectory recognition of vehicle which mainly includes two parts： the identification of vehicle targets and the position of the vehicle between the continuous video frames.Using the HSV color model and the Sobel operator can quickly extract the candidate region＇s information of the vehicle target and the information is input to the trained CNN model for extraction of the information of vehicle target. And then according to the correlation between the front and rear frames,the position of the vehicle in each frame can be extracted.Through the above steps we can calculate a series of continuous track points of the vehicle,and then we can analyze the state of the vehicle on the road.The experiment shows that the system can not only identify the vehicle target in the video,but also can track the vehicle trajectory in the continuous frames.