基于运动目标识别的摩托车视频监控驾考识别技术研究

Research on Motorcycle Video Surveillance Driving Test Recognition Technology Based on Moving Target Recognition

针对人工监考摩托车驾考往往存在徇私舞弊、判断不准确等缺点,提出一种基于运动目标识别的摩托车视频监控驾考识别技术.基于改进ORB图像特征匹配算法、消除误匹配点等策略对摩托车驾考监控视频进行稳像处理;使用边缘轮廓检测的运动目标识别算法,完成运动中摩托车目标图像边缘轮廓特征检测,利用双目视觉极线约束模型对摩托车目标特征点进行立体匹配与三维重构,提高轮廓检测精度.测试结果显示,该方法检测摩托车驾考目标的最优准确率可达96%,检测算法运行的响应时间较少、实时性优,具有良好的视频监控驾考适应性.

Manual invigilation of motorcycle driving test often has the shortcomings such as favoritism and inaccurate judgment.For this reason,a modern motorcycle video surveillance driving test recognition technology based on the research of moving target recognition technology is proposed.Based on the strategy of improving the ORB image feature matching algorithm and eliminating mismatched points,the motorcycle driving test surveillance video is stabilized;the moving target recognition algorithm based on edge contour detection completes the edge contour feature detection of the motorcycle target image in motion.Binocular visual epipolar constraint model is used to perform stereo matching and three-dimensional reconstruction on the target feature points of motorcycles,which improves the accuracy of contour detection.The test results show that the optimal accuracy rate of this method for detecting motorcycle driving test targets can reach 96%;the detection algorithm runs less response time,has and excellent real-time performance and good video surveillance driving test adaptability are achieved.