【摘要】 为解决传统车辆检测和计数方法局限性的问题,设计一种基于人机交互界面的车辆检测与计数方法.利用灰度世界算法进行图像预处理,去除灰度突变;利用改进混合高斯法建立背景模型,并选取像素信息进行背景自适应更新;利用五帧差分法提取前景目标,提高检测的准确性;划定感兴趣区域并设置虚拟绊线使检测更有针对性;利用目标团块分析法进行目标跟踪,对满足条件的车辆进行计数,提高了跟踪鲁棒性.在实际交通场景中进行试验,将过程和结果显示在人机交互界面上,增强信息传递效率.试验结果表明,车辆计数的准确度较高.更多还原

【Abstract】 To solve the problem with the limitation of traditional vehicle detection and counting methods, vehicle detection and counting technology is designed based on human-computer interaction interface. Gray world algorithm for image preprocessing is used to remove gray-scale mutations, the improved Gaussian mixture is used to establish the background model, and current pixel information is selected to realize the adaptive update of the background. Five-frame difference is applied to extract foreground objects to improve the accuracy of detection. The detection is made more targeted by delineating the area of interest and setting up virtual tripwires. Target clump analysis method is then used to track the target to improve the tracking robustness and count the vehicles that meet the conditions. Experiments are carried out in different actual traffic scenarios, the process and results of which are displayed on the human-computer interaction interface to increase efficiency of information transmission. The results show that the accuracy of vehicle counting is high.更多还原