高速道路交通视频中车辆目标提取研究

为了能够从监控视频中快速、准确地分析车辆目标,提出了基于感兴趣区域(ROI)的车辆目标提取方法.针对高速公路监控视频,利用混合高斯背景建模,在视频中划定ROI,以排除逆向车道车辆目标的影响,应用图像形态学进行干扰点排除与前景图像轮廓空洞填充,对运动车辆目标进行检测后,用最小矩形方框法自动截取目标,最终,通过图像尺度归一化建立车辆样本数据库,为车型分类和识别提供目标图像.实验结果表明:该方法对车辆目标提取准确率高,且图像数据库样本丰富.

Research on video vehicle object extraction based on high-speed road traffic

In order to analyze on vehicle target from the surveillance video quickly and accurately,vehicle target extraction method based on region of interest (ROI)area is proposed. Aiming at highway surveillance video,use Gaussian mixture background for modeling,draw ROI area invideo,to eliminate the influence of the reverse drive vehicle target,by method of image morphology for eliminating interference points and image contour cavity filling, after detecting vehicle moving target automatically intercept target with minimum rectangular box method,and finally build vehicles sample database by normalized image scale,provide target image for vehicle classification and recognition. The experimental results show that extraction accuracy of vehicle target by the method is high,and be able to rich the image database sample.