基于序列图像运动分割的车辆边界轮廓提取算法

Edge detection algorithm of moving vehicle based on sequential image motion segmentation

为了准确获得图像感兴趣区中运动车辆的形状特征,提出了一种新的车辆边界轮廓提取算法。利用连续3帧图像,对包含同一运动车辆的图像感兴趣区进行光流场分割,以获取目标运动区域,通过平移运动区域的左、右边界获得正确的车辆区域及其封闭边界轮廓,通过放大运动矢量计算公式的阈值来提高其运行效率。试验结果表明:该算法可从具有复杂自然场景的图像序列中检测出完整的运动车辆边界轮廓,检测正确率在95%以上。

To obtain the shape feature of moving vehicle in interested region, a novel algorithm of edge detection was proposed..After inputting three consecutive frames with interested vehicle and stationary complex background, the motion region of vehicle was detected by image motion segmentation and median filtering operator. The complete region and enclosed contour edge of vehicle were gotten by displacing the left edge or the right one of the motion region in horizontal detection. The run time of the algorithm was reduced by enlarging the threshold of iterative equation for motion vectors. Experiment result shows that in a 30 min sequence of urban traffic, 95% of vehicles are accurately detected in complex background. 1 tab, 10 figs, 10 refs.