摘要
在传统的基于块匹配算法的运动车辆检测中,因为直接对灰度图像进行匹配而带来计算量大、光照变化较敏感和实时性差的缺陷。针对这些问题,提出了基于概率松弛标记算法的块匹配算法,先通过概率松弛标记算法求取灰度图像的边缘,然后通过块匹配算法根据匹配的相似度得到位移矢量场,并对其进行矢量中值滤波,最后通过顺序区域增长把运动车辆分割出来。实验结果表明,这种算法鲁棒性强、实时性好,与传统算法比较在效率大幅度提高的情况下准确率也得到了一定的提升。
In the traditional moving vehicle object detection based on block matching algorithm,its disadvantages are the long computing time,sensitivity to illumination changes and poor real-time performance because of directly matching gray pictures.So aimed at this problem,we present the block matching algorithm based on the probability relaxation labeling algorithm.First using the probability relaxation labeling algorithm to get the edge of gray pictures,then obtain the displacement vector field by block matching algorithm and vector median filter,finally split out the moving vehicle object by sequence region growth method.Experiment results show the method is robust and real-time.Compared with the traditional algorithm,the efficiency has been improved greatly,and the accuracy has been enhanced too.
出处
《电子测量技术》
2016年第8期75-78,共4页
Electronic Measurement Technology
关键词
矢量中值滤波
运动车辆检测
概率松弛标记算法
vector median filter
moving vehicle detection
probability relaxation labeling algorithm
