改进的ORB算法在有遮挡的车辆跟踪上的应用

Partially occluded vehicle tracking based on ORB

为解决部分遮挡情况下车辆实时跟踪丢失的问题,提出一种基于特征点匹配的改进ORB(improved oriented FAST and rotated BRIEF)算法。在FAST检测角点后用拉普拉斯极值去除虚假角点,相比ORB算法提高了匹配的正确率和检测速度。改进的FAST检测角点速度快,BRIEF(binary robust independent elementary features)描述子缩短建立描述符的时间,减少存储空间,提高特征点匹配的速度,满足实时跟踪的需要。实验结果表明,在有光照变化和噪声干扰的情况下,该算法能够快速准确地跟踪有部分遮挡的车辆。

To solve the real-time tracking problem of vehicle occluded effectively,a method based on improved ORB（improved oriented FAST and rotated BRIEF）feature matching was proposed.FAST was used to detect feature points and Laplace values of the pixel were used to eliminate false corners.Compared with ORB,it improved the accuracy and the speed of matching.Improved FAST got feature points quickly,and BRIEF（binary robust independent elementary features）shortened the time cost of describing the feature and reduced the space of storage to improve the speed of matching.It met the demand of real-time applications.Results of tests show that ORB can track the partially occluded vehicle,even though noise or light changes.