背景运动补偿和假设检验的目标检测算法

根据背景与目标具有不同的运动特性这一事实,提出一种基于背景运动补偿和假设检验的目标检测算法。首先,采用特征点对应法求相邻帧间的运动;然后,用最小二乘法计算出摄像机仿射运动参数,利用此参数进行帧间背景运动补偿后得到稳定的背景;最后,用假设检验方法检测补偿后的帧差图像,经过简单的形态学和连通区域处理后检测出运动目标。仿真表明,该运动补偿算法能有效消除背景突出目标,补偿前后的差分图像信噪比提高了14.71 dB。该算法计算量小,可以成为一种通用的实时目标检测算法。

Object detection algorithm based on background motion compensation and hypothesis test

According to the fact that background and object have different motion features, an object detection algorithm was proposed based on background motion compersation and hypothesis test. Motion vector was detected by a feature point corresponding algorithm and affine motion parameters were calculated through least square method. Difference image after compensation was detected by hypothesis test. After a series of morphological process, we obtained the intact moving object. Experimental results indicate that this algorithm can eliminate background motion effectivelly because of moving cameras and segment object under complex environment. The peak signal noise ratio has been improved 14.71 dB compared with two difference images.