基于相位一致性检测的目标分割和跟踪方法

在复杂背景下，目标和背景的亮度或对比度变化以及光照不均匀等因素都会影响目标分割的结果，导致跟踪不稳定或失败。针对这些问题，介绍了基于相位一致性检测的目标跟踪算法。该算法利用log Gabor滤波器获取多方向相位一致图像，并根据最小尺度下6个方向log Gabor小波滤波图像的中值平方确定目标响应最大的3个方向，进行目标分割；利用分割结果确定目标大小、质心位置，进而实现波门自适应跟踪。实验表明，该算法不受亮度和对比度变化影响，采用一个固定的分割阈值，就能得到很好的分割结果，实现了对目标的稳定跟踪，具有较高的跟踪精度。

Phase congruency based on target segmentation and tracking method

In complex scene, there are several factors affecting the performance of target segmentation, lead to unstable tracking, such as change of brightness/contrast and un-uniform illumination. To deal with such problems, a phase congruency(PC) based on target tracking method is proposed, which adopts log Gabor to get multi-directional PC images. The algorithm computes the Median Squared Response (MSR) obtained from multi-direction log Gabor filters on minimal transform level, selects three directions with maximal MSR,and performs target segmentation. Based on the segmentation result, target size and its mass center are determined, and adaptive gate target tracking is realized. Experimental results show that the method does not be influenced by the change of target intensity or image contrast, is capable of achieving a satisfied segmentation performance with a constant segmenting threshold, and realizes stable tracking with high tracking precision.