基于特征融合的红外目标跟踪方法

Infrared Target Tracking Algorithm Based on Feature Fusion

为了能够对红外目标实现稳健跟踪,提出一种特征融合算法,将目标的LBP纹理特征和灰度特征空间融合,构造了由基于区域置信度及空间距离加权的LBP特征直方图,以及灰度加权直方图联合表示的目标模型。以均值偏移算法为跟踪框架,在当前帧中迭代搜索目标位置。实验证明该方法对于目标和背景灰度相似以及目标被部分遮挡时的跟踪是有效和稳健的。

An infrared target tracking algorithm is proposed based on feature fusion. By combining grayscale feature space with LBP texture feature space of the target, a target model is constructed and represented by LBP feature histogram and grayscale weighting histogram. The grayscale histogram information and LBP texture feature of the target are integrated in Mean Shift tracking frame after weighted, and the model is updated in time. Experiments showed the algorithm works efficiently and robustly for the obscured object tracking and when the grayscales of target and background are similar.