对光照变化鲁棒的多特征动态提取跟踪算法

针对视觉跟踪算法光照自适应能力差的问题,提出了一种对光照变化鲁棒的多特征动态提取跟踪算法。该算法采用高效克服光照影响的特征提取方法,颜色子模型采用模糊直方图方法获取,在同态滤波基础上建立边缘子模型,运动子模型采用改进的三帧差分法提取。该算法还定义了一个新的特征融合模型,把多种互补的观测子模型动态融合,增强了观测模型的准确性,合理量化特征的可靠性使跟踪更稳定。同时采用改进的粒子重采样方法提高了跟踪准确度。实验结果表明,该算法能有效地避免光照变化对跟踪的影响,具有较好的鲁棒性。

A Robust Multi-feature Dynamic Extraction Tracking Algorithm under Varying Illumination

A robust tracking algorithm based on multi-feature dynamic extraction was proposed to solve the problem of the poor illumination adaptive ability.The accurate and effective feature extraction method was used in the algorithm to overcome the influence of illumination.The color model was built by fuzzy histogram;edge the model was built by homomorphic filtering and gradient operator;and the motion model was built by the improved frame difference method.A new feature fusion model was introduced to put the multiple complementary observation models be fused in the particle filter framework.The feature reliability was then improved;and the tracking was more stable.The improved particle resampling strategy was used to improve the tracking accuracy.Experimental results show that the proposed algorithm is more robust,and can effectively avoid the influence of varying illumination.