基于局部不变特征与Camshift复杂环境跟踪技术研究

Research on complex environment vehicle tracking technology based on local invariant feature and Camshift

针对城市道路交通中广泛存在由于车辆形变、雾霾天气、背景噪声以及光照变化、遮挡干扰等复杂环境导致传统方法跟踪失败的问题,提出在连续自适应均值漂移算法的基础上融合HSV颜色模型、局部不变特征的城市道路交通车辆目标跟踪新算法。新算法通过建立基于颜色、灰度的多特征模板,利用统计分析学样本主成分分析降维提高特征匹配的效率,定义并计算加权特征在空间上的分布,使得算法对复杂环境亦能较好的应对,保证跟踪稳定、准确。实验表明新算法识别率高、实时性好,抗环境干扰鲁棒性强。

Since the traditional tracking methods are failed due to the complex environments of vehicle deformation, haze weather, background noise, illumination variation and occlusion interference existing in urban road traffic, the vehicle target tracking new algorithm of urban road traffic is proposed, which integrates HSV color model and local invariant feature with the continuously adaptive mean shift （Camshift） algorithm. The multi-feature template based on color and gray level is established. The sample principal component analysis method of statistic analysis is used to reduce the feature dimension to improve the effi- ciency of feature matching. The spatial distribution of weighted features is defined and calculated, so the algorithm can better deal with the complex traffic environment, and ensure the stable and accurate tracking. The experimental results show that the new algorithm has high recognition rate, good real-time performance, and strong robust to resist environmental interference.