为保障工人的安全,对工人是否穿着工作服进行检测与跟踪具有重要意义。针对均值漂移算法中颜色直方图对工作服特征描述不准确、固定核窗宽不能对尺寸逐渐增大的目标进行有效跟踪两方面问题,提出一种基于目标像素中心加权与背景颜色削弱权值的双权值颜色直方图建模法,建立颜色特征突出的工作服模型,并结合工作服两种显著颜色的质心距离变化,在跟踪中调节核窗宽的大小。实验结果表明,提出的方法能够更准确地检测和跟踪工作服。
In order to guarantee the safety of the workers, it has an important significance for detecting and tracking whether the workers dress working clothes. Aiming at the problem of the inaccurate description of working clothes by classic color histogram, and limited performance when tracking a gradual increased target with fixed kernel-bandwidth, a new color histogram model with double weight is proposed, which is weighted by target pixel center and background color weaken coefficient. For adapting to the change of target size, two reliable color centroid distance is used to fix kernel-bandwidth. Experiment results show successful adapt to detecting and tracking of working clothes.