改进局部二值模式算法与Camshift结合的目标跟踪方法

为解决视频实景监视系统中因场景光照、阴影及远距离小目标跟踪易丢失问题,提出一种改进局部二值模式(local binary patterns, LBP)算法与Camshift结合的目标跟踪方法。利用LBP算子纹理和颜色对阴影不敏感的特性,采用改进的LBP算子与高斯混合模型结合进行背景建模和目标检测,以抑制阴影的干扰;同时将LBP算子的纹理和颜色融入Camshift算法中,结合Kalman滤波进行目标运动状态的预测,最终实现对监视场景中运动目标的可靠、稳定跟踪。采集行人、车辆及航空器等不同类目标进行实验,验证了本文方法不仅能够稳定、精确地跟踪运动目标,同时可适用于场景雾天低能见度条件下的目标跟踪。

An moving target tracking method based on improved LBP algorithm and Camshift

To solve the moving target tracking loss problem of the illumination, shadow, distant and small target in video real-scene surveillance system, an moving target tracking method based on improved LBP algorithm and Camshift is proposed. Firstly, the improved LBP Operator is combined with Gaussian mixture model for background modeling and target detection to suppress the interference of shadow by that the texture and color of LBP Operator are insensitive to shadow. At the same time, the texture and color of LBP Operator are integrated into Camshift algorithm and the Kalman filter is combined to predict the state of moving target. Finally, moving target is tracked reliably in video surveillance scene. Experiments on pedestrians, vehicles and aircraft show that the proposed method can not only track moving targets steadily and accurately, but also be applied in the foggy low visibility conditions.