采用形态学算子的运动目标检测跟踪方法研究及实现

Research and implementation of moving objects detection and tracking by using morphologic operators

针对背景动态变化的场景,提出了一种基于全方位视觉的运动目标检测跟踪方法.采用统计方法建立背景模型,并实现背景模型的实时更新;利用减背景法和改进的二值图像连通域算法实现运动区域提取、分割;引入形态学算子计算目标区域体态比和紧密度,过滤背景干扰物;采用卡尔曼滤波与匹配矩阵相结合实现多个运动目标的跟踪;通过目标在HSV颜色空间中的H值、目标间的欧氏距离和目标相交面积等特征融合,提高目标跟踪的鲁棒性.实验表明,所设计的方法能实现实时准确的运动目标检测与跟踪.

Aiming at the dynamic changing scene of background, a novel method for moving objects detection and tracing based on omnidirectional vision is presented in this paper. Firstly background is modeled by statistical method and the background model is real-time update. The moving objects extraction and segmentation are implemented by background subtraction technique and improved connected domain algorithm of binary image. In order to filter disruptors in the background, the shape ratio and compactness of object are calculated with morphologic operators. Further more, multi-objects tracking is achieved by combination of kalman filter and matching matrix. Robustness of tracing objects is improved by the fusing the value of hue in HSV color space, the Euclidean distance and the intersection area between objects. The experiment results testify that the method is practicable and accurate in real-time detection and tracking of rmoving objects.