An Improved Method of Tracking and Counting Moving Objects Using Graph Cuts

To improve the efficiency of tracking and counting moving objects under occlusion conditions, an improved tracking and counting method is proposed. First, a graph cut method is employed to segment an image from a static scene, and foreground objects are identified by the sizes and positions of foreground areas obtained. Second, to distinguish moving objects, object classification based on shape is applied. In addition, in the object tracking phase, the proposed tracking method is used to calculate the centroid distance of neighboring objects and facilitate object tracking and people counting under occlusion conditions. In the experiments of moving object tracking and people counting in two video clips, compared with traditional methods, the experimental results show that the proposed method can increase the averaged detection ratio by approximately 11 %. Thus, the method can be used to reliably track and count.