Moving object tracking based on mean shift algorithm and features fusion

This paper presents a method for moving objects tracking in complicated areas based on features fusion and mean shift algorithm. Primary mean shift algorithm is only based on colour feature, and has a suitable performance especially in partial occlusions. However, primary mean shift algorithm would fail to track in some conditions. This paper exhibits a method to solve mean shift algorithm problems with combination of colour and edge features in light change conditions and in presence of several objects with same colour. In the proposed method, after extracting the histogram for each feature of moving object, mean shift algorithm is applied separately. In the next step, we combined mean shift algorithm outputs with suitable coefficients. We defined colour and edge coefficient using Bhattacharyya concept so that each feature has the best outcome on the tracking final target location. Results show whereas primary mean shift algorithm misses the target, the proposed algorithm has reasonable performance.