A new moving objects detection method based on improved SURF algorithm

In this paper, we present a new moving objects detection method in dynamic scenes to meet the real-time requirements. In the method, we propose an improved SURF algorithm for feature extraction. The SURF is improved via limiting the number of detected feature points, and adopting a fast method to reduce the repeated calculation when calculating the feature point´s dominant orientation. This algorithm improves the speed and precision of the original SURF. To reduce the computation complexity in the global-search method, an improved matching method is proposed. This improved method reduces the matching time and improves the precision of matching. Experimental results demonstrate that our proposed moving objects detection method is able to successfully detect the moving objects in dynamic scenes. It not only has higher accuracy and robustness, but also has a good advantage of time compared with the existing moving objects detection methods based on SIFT and SURF.