Moving Small Object Detection Algorithm Based on Three-frame Difference and Improved Twice Image Segmentations in HSV Space

For the detection of moving small object, the traditional inter-frame difference method often leads to object hole and contour loss. In this paper, a moving small object detection algorithm based on three-frame difference and improved twice image segmentations in HSV space is proposed. Firstly, the three-frame difference method is used to extract the object region. Secondly, an improved twice image segmentations algorithm is proposed in HSV space. The results of three-frame difference method are used as seed points, and the H component is used for region growing in order to improve connectivity of object region. Finally, according to the characteristics of small object, an improved adaptive threshold segmentation algorithm is designed based on S component gradations in order to make the extracted object closer to the real object. Using the adaptive threshold as the segmentation criteria, the coarse segmented images are segmented again. The moving small object can be obtained. The experimental results show that compared with traditional GMM algorithm and three-frame difference method, the proposed algorithm can efficiently obtain the moving small object with higher accuracy and integrity. The detection rate can reach up to 99.58%, and the false acceptance rate can be as low as 0.06%.