A rapid texture-based moving object detection method
Yea‐Shuan Huang, Zhi-Hong Ou, Hsiang-Wen Hsieh, Hung-Hsiu Yu · Asian Control Conference · 2011
This paper presents a moving object detection method based on texture information extracted from images. Every image captured from the camera is converted into Local Binary Pattern (LBP). Salient feature points extracted from previous LBP are compared with those features found from the current LBP with a block matching approach so that the corresponding feature points from the successive image frames can be identified. If multiple correspondences exist between feature points, the motion vectors of each feature points are then calculated to determine the best corresponding features on the current LBP. Finally, with clustering of motion vectors, all the moving objects on image frames can be successfully detected and identified. Experimental results show that the average matching accuracy rate is 95.12%, and the average processing time for moving object detection is 46.2ms.