Moving object detection method using background Gaussian kernel density estimation

Moving object detection is one of the focuses of the study on computer vision, image comprehension and signal processing. After studying and comparing the existent algorithms, an improved method was proposed to detect moving object, which uses the gray information of image to estimate the Gaussian kernel density, and it was easily realized in real engineering. In the algorithm, the background's multi-mode would be preserved, and area threshold was introduced to eliminate the big noises and judge a sudden change in the background. If the big noises wasn't eliminated or the sudden change happened in the background, the background samples would be reupdated to adapt its change. Experimental results show the proposed method is valid in treating the recurrent moving disturbance and sudden change in the background, eliminating the false alarm, and reducing the ratio of mistaking in the complex background, such as camera dithering, branch swing, raining day etc., the moving object is detected accurately.