A MOVING OBJECT DETECTION METHOD BASED ON THE OTSU METHOD

Moving object detection is to recognize the physical movements of the objects in a video or image sequence. It is a preprocessing in our pedestrian re-identification task. We proposed a discrete Fourier transform-based moving object detection method in our former research. However, it depended on an experimental threshold and remained a ghost problem. This paper proposed an Otsu method-based moving object detection method. Specifically, it compared the Otsu method with the entropic method, minimum error method, and moment-preserving method, and the experimental result showed the Ostu method performed the best, and it could determine the threshold automatically and deal with the ghost problem. Moreover, the average nearest neighbor index was used to handle the misclassification problem of no foreground images.