The Research And Implementation Of Multi-Target Detection Under Complex Scene In Video Surveillance System

In recent years, the multi-target detection task under complex scene in video surveillance system becomes a hot research direction in the field of computer vision. Because of the influence of various factors in complex scene, there will be undesirable problems including incomplete detection, over detection and missed detection in the object detection processes. This phenomenon affects its applications in computer vision and related fields. The undesirable situation will also happen when this detection technology is used in the multi-target detection system. This paper describes in detail the design and implementation of multi-target detection system based on Ncut(Normalized Cut) segmentation method and Part-based Model method of object detection.This paper proposes an improved method based on Part-based Model. First, we segment images based on Ncut algorithm and detect objects based on Part-based Model. Then, we dilate and mark the segmented images, adjust the score and threshold. Finally, we use segmentation for detection technology to realize the multi-target detection system, solving the undesirable problems to a certain extent and improving the detection accuracy. Experimental results show that in traffic scenes, our proposed method can detect foreground multi-target correctly and get an ideal effect.