Abandoned Object Detection Based On Single-View Video Analysis

Intelligent video surveillance based on video analysis, being an important research branch in the field of computer vision and pattern recognition, has become one of the key technologies which is paid much more attention in recent years. With the ever increasing requirements in public security, abandoned object detection has become one of the important tasks in intelligent video surveillance. Starting with the survey of the state-of-art, the thesis focuses on the study of abandoned object detection under single-view camera. The main contributions can be summarized as following:1. Abandoned object detection based on dual-foreground information fusionThis part regards abandoned object as one that does originally not belong to the background, but enters into the view of the camera, stops and remains static for enough time, and finally be absorbed into the new background. First, Codebook is used to model long-term and short-term background. By means of background subtraction, dual-foreground information can be extracted and fused. Finally, abandoned object can be detected on the base of morphology post-processing and adjacent neighbor connecting.2. Abandoned object detection based on single-layer foreground analysisIn this part, abandoned object is viewed as static foreground that doesn’t belong to the surveilled scene. First, this proposed algorithm does foreground extraction based on Codebook Model. Then, the static foreground region can be detected according to the given decision criteria. Therefore, abandoned object can be discriminated by means of region shrinking and out-growing.Experiments demonstrate the effectiveness in both above algorithms for abandoned object detection.