Target Detection In Video Image Serial And Its Application In Service Monitoring System

The intelligent video monitoring technology is an emerging research in computer vision domain and receives the attention as the front topic, which involves computer science, machine vision, image processing, pattern recognition, artificial intelligence and many other disciplines. It uses the methods in computer vision and sequence image analysis to process, analysis and recognize the image sequence automatically.The service monitoring system mainly depends upon the human to examine and recognize while watching the captured video. It cannot meet the requirements for carrying out security guard, detaining and guarding service because its simplicity. Therefore, we apply the intelligent video monitoring technology into the service monitoring system. This dissertation conducts a research, designed according to specific tasks, on some essential technology of intelligent video monitoring technology, and then tries to apply the intelligent video monitoring technology to the service monitoring system. The intelligent service monitoring system automatically examines the objects appearing in the scene and judge whether they are persons or not. Then certain warning measure would be taken by judging results. The new system improves the efficiency of performing duties.There are several contributions in the thesis.Firstly, the questions of updating the background are analyzed based on the Gaussian model through experiments, and moving objects are detected by background subtraction from the video. We also proposed a method based on C-means clustering to extract the objectives. Compared with the typical iteration method and Otsu method, experiments show that this method can obtain the ideal effect of segmentation similarly.Secondly, characteristic parameters of human are established in the target prospects images after segmentation, such as the ratio of height to width, the area ratio, and the ratio of circumference. We also rebuild the Objective Recognition Algorithm based on the BP neural network, and create a three-tier BP neural network. The algorithm can judge whether the detected object is people or other objectives after training on the samples we collect.Finally, through analyzing the requirements of the service monitoring system, we design the overall structure of the system, hardware and software. Experiments implement some parts of them.