Dynamic Background Based Moving Object Detection And Recogition

In the field of computer vision, video processing is the most basic operation.Inspecting and tracing the moving object in video have great application and research values. In today’s life, video processing technology has been applied in various fields,including the daily monitoring, military transportation, intelligent robot, etc. The computer vision method has great development space in today’s complicated traffic environment.Advanced Driving Assistance System is put forward as the important security assurance of people. In the system, the vehicle collision detecting system is an important component, which has two methods: one is the radar-based method, while the other is the vision-based method. Compared with the radar method, the vision method is more likely to be widely used because of its low price, large amount of information, and long sensing distance. Based on theories of Computer Vision and Machine Learning, the paper put forward moving object detecting and identifying methods in the dynamic background through the vision-based sensor, and applied it into ADAS to realize the vehicle collision detection. The major contributions of this paper are as follows:(1)Based on the study of basic theories of motion analysis in computer vision(background subtraction, modeling method and optical flow), and feature detection and basic learning methods of machine learning, the paper proposed the collision detecting method, different from the independent feature detection method commonly used in today, for the driving assistance system by modeling based on the process of hypothetical generation and hypothetical verification. Through experiments, it is proved that the system is practical with the steady illumination.(2)After the detailed study of literatures regarding collision time detection and short range communication technology, the collision time detecting method was proposed to be used in the driving assistance system and the short range communication technology was proposed to be used for communication. Different from traditional methods that only detect the vehicle, the system is improved by these two methods. A complete collision detecting system is proposed in the paper.