Detection Of Vehicle Tracking Method Based Car Video

In recent years, our family car ownership grow speedy,with the auto-relatedtechnology gradually matured,urbanization process develop more fast and people’slife quality of life greatly improved. At the same time,the urban road traffic bemore congested and traffic speed be faster, traffic conflicts is more serious and thetraffic accidents occur frequently that threaten on people’s lives and caused theloss of national economy property. Thus the automatic driving and safe driving hasincreasingly become a research hotspot, vehicle detection and tracking technologyin intelligent transportation system are increasingly concerned by domestic andforeign scholars.Video-based moving object recognition algorithm has been the focus ofcomputer vision research. Different researchers of countries gives a differentsolution on this issue from different angles. However, due to the complexity oftarget movement identification based on visual, there are many difficulties onreal-time identification and recognition accuracy,till now have not find a specificmethod that is suitable for all scenarios,there remains difficulties on moving targetidentification especially for rain, snow, fog and other unconventional weather.This paper first summarizes and analyzes the research status and the commonsense of video moving target identification, compare and analyze the detectionalgorithms,give a detailed description of feature extraction and time domaintracking method,give the comparerence of traditional detection and trackingalgorithms and the characteristics of optical flow. Then made the introduction ofbackground modeling based on Gaussian distribution,followed complete trajectorytracking by the use of hidden Markov models.Finally, we propose a moving object recognition method which suit fog andhaze days with the basis of a combination of domestic and foreign research results,this method can be in a bad light, severe fog and haze obscured days to completethe identification of moving targets, at the same time we make a lot of experimentsand analyze the experimental results.By analyze the experimental results,we canverify feature-based optical flow method and the moving target tracking algorithmto identify low-dimensional feature extraction accuracy in real-time and betterperformance. Meanwhile, the paper also still summarizes the problems, while thedirection of future research to make. In summary, This paper presents a video recognition method based onmoving targets that recognition accuracy under the premise of ensuring thecompletion of a moving target in bad weather to identify unconventional methodsto prove the usefulness of this method.