Study And Realization Of Intelligent Vision Monitor For The City Road

In order to meet the requirement of city's traffic intelligent control, this thesis will present a study on how to realize the road traffic intelligent control of city using computer vision.First, this thesis will study how to recognize the double full line. After point out the difficulties of recognizing the double full line in the road's image, we put forward a distorted Sobel operator, based on its feature in the image. Then this operator will be carried on template operation to the road's image. A texture image will be created in the double full line region. We can determine the possible region of the line by using texture segmentation technology. Then this region in the image is binary, mathematical morphology operation, totaling several images for optimizing the binary image result. At last, with fitting two straight lines, the double solid line in the image is got.It always needs to segment the moving vehicles and other objects from the video in order to monitor road of city. Then, the thesis will study how to segment the moving object from image sequences. Before our plan is showed, we introduce other people's methods in segmenting moving object. Our methods need acquire three continuous frames. Each two adjacent frames subtract to get difference image. A new binary image way based on grid will put forward. After the difference image is binary, we use the following ways to analysis the binary image. That is or operation, Markov random field analysis, region operation, active contour model constricting original dot position. Then accurate object's contour will be found.In the last chapter, the thesis haves a study on how to comprehend the object's moving trend and track objects in image sequences. We can get results of the moving object, based on the video segmentation. Our plan uses the Horn-Schunck algorithm to get optical flow in the object contour. Then the optical flow will be applied to analysis object moving trend, and guidance moving object's tracking questions. The experiments show all our works get well results.