Design And Implementation Of Traffic Intersection Image Stitching System Based On ARM

Image stitching is widely used in computer vision,graphic remote sensing and virtual reality,and it is one of the hot research fields of image processing in recent years.With the development of electronic information technology,Intelligent Transportation has been paid more and more attention by the state in recent years.Today,illegal image acquisition at traffic intersections has the advantages of high definition and high exposure speed.However,when taking illegal images,because the camera has a high resolution lens,the perspective of the image taken by this camera will not be very wide,and it cannot obtain a wider range of vehicle surroundings.Therefore,we need to use image stitching technology to expand the vision of the intersection image while keeping the image clear.In this thesis,image registration and image fusion methods in image stitching technology are studied in depth.SURP-PROSAC image registration method and fast seamline location algorithm based on distance and value matrix are used.An embedded hardware platform based on ARM is designed to realize the image stitching system of traffic intersections.The main work includes the following:1.In view of the mismatches in image registration,this thesis presents a Speeded Up Robust Features(SURF)as a feature extraction method,and uses the Random Sample Consensus(RANSAC)improved algorithm PROSAC(Progressive Sample Consensus)to refine the feature points and optimize the feature extraction scheme for feature matching.Experiments show that in the image stitching process,the scheme effectively eliminates mismatched points and is faster than RANSAC,effectively reducing the consumption of hardware computing power.2.To solve the slow seamline positioning problem in image fusion,a fast seamline positioning algorithm based on distance and value matrix is proposed.Seamline estimation is one of the key steps in image stitching,which can eliminate faults in overlapping areas or the ghosting of moving objects.In this thesis,a fast seamline extraction method is presented by defining gray weighted distance and gradient difference regions.By introducing distance weighted matrix and value weighted matrix,the extraction speed and quality of seamlines are improved based on the values in the matrix,and the optimal seamlines are estimated to eliminate seamlines and ghosting.In the experiment,the image data collection made by the intersection of Weihe Road and Xijing Road in Xi’an and the intersection of Xijing Road and Shendian Du Street is used.Experiments show that the fast seamline positioning method has better speed and robustness than other commonly used algorithms,and achieves the expected results in seamline positioning and fusion.3.This paper implements the image stitching system of traffic intersection on the embedded ARM hardware platform.As an embedded processor,ARM has the advantages of small size,low cost and wide application range.This paper builds development environment based on ARM platform.The library was transplanted and compiled.Complete communication between different platforms by installing virtual machines;Designing PC interface and realizing image stitching system of traffic intersection.The test system can run completely and achieve the expected results.