| It has become an urgent request to realize the intelligent,real-time and informationization of the over-limit detection of goods vehicles.Based on this research background,this paper uses binocular vision technology as the research direction,and improves the algorithm with stereo matching algorithm.Stereo matching is one of the main problems in binocular stereo vision research,and also a key step to realize the three-dimensional reconstruction of the target.SIFT algorithm is considered to be one of the most common and robust methods because it overcomes the influence of environmental factors to a great extent.However,SIFT algorithm has the problems of less matching points and long computation time,which affects the real-time implementation of 3D reconstruction and detection.Therefore,this paper makes an improvement on the above two points.First of all,we bring an improved plan to overcome the problem of SIFT to obtain less matching points.Using SIFT algorithm to generate the initial matching points at first,then use RANSAC algorithm to remove the error matching point pairs and generate the second matching point pairs,finally using the second matching point pairs as the seed points for regional growth by region growing theory.Then,to solve the problem of taking a long time,we put forward the method to implement the hardware platform of the algorithm,using FPGA has the advantages of parallelism,hardware reuse and design flexibility,parallel analysis and modular design based on FPGA hardware platform for algorithm.Simulate the improved algorithm in the MATLAB software,the experimental results show that the improved algorithm is matching a substantial increase in thenumber of points and verify that the improved algorithm has high matching accuracy than traditional algorithm.To verify the improved algorithm based on FPGA platform in Quartus II +ModelSim environment,according to the output results in MATLAB reduction analysis,verify the feasibility of the improved algorithm on FPGA. |
PDF Full Text Request