The Design Of Railway Defect Image Detection System Based On Machine Vision

With the rapid development of national economy, railway transports play an increasingly important role, on April18,2007sixth national railway system, China railway opened "zhui feng era". Railway defect detection due to train speed is becoming more and more short of effective time, the rail flaw detection equipment requirements have become higher.For railway defect detection car performance requirements, study its mileage device, make its reasonable installation of photoelectric encoder, realizes the mileage wheel and photoelectric encoder coaxial rotation, photoelectric encoder to get mileage pulse information; through to the pulse frequency doubling, guide to the processing of information, realize the railway defect detection car mileage count and direction discrimination; based on the linear array camera and lens element selection structures, image acquisition system, the accuracy at the same time, and image acquisition is not affected by the car speed, the output pulse by single chip microcomputer control for linear CCD array camera trigger pulse interval, collect image information.Research railway defects image processing algorithms, the railway surface defects, first USES the weighted median filter, remove the noise of the image information, and then by splitting the rail track vertical projection image plane, finally using the improved area connecting rail flaw information extraction algorithm, and using the LVQ neural network to realize classification of defects. Defect inspection for fasteners, fastener improvement "cross localization" legal fastener target area, and solve transcendental knowledge inaccurate positioning of faults. Then to extract the area of the fasteners, fastener contour direction field identification algorithm is put forward, according to the fastener contour matching, gradient direction identification of defects.Prote199se to complete single-chip computer control hardware circuit design, this paper, using Matlab to track defects related algorithms, and validate the feasibility of the algorithm. Of field of rail defects image processing, continuous optimization algorithm, complete the correct recognition and matching of defect information. For radio and television encoder pulse counting, learning to program using Visual c++6.0to write, then burn-in MCU to implement functional requirements. Finally, complete the system platform, integration of each system, accurately detect the defect information.