Image Based Intelligent Recognition For Environment Status Of High Speed Railway Lines

The technologies relatied to the intelligent recognition of the environment status in high speed railway, mainly including the real-time image processing and the image based intelligent recognition, were reaserched in this thesis. The intelligent recognizing system for high speed railway environment status based on these technologies was also built in this thesis. This system acquires images of railway line by using the camera installed on the comprehensive inspection train, then processes and analyses these images to realize the intelligent recognition of the environment status.The general design of the intelligent recognition system for the environment status of the high speed railway lines, including the component of the system, the functions of each part, and some key technologies, was firstly introduced in this thesis. The issues about camera selection and installation were also discussed. Then the aspects of real-time image processing and image based intelligent recognition were reasherched in deatail.In the aspect of real-time image processing, a FPGA based hardware processors, which highly accelerates the processing speed, was used in this system. To simplify the design of image processing hardwire and makes the hardware designed more flexible, a new hardwire implement technique of image processing algorithms based on image algebra was proposed in this thesis. This method was used to implement some algorithms used in system, including 2-D image filter, gray-scale morphology and edge detection, the simulation results and testing results were given in this thesis.In the aspect of the recognition, a method combines 2-D image recognition and 3-D stereo recognition was proposed in this thesis. Firstly,2-D image recognition was used to abstract the abnormal features of the railway line rapidly, and then the stereo vision was used to obtain the 3-D information of them, which confirm if these features are defects of the railway line or not. Therefore, this method has both high recognizing speed and high reliability. Fence completeness recognition algorithm and foreign object recognition algorithm were discussed in this thesis, the principles, features and results are given. The architecture and principle of stereo vision, calibration experiment and 3-D information obtain are introduced, and the recognition based on 3-D information is also discussed in this thesis.