Study On Switch Machines Gap Detection System Based On Video Image Recognition Technology

Turnout is the key equipment in the process of railway and also an important link toensure the safety of train operation. With the development of high-speed railway in our country,stability and accuracy is particularly important for the turnout, not only does it relate to trainoperation safety, but also relates to the safety of the passengers, so the first task is to build asuitable and stable working system. In recent years, turnout gap detection has been the hotspot in the railway signal system, especially in the maintenance department. For signal mainte-nance management department, it is a key to break through the switch state bottlenecks anddissolve the turnout contradiction; For signal research department, it is also a problem thatcan’t be solved for a long time.According to the practical difficulties, we designe switch machine gap detection systembased on video image recognition technology, and focuse on the key technology of the system,including the rapid detection of target recognition pretreatment method, extract the image fea-tures and enhance image value, through the analysis to extract the target charcteristic curve,and the rail gap of background and target image segmentation, finally, analyse feature extrationand so on, then we make every effort to balance the effectiveness and quickness in the algorithmdesign.The thesis presents a high performance embedded intelligent visual inspection moduledesign scheme, the hardware design of the switch machine gap detection system adopts thesurface of the charge coupled device (CCD) camera and high-speed digital signal processing(DSP) processor as the core of hardware system, and also design the corresponding embeddedintelligent visual inspection software framework. Through the study of wavelet analysis theory,the thesis proposes a new fast algorithm, comparing with various algorithms for digital image,the pretreatment of the median filtering method can effectively reduce the noise interferenceimage, after the noise reduction of the image using the Sobel operator template convolutionoperation and extraction of target edge curve; gray image threshold method is used to finishedge image binarization operation. Finally, through using image deletion algorithm, pruningand refining a variety of small area algorithm of digital image to get images of single pixeledge curve, complete monitoring target image characteristic curve extraction, and then theoffset can be calculate conveniently.The system solves the problem of turnout maintance and promotes the development ofrailway system which becomes a key to break through the switch state bottlenecks, the theore-tical basis not only provides a reference for engineering design and further research, but also awide range of benefits for the social and economic. Finally, the main work of this thesis issummarized, and the research in the future is prospected.