Research On The Technology Of The Wheel Set Image Online Processing Algorithm Based On Parallel Programming

Rail transit is one of the main domestic transportation, ensure its safe and efficient is verynecessary. Wheel set is the key moving part of rail transit, online detection of wheel set wearparameters is of great significance to the Rail transit. In recent years, the train speed is increasing,the use of artificial measurement gradually can not meet the requirements of the vehiclemaintenance. In order to be able to know the situation of wheel set when moving on time, improvethe efficiency and precision of detection, the automatic online detection is widely investigated.Wheel set wear online detection method based on optoelectronic image detection has advantagessuch as non-contact, fast detection speed, higher precision.During the process of wheel set wear online detection based on optoelectronic image, wheelset image acquired by CCD, have to through the technology of digital image processing to gain thewheel set profile curve, and then get the parameters of wheel set wear. Wheel set image realize theimage processing such as segmentation and related parameters calculation by the computer software,the research on the architecture of digital image processing and algorithm affects treatment effect ofwheel set image.The paper were the research of wheel set image online processing algorithm optimizationtechnology based on the parallel programming, according to the wheel set image acquired by wheelset wear detection, used the soft development exert the advantage of processors, used the paralleloptimization technology to reduce the time of wheel set image processing, optimized the algorithm,improved the accuracy of wheel set image processing. The main contents were as follows.(1) Studied on the characteristics of wheel set image acquired by the online detection of wheelset wear based on the optoelectronic image, make a classification of wheel set image according tothe characteristics. Studied the three kinds of parallel programming technology, respectively wereWindows multi-thread programming, OpenMP technology based on CPU parallel programming,and CUDA technology based on GPU parallel programming.(2) Analyzed the related image processing algorithms of wheel set, image classificationsegmentation algorithm was proposed. Used the sub-pixel technology for the pre-processing ofwheel set image, gained the pixel increased image, then made later operation for the image. Aimedat three different multi-thread technology, designed parallel optimization scheme.(3) Designed the experiment schemes of software, and designed the experiment of wheel setimage processing; through the experiment, analyzed the advantage of sub-pixel technology by qualitative and quantitative. Used Intel VTune Amplifier performance analysis tool to analyze thehot points and bottlenecks of the image processing procedures; designed the parallel optimizationexperiments.Finally, accorded to the experiment of wheel set image processing, the optimization contrastexperiment of sub-pixel image processing, the experiment of parallel optimization of wheel setimage, analyzed the experiment results by qualitative and quantitative.