轨道表面图像处理算法研究  

Research on Algorithm of Track Surface Image Processing

作　　者：汪昡紫 孙宪坤[1] 高飞[1] 

机构地区：[1]上海工程技术大学电子电气工程学院,上海201620

出　　处：《计算机技术与发展》2015年第9期182-186,193,共6页Computer Technology and Development

基　　金：上海市教育重点创新项目(14zz156);上海工程技术大学"十二五"内涵建设项目(B-8932-13-0124)

摘　　要：轨道检测是我国快速发展的铁路运输和城市轨道交通运输的重要保障。为了更好地完成轨道检测工作,首先介绍了轨道检测发展的趋势和现有的轨道探伤标准,然后基于计算机视觉中的双目视觉原理,搭建了轨道图像采集系统的模型。针对此系统采集到的同一根轨道的角度不同的左右两幅图像,研究针对轨道表面图像的处理算法。对比分析基于区域的Otsu分割算法和基于改进蚁群优化算法的边缘检测算法的原理以及算法表现,实验结果表明,文中改进算法的结果更适合此应用。在此基础上,根据相似度函数将分割好的左右图像进行图像匹配,获取双目视觉的视差图,为后续的深度信息提取与图像的三维重建和全面测量打下了良好的基础。Track detection is the basic protection of the railway and the subway development in China. In order to finish the track detecting work better,firstly introduce the developing trends and up to date method used in the track detection. Then based on the binocular vision system in computer vision technology,build a image collecting system. Aiming at the two images collected of a same track from two dif-ferent angles,have done researches on track surface image process algorithm. Compare and analyze the Otsu algorithm and the modified ACO algorithm,the experimental result shows that the latter performs better in this application. Based on this,finish the image matching using similarity function and obtain the disparity map of these two images,which will definitely make a contribution to the following im-age deep information extraction and image 3D rebuilding,analysis and totally measuring of the track.

关 键 词：双目视觉 图像分割 OTSU算法 蚁群算法 图像匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]