基于优化蚁群算法的钢轨轮廓识别

摘要

Aiming at the existing problems of traditional ant colony algorithm in rail image recognition,the ant colony algorithm is optimized in four aspects.For the optimization of initialization process,the nonlinear iterative equation of one-dimensional Logistic chaotic mapping sequence is adopted to make the initialization distribution of ant colony more uniform,so that a large number of independent operations are avoided.For the optimization of search process,random search strategy is used at the beginning of ant colony search.The threshold is automatically set according to the gray gradient value of the image.The pixels of rail edge in the image are determined preliminarily,and then a region search model is set up to search and depict the rail edge accurately.For the optimization of search step length,in the early stage of the search,large step random search strategy is used to recognize the pixels of rail edge.Then small step region search strategy is used to recognize more accurately the pixels of rail edge so as to realize the accurate recognition of rail profile as well as reduce the search time and the convergence time of the algorithrr.For the optimization of pheromone update strategy,to prevent falling into local optimum,the pheromone is updated according to the maximum and minimum pheromone concentration of the pheromone which is set automatically after each search.A contrast test of track profile recognition for rail images acquired on straight and curve lines is conducted by Canny edge detection operator,traditional and optimization algorithm respectively.Results show that the proposed optimization algorithm has better robustness and higher recognition efficiency.%针对传统蚁群算法在钢轨图像识别中存在的问题,对蚁群算法进行4个方面的优化.初始化过程优化:采用一维Logistic混沌映射序列非线性迭代方程,使蚁群的初始化分布更加均匀,以避免大量的无关运算;搜索过程优化:在蚁群的搜索初期采用随机搜索策略,根据图像灰度梯度值自动设置阈值,初步确定图像中钢轨边缘的像素点,然后建立区域搜索模型,以进行钢轨边缘的精确搜索和描绘;搜索步长优化:在搜索初期,采用大步长随机搜索策略识别钢轨边缘的像素点,然后利用小步长区域搜索策略对钢轨边缘像素点做更精确地识别,从而实现钢轨轮廓的精确识别,并减少了搜索时间和算法的收敛时间;信息素更新策略优化:每完成1次搜索,根据自动设置的信息素最大、最小浓度值更新信息素,防止陷入局部最优.对实际采集到的直线和曲线线路上的钢轨图像分别用Canny边缘检测算子、传统算法和优化算法进行钢轨轮廓识别的对比试验,结果表明:优化算法具有更好的健壮性和识别效率.