%0 Journal Article
%T 基于图像色相值突变特征的钢轨区域快速识别方法<br>Fast recognition method of rail region based on hue value mutation feature of image
%A 闵永智
%A 殷
%A 超
%A 党建武
%A 程天栋
%J 交通运输工程学报
%D 2016
%X 应用彩色图像中不同区域HSL色彩空间中色相值突变特征提取轨检图像中钢轨边界点,对多条不同等分线处钢轨边界点进行直线拟合以确定钢轨边缘,识别目标钢轨区域。分析了机器视觉轨检系统序列图像中轨枕、砟石、扣件与钢轨的分布特征及不同特征区域图像色相值的突变特征,研究了轨检图像不同等分数值下等分线处色相值突变点与钢轨边界点的对应关系,讨论了不同等分值对识别时间与识别失败率的影响。在不同光照条件下对识别方法与传统方法进行了对比分析。分析结果表明:当等分值为8时识别效果最优,识别失败率为5.0%,识别时间为4.65 ms; 在500～1 000、1 000～10 000、10 000～100 000 lx三个特征光照强度区间,识别方法在木枕与混凝土枕轨道中钢轨区域的平均最大识别时间分别为4.57、4.48 ms,比传统方法分别减少了44.4%、47.1%,识别时间标准差分别为0.15、0.12 ms,比传统方法分别降低了91.8%、93.6%,平均最大识别失败率分别为3.5%、3.3%,比传统方法分别降低了66.0%、76.9%,识别失败率标准差均为1.6%,比传统方法分别降低了68.9%、71.1%。可见,本文方法是一种机器视觉轨检系统中目标钢轨区域识别的有效方法。<br>The rail boundary points in track inspection image were extracted by detecting the hue value mutation features of different regions in the HSL color space of color images. The rail's edges were determined by the linear fitting of multiple rail boundary points of different bisectors, and then the target rail region was recognized. The distribution features of sleeper, ballast, fastener and rail and the hue value mutation features of different characteristic regions were analyzed in the captured sequence images of track inspection system using machine vision. The correspondences between the hue value mutation point and the rail boundary points of different bisectors under different numbers of equal parts were researched. The influences of different numbers of equal parts on recognition time and recognition failure rate were discussed. The recognition method was compared with the traditional method under different light conditions. Analysis result indicates that when the number of equal parts is 8, the optimal recognition effect is obtained, and the recognition failure rate is 5.0%, the recognition time is 4.65 ms. In the three intervals of characteristic light intensity, such as 500-1 000, 1 000-10 000, and 10 000-100 000 lx, the average maximum recognition times of recognition method in the rail regions of tracks with wood sleepers and concrete sleepers are 4.57 ms and 4.48 ms respectively, and are 44.4% and 47.1% less than the values of traditional method respectively. The standard deviations of recognition times are 0.15 ms and 0.12 ms respectively, and are 91.8% and 93.6% lower than the values of traditional method respectively. The average biggest recognition failure rates of proposed method are 3.5% and 3.3% respectively, and are 66.0% and 76.9% less than the values of traditional method respectively. The standard deviations of recognition failure rate are 1.6%, and are 68.9% and 71.1% lower than the values of traditional method respectively. So the proposed method is an effective recognition method of target rail
%K 轨道检测
%K 钢轨区域
%K 机器视觉
%K 区域识别
%K 色相值突变
%K 直线拟合
%K 识别时间
%K 识别失败率&lt
%K br&gt
%K track inspection
%K rail region
%K machine vision
%K region recognition
%K hue value mutation
%K linear fitting
%K recognition time
%K recognition failure rate
%U http://transport.chd.edu.cn/oa/DArticle.aspx?type=view&id=201601006