提出一种改进的基于激光投射成像法的管道内壁障碍物检测算法。通过激光特征分析与比较法结合提取激光圆环轨迹，利用 Otsu算法滤波，最小二乘法确定圆环参数，最后通过离线计算，在线查表方式获得障碍物高度。算法对管道图像亮度无特殊要求，克服了原系统观察与检测不能实时同步的问题。 MATLAB运算结果表明：算法的准确性，实时性和鲁棒性相对原设计有了很大的提升。每帧平均运算时间小于 0.5 s；计算的障碍物高度误差范围± 2 mm；算法适用于绝大部分场景的管道图片。

An improved method of laser projection photograph to judge the situation of pipeline in service is put forward. The algorithm detects the height of protuberance (which is also called obstruction) through the procedure of track detection which is finished by the combination of laser feature analysis and the comparison method. By Otsu algorithm filtering, parameters of circle are got with least square method. Finally, obstruction height is obtained through obstruction-height table which is calculated off-line. There is no strict light requirement to the pipeline image avoiding the original system’s observation and tests are not synchronous. Three advantages are demonstrated by MATLAB experiments which are obviously better than the original design: higher accuracy, better real-time performance and nice robustness. The average operation time of each frame is less than 0.5 seconds, and the range of height error is 2 mm. Images of almost all situations can be processed through the algorithm.