Abstract: Because the traditional image recognition algorithms have a large time-consuming and poor recognition effect in a complex environment, this paper proposes a track detection method for the coal mine underground environment. For the uneven light intensity in the underground environment, a grey level stretch and edge detection algorithm is designed for the complex underground environment to improve the effectiveness of the track detection. It proposes a priority-based search algorithm for inner edge of the track. The next image is detecting tracks in the Region of Interest(ROI) which is built by the detection result of the last image. In this way, the calculated amount is reduced and timeliness is improved. Field experimental results show that this method can detect the locomotive tracks effectively and has an obvious reduction of time-consuming than the traditional methods.

Key words: track detection, equidistant division, gray level stretch, edge extraction, inside edge search, Region of Interest(ROI)

摘要： 针对传统图像识别算法耗时大、对复杂环境识别效果差等缺点，提出一种针对煤矿井下环境的轨道检测方法。根据井下光线亮度不均匀的特点，设计井下复杂环境下的灰度拉伸与边缘提取算法，提高轨道检测的有效性。给出基于优先级的轨道内侧边缘搜索算法。后帧图像在基于前帧图像检测结果建立的感兴趣区中进行轨道检测，可降低计算量，提高实时性。现场实验结果证明，该方法能有效检测出机车轨道，且相比传统方法耗时明显减小。

关键词: 轨道检测, 等距离分割, 灰度拉伸, 边缘提取, 内侧边缘搜索, 感兴趣区

CLC Number: 

• TP391