带式输送机托辊红外图像分割与定位算法

摘要

Key parts on Belt conveyor are lack of effective monitoring at present.The traditional monitoring methods such as visual inspection and temperature measurement have shortcoming with difficulty to solve the huge workload,blind spots and other issues.This paper proposes a kind of automatic segmentation and location algorithm for conveyor belt' s roller infrared image based on connected component.Conveyor belt infrared image were captured by inspection robot along the inspection track,and infrared images were processed using image signal processing technology.Using horizontal and vertical projection to extract roller area,we can reduce influence of bracket,conveyor belt as well as background on the follow-up image processing;the false edge and retain true edge were filtered based on axial ratio and area of extracted roller image edge' s connected components;edge crack was connected using closed operation;executed boundary tracking to obtain roller closed contour,and carried out seed filling to realize automatic segmentation of roller.Finally,according to pixel coordinates of closed contour,roller location can be realized in original infrared image.Experimental results show that this method can realize automatic segmentation and location of roller quickly and accurately,thus establishing a foundation for roller's running status monitoring.%由于煤矿带式输送机关键部件缺乏有效监测,而传统目检、温度监测方法存在工作量大、盲点多等问题,文中提出一种基于连通分量的带式输送机托辊红外图像自动分割与定位算法,对巡检机器人沿巡检轨道采集的带式输送机红外图像进行处理,利用垂直和水平投影截取托辊所在区域,减少支架、输送带以及背景对后续图像处理的影响;采用基于连通分量的长短轴比和面积信息对图像边缘进行过滤,消除对上述截取图像进行边缘检测形成的伪边缘,保留托辊的真实边缘;利用形态学闭运算连接托辊边缘缝隙,通过边界跟踪获得托辊闭合轮廓并进行种子区域填充,实现托辊自动分割;最后根据所得托辊二值图像闭合轮廓,基于轮廓像素点遍历在原红外图像完成托辊的自动定位.实验表明,本方法可快速实现托辊的自动分割和定位,为带式输送机托辊的运行状态监测奠定了基础.