分水岭算法耦合感兴趣区域识别的智能机械手工件定位∗

摘要

为了解决基于图像处理技术的工件定位技术严重依赖机械定位与缺乏视觉功能而导致算法的自适应与定位精度不佳等不足，文章提出一种基于图像处理的智能机械手定位算法，并对其中的图像处理算法进行研究和实现。首先，根据经验值设定感兴趣区域，并实时计算像素灰度平均值，判断是否存在材料工件；然后将阈值控制嵌入分水岭算法中，避免过渡分割，凸显材料工件区域，计算出材料工件中心点坐标；并引入强度变化阈值和最小允许距离，基于Harris角点检测算法，刷选出最佳角点，定位材料工件的边角坐标，从而计算材料工件旋转角度。最后编程实现文中算法。实验测试表明：并与当前技术相比较，文中算法定位精度更高，可以应用于智能机械手的视觉定位系统。%In order to solve the current workpiece positioning technique based on image processing tech-niques are heavily dependent on mechanical positioning and algorithm of adaptive resulted from the lack of visual function and poor positioning accuracy is insufficient, this paper puts forward a kind of intelligent ro-bot localization algorithm based on image processing, and the image processing algorithm are studied and implemented. First of all, according to the standard value setting interest area, and real-time computing pix-el grayscale average, to determine whether there is material artifacts; Then embedding threshold control in the watershed algorithm, avoid transition segment, highlight material artifacts area, calculate the material workpiece center coordinates;And introduce threshold intensity change, and the minimum allowed distance based on Harris corner detection algorithm, brush chooses the best Angle of point, positioning material workpiece coordinates at the edges, thus the material workpiece rotation Angle is calculated. This algorithm is realized by programming. Experiment and test show that compared with current technology, the algorithm is higher location accuracy, can be applied to intelligent robot vision positioning system.