挖掘机器人铲斗目标图像的改进分水岭分割

Improved watershed segmentation for bucket object images of excavator robots

挖掘机器人自主挖掘目标的实现过程中,利用视觉信息跟踪、识别挖掘机器人铲斗目标是关键技术之一.传统的阈值分割方法很难将铲斗从复杂环境中分割出来,提出了改进的分水岭铲斗图像目标分割方法,首先对铲斗目标图像进行模糊C-均值(C为预定的类别数目)聚类分割,再以初步分割后的图像为基础得到梯度图像,将梯度值与设定的阈值比较得到标记点,最后以标记点作为极小值点进行分水岭分割.实验表明分割效果得到了改善.

During the implementation process of autonomous object excavation via excavator robots, the bucket object secures a critical position in vision information tracking. By employing the traditional threshold segmentation method, it is difficult to segment the bucket from the complicated environment. As such, the improved watershed segmentation method is proposed. Firstly, the bucket object images are processed using the fuzzy C-means clustering. Then, the gradient images are obtained from the initiallysegmented images. Next, the mark points are attained through comparing the gradient values and thresholding values. Finally, the mark points are used as the minimum points for watershed segmentation. Therein, its is found from experimental results that the segmentation effect is improved.