面向软面料自主抓取的机器人视觉检测与定位

The Visual Detection and Localization of Robots’ Grasps Automatically on Soft Fabric

摘要:

服装软面料容易发生毛刺、褶皱、折叠和模糊等现象,给服装机器人自主上下料系统的视觉边缘检测与定位的准确性带来了较大的挑战。为了克服上述问题,本文提出了一种基于双重滤波的自适应边缘定位方法,可用于引导服装机器人实现软面料的自主抓取操作。首先采用高斯滤波、形态学滤波对图像进行平滑处理,然后基于边缘梯度信息计算Canny算子自适应阈值完成边缘检测,最后对提取到的边缘特征坐标进行定位。实验结果显示：方法在服装软面料构成的真实数据集上测试时,表现出良好的有效性(准确率≥96%)和鲁棒性。该方法对服装机器人的自主上下料操作具有良好的应用价值,对智能服装工厂的构建具有积极意义。

Abstract:

The soft fabric is prone to burr, fold, enfold and blur, which poses a great challenge to the accuracy of visual edge detection and location of the autonomous loading and unloading system of garment robot. In order to overcome the above problems, it proposes an adaptive edge location method based on double filtering in this paper, which can be used to guide the garment robot to achieve grasping of soft fabric automatically. Firstly, the Gauss filtering and morphological filtering are used to smooth the image, then it calculates the canny adaptive threshold based on the edge gradient information and detect the edge, finally it locates the coordinates of edge feature extracted. The experiment results shows that the method is effective(accuracy greater than 96%)and robust tested on the real data sets composed of soft fabric. The method has good application value for the autonomous loading and unloading operation of the garment robot and has positive significance for the construction of the intelligent garment factory.