摘要
为了提高采摘机器人自主导航和路径规划能力,提出了基于计算机视觉路径规划和RBF神经网络自适应逼近算法的导航方法。使用图像分割、平滑处理和边缘检测技术,根据图像像素灰度值确定了导航线的位置,利用逐行扫描的方法得到了导航离散点。路径规划和跟踪使用RBF神经网络逼近算法,通过逼近误差和权值控制路径跟踪的精度,系统响应的执行端使用液压伺服系统,提高了机器人自主导航的精度。以黄瓜采摘作为研究对象,在日光温室对机器人采摘作业进行了测试,通过测试得到了RBF神经网络的路径跟踪误差曲线。测试结果表明:机器人可以很好地逼近跟踪规划路径,其计算精度较高,跟踪效果较好。
In order to improve the ability of autonomous navigation and path planning of picking robot,a navigation method is proposed based on computer vision path planning and RBF neural network adaptive approximation algorithm. The use of image segmentation,smoothing and edge detection technology,the navigation line positions are determined according to the image pixel gray value using progressive scan method of navigation discrete points. The path planning and tracking using RBF neural network approximation algorithm,the accuracy of the system response is controlled by the accuracy of the error and weight control. Taking cucumber as the research object,it tested the robot picking operation in greenhouse,and obtained the path tracking error curve of RBF neural network. The test results show that the robot can get a good approximation of the path.
出处
《农机化研究》
北大核心
2016年第11期234-238,共5页
Journal of Agricultural Mechanization Research
基金
河北省自然科学基金项目(E2013203271)
