林果采摘机器人视觉系统研究进展

Research Advance on Vision System of Forest Fruit Picking Robot

视觉系统是林果采摘机器人的重要组成部分之一,它的林果目标识别和定位能力直接决定了采摘机器人的采摘效率和质量。首先介绍了典型的林果采摘机器人及其视觉系统,其次系统地梳理分析了现有林果识别与定位的几种主要方法。林果识别方面包括传统的数字图像处理技术和机器学习的图像识别,深度学习的图像识别;三维定位方面包括单目与双目的目标定位,RGB-D深度相机的目标定位。最后针对现有视觉系统中存在的不足提出几点建议。

Vision system is one of the important components of the forest fruit picking robot, and its ability to identify and locate the target of fruit directly determines the picking efficiency and quality of the robot.Firstly, this paper introduces the typical forest fruit picking robot and its vision system with examples.Secondly, it systematically analyzes and sorts out several existing main methods of forest fruit recognition and positioning.Forest fruit recognition includes traditional digital image processing technology, machine learning image recognition and deep learning image recognition.Three-dimensional positioning includes monocular and binocular target positioning and RGB-D depth camera target positioning.Finally, some suggestions are put forward aiming at the existing deficiencies in the visual system.