采摘机器人分割与识别算法的研究现状

Research Status of Picking Robot Segmentation and Recognition Algorithms

针对人工采摘效率低、成本高且劳动力缺乏的问题,利用采摘机器人实现果实的自动化采摘日益成为研究的热点。视觉识别作为采摘机器人的关键技术之一,其发展对实现自动化采摘具有重要意义。采摘环境的复杂性为机器人正确识别目标带来了困难,使得目前的识别算法研究仍面临着较大的挑战。本文阐述了常见的几种目标分割识别算法,并将其归纳为基于特征、像素和深度学习的三类识别方法,综述了国内外学者在采摘机器人识别目标时应用到的分割和识别算法,并对目前存在的问题进行总结,同时对今后的研究趋势进行展望,以期为采摘机器人的研究设计提供参考。

In view of the problems of low efficiency,high cost and lack of labor force of manual picking,the automatic picking of fruits by picking robots has become a hot research topic.As one of the key technologies of picking robot,the development of visual recognition is of great significance to realize automatic picking.The complexity of picking environment makes it difficult for robot to correctly identify the target,so the research on the current recognition algorithm is still facing great challenges.This paper expounds several common kinds of target segmentation and recognition algorithms,and they were summarized as three kinds of recognition methods based on feature,pixels and deep learning approach.The scholars at home and abroad were reviewed in segmentation and recognition algorithms for picking robots and applied to identify the targets.The existing problems were also summerized,at the same time,the research trend in the future was pointed out,which was expected to provide references for the research and design of picking robot.