非结构农田环境机器视觉障碍物检测方法

Machine vision obstacle detection method in unstructured farmland environment

非结构农田环境准确感知与理解,是农业生产自动化向自主化进程中的关键技术支撑。针对农田作业机器人障碍物感知检测技术,提出3个田间避障路径规划的基本原则。分析阐述非结构复杂农田背景下基于视觉检测田间障碍物的优劣势,概述目前基于单目视觉和双目视觉检测障碍物的主要方法及各方法检测田间障碍物的局限性。探讨基于机器视觉的田间障碍物检测技术的实现方式、优缺点对比和局限性等问题,指出基于视觉的多传感器融合是农田作业机器人环境感知、认知进而自主作业的发展方向。

Accurate perception and understanding of unstructured farmland environment is the technical key in agricultural production automation to autonomy.Aiming at the obstacle perception and detection technology of farmland operation robots,three basic principles of obstacle avoidance path planning are proposed in this paper.This paper analyzes and expounds on the advantages and disadvantages of field obstacle detection based on vision under the background of unstructured,complex farmland and summarizes the main methods of obstacle detection based on monocular vision and binocular vision and the limitations of each method.This paper discusses the implementation,advantages,disadvantages,and limitations of field obstacle detection technology based on machine vision and points out that vision-based multi-sensor fusion is the development direction of environmental perception,cognition,and autonomous operation of farmland operation robots.