农业机械学报2024，Vol.55Issue(z1)：9-17,9.DOI:10.6041/j.issn.1000-1298.2024.S1.002
基于YOLO v8-Seg的地栽草莓采摘机器人垄面视觉导航控制方法
Ridge Visual Navigation Control Method for Ground-planted Strawberry Picking Robots Based on YOLO v8-Seg Algorithm
摘要
Abstract
The unmanned operation of agricultural machinery is inseparable from autonomous navigation technology.With the development of sensors and the improvement of computer vision technology,the autonomous visual navigation operation of agricultural robots in greenhouses has gradually become possible.Research on the visual navigation control method for ridge-surface operation of strawberry picking robots planted in the field was conducted.It analyzed the agricultural techniques of field-grown strawberries and acquired the features of strawberry ridges based on the YOLO v8 instance segmentation algorithm.The Canny edge detection algorithm was employed to extract the edge information of the ridge surface.Two straight lines with slopes of 1 and-1 were used to traverse the ridge surface,and the intercept information was statistically obtained to acquire the upper and lower endpoints of the ridge surface.The center point coordinates of the upper and lower endpoints on the ridge surface were then obtained.By connecting the upper and lower center points of the ridge surface into a straight line,the corresponding navigation line of the ridge can be obtained.An image dataset of the ridge surface of field-grown strawberries in the greenhouse environment was collected.After testing,the extraction accuracy of the navigation path was 96%,and the algorithm took 30 ms.The algorithm was deployed to the strawberry picking robot with a four-wheel Ackerman steering chassis.Combined with the preview point tracking algorithm,a navigation test was carried out on the simulated strawberry ridge.After testing,the extraction accuracy of the navigation path was 94%,and the algorithm took 30 ms.When the driving speed was 0.2 m/s,the maximum lateral offset was 32.69 mm,the average value was 22.12 mm,and the root mean square error(RMSE)was 5.37 mm,meeting the requirements for autonomous navigation control of the strawberry picking robot on the ridge surface.This control method,in conjunction with the autonomous picking function of the picking robot,can enable the unmanned autonomous operation of the strawberry picking robot.关键词
地栽草莓/采摘机器人/实例分割/垄面视觉导航/预瞄点跟踪/YOLO v8-SegKey words
ground-planted strawberries/picking robot/instance segmentation/ridge visual navigation/preview point tracking/YOLO v8-Seg分类
信息技术与安全科学引用本文复制引用
应仇凯,程泓超,马锃宏,杜小强..基于YOLO v8-Seg的地栽草莓采摘机器人垄面视觉导航控制方法[J].农业机械学报,2024,55(z1):9-17,9.基金项目
浙江省"三农九方"科技协作计划项目(2024SNJF070-1) （2024SNJF070-1）
