基于机器视觉的柑橘果形测控系统设计  

Design of Citrus Fruit Shape Measurement and Control System Based on Machine Vision

作　　者：王志宇 文韬[1] 李浪 聂齐毅 刘豪 龚中良[1] Wang Zhiyu;Wen Tao;Li Lang;Nie Qiyi;Liu Hao;Gong Zhongliang(School of Mechanical and Electrical Engineering,Central South University of Forestry and Technology,Changsha 410004,China)

机构地区：[1]中南林业科技大学机电工程学院,长沙410004

出　　处：《农机化研究》2024年第1期120-125,共6页Journal of Agricultural Mechanization Research

基　　金：湖南省自然科学基金项目(2020JJ4142);湖南省林业杰青培养科研项目(XLK2021-7);湖南省教育厅科学研究重点项目(20A515)。

摘　　要：针对柑橘果形特征中圆度和果径检测精度低、姿态定位时间长的问题,设计了一种嵌入式快速检测与控制系统。系统以STM32单片机为系统控制核心,测量了单目相机图像的半径误差和形状误差,并采用高斯滤波、数字图像形态学、Hu矩和Canny算法,对动态的柑橘图像进行姿态识别与柑橘果形检测。检测结果表明:视觉检测系统半径误差控制在1.8%以内,形状误差为2.71%~3.69%;在5个/s柑橘的检测速度下,柑橘圆度和果径的在线分级正确率分别为81%和91.92%。本研究结合机器视觉无损检测技术,实现了动态下柑橘圆度和果径特征的综合检测与分级。Aiming at the problems of low detection accuracy of roundness and fruit diameter and long attitude positioning time in citrus fruit shape features,an embedded fast detection and control system was designed.The system used STM32 micro-controller as the system control core,and measured the radius error and shape error of the monocular camera image.The dynamic citrus images were processed in real time using Gaussian filtering,digital image morphology,Hu moments and Canny algorithm.Experiments shows that the radius error of the visual detection system is controlled within 1.8%,and the shape error is 2.71%-3.69%.Under the conveying speed of 5 citrus per second,the online grading accuracy of citrus roundness and fruit diameter are 81%and 91.92%,respectively.This study combined the machine vision non-destructive testing technology to realize the comprehensive detection and classification of citrus roundness and fruit diameter characteristics under dynamic conditions.

关 键 词：机器视觉 STM32 HU矩 圆度 果径 柑橘 

分 类 号：S126[农业科学—农业基础科学] TP271.4[自动化与计算机技术—检测技术与自动化装置]