水果视觉分级检测线控制器设计  

Design of Controller for Fruit Visual Grading Inspection Line

作　　者：冉涌[1] 郎朗[1] 徐明灿[1] Ran Yong;Lang Lang;Xu Mingcan(School of Intelligent Manufacturing,Chongqing Three Gorges Vocational College,Chongqing 404155,China)

机构地区：[1]重庆三峡职业学院智能制造学院,重庆404155

出　　处：《安徽电子信息职业技术学院学报》2023年第1期1-5,共5页Journal of Anhui Vocational College of Electronics & Information Technology

基　　金：2020年重庆市教委科学研究项目“基于机器视觉的农产品智能检测与分级关键技术研究与应用”(KJQN202003501)。

摘　　要：对水果品质分级是提升农产品品牌价值的一个重要途径。传统的人工分拣或者纯机械分拣都无法满足大规模生产和精细化品质分级的要求。基于机器视觉的智能分级检测生产线是解决水果大规模快速分级的重要途径,为此,设计了一款检测控制器,选择边缘计算平台Jetson Nano作为核心计算单元,并部署YOLOv5模型用于视觉图像中目标检测,通过深度学习推理引擎TensorRT提升运行速度,对水果的果径、色泽、缺陷等几方面综合打分,量化分级准则。实验结果表明,该检测线对玫瑰香橙的分级速度远远高于人工分级,分级准确率也优于人工。Fruit quality grading is an important way to enhance the brand value of agricultural products.Traditional manual sorting or pure mechanical sorting cannot meet the requirements of mass production and fine quality grading.An intelligent grading production line based on machine vision is an important way to solve large-scale and rapid grading.Therefore,the controller in the inspection line is designed.The the edge computing platform Jetson Nano is selected as the core computing unit,and YOLOv5 model is deployed for target detection.TensorRT,as a deep learning reasoning engine,improves the running speed,comprehensively scores the fruit diameter,color,defect and other aspects,and quantifies the grading criteria.The experimental results show that the speed of orange grading with the detection line equipped with this controller is far higher than that of manual grading,and the grading accuracy is also better than that of manual grading.

关 键 词：果蔬分级 机器视觉 ARM 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]