摘要
水果品质检测与识别是智慧农业的重要应用场景之一,涉及智能水果分拣、仓储管理优化等核心任务。鉴于手工设计特征与分类器的方法不能有效覆盖各种水果,提出一种基于机器视觉的端到端水果识别与品质检测系统。系统由C#开发的上位机图形化界面、STM32F429开发板、单目摄像头和用于水果品质检测的深度神经网络模型组成。经过为期一个月对苹果、香蕉和柑橘三类水果的跟踪记录,完成了水果新鲜度检测数据集的制作。实验结果表明,本系统在自建数据集的三种水果新鲜度的识别任务中Macro-Precision达到了97.83%,可以被有效地应用到诸如智慧水果仓储或智能水果分拣等场景,具有一定的现实应用意义。
Fruit quality detection and identification is one of the important application scenarios in smart agriculture,involving core tasks such as intelligent fruit sorting and optimized warehouse management.Given that the manual design of features and classifiers cannot effectively cover various fruits,an end-to-end fruit identification and quality detection system based on machine vision is proposed.The system consists of a graphical interface developed in C#for the upper computer,an STM32F429 development board,a monocular camera,and a deep neural network model for fruit quality detection.After tracking and recording for three types of fruits-apples,bananas,and citrus fruits for a period of one month,the dataset for fruit freshness detection was completed.Experimental results show that this system achieved a Macro-Precision of 97.83%in the identification tasks of the freshness of three types of fruits in the self-built dataset,and can be effectively applied to scenarios such as smart fruit storage or intelligent fruit sorting,possessing certain practical significance.
作者
马磊
管文范
MA Lei;GUAN Wenfan(Tongren Polytechnic College,Tongren,Guizhou,China 554300;Huawei Technologies Co.,Ltd.,Hangzhou,Zhejiang,China 310007)
出处
《湖南邮电职业技术学院学报》
2025年第2期31-36,共6页
Journal of Hunan Post and Telecommunication College
关键词
智慧农业
水果品质检测
深度神经网络
smart agriculture
fruit quality detection
deep neural network
