基于坐标注意力机制和残差网络的苹果外观品质检测
Apple Appearance Quality Detection Based on Coordinate Attention Mechanism and Residual Network
齐永兰 1李仁惠 1李学伟1
作者信息
- 1. 河南工学院智能工程学院,河南 新乡 453000
- 折叠
摘要
随着机器视觉技术的发展,利用卷积神经网络实现苹果品质分级已成为较优的应用技术.本研究以苹果外观品质特征为对象,提出了一种基于残差神经网络和坐标注意力机制的苹果品质检测方法.实验结果显示,引入坐标注意力机制后的ResNet18 网络模型平均准确率达到 91.4%,损失值为 0.1.该方法在各项性能上优于ResNet18、34、50 网络模型,能够有效实现苹果品质分级.
Abstract
With the development of machine vision technology,the use of convolutional neural network to achieve apple quality classification has become a better technology.In this paper,an apple quality detection method based on residual neural network and coordinate attention mechanism is proposed.The experimental results show that the average accuracy of ResNet18 network model after the introduction of coordinate attention mechanism reaches 91.4%,and the loss value is 0.1.This method has better performance than ResNet18,34 and 50 network models,and can effectively realize Apple quality classification.
关键词
坐标注意力机制/残差神经网络/机器视觉/水果分级Key words
coordinate attention mechanism/residual neural network/machine vision/fruit classification引用本文复制引用
出版年
2024
NETL
NSTL
万方数据