基于深度学习的冰鲜红虾新鲜度智能识别  

Intelligent Recognition of Freshness of Chilled Red Shrimp Based on Deep Learning

作　　者：王可 张存喜 王瑞 WANG Ke;ZHANG Cunxi;WANG Rui(School of Marine Engineering Equipment,Zhejiang Ocean University,Zhoushan 316022,China)

机构地区：[1]浙江海洋大学海洋工程装备学院,舟山316022

出　　处：《自动化与仪表》2023年第11期84-87,114,共5页Automation & Instrumentation

基　　金：舟山市科学技术局浙江海洋大学科技重大专项项目(2021C21001);浙江省虚拟仿真实验教学项目(11184890421)。

摘　　要：水产品的安全日益受到人们的关注,针对传统的水产品新鲜度检测方法存在着破坏样本、耗时长、成本高、检测效率低等问题,冷链储运的发展急需一种快速、准确的虾类新鲜度检测技术。以Resnet50网络模型为基础,构建冰鲜红虾新鲜度检测模型,引入ECA注意力机制给予目标特征更高的权重以提高目标识别精度,并使用迁移学习以更好地进行训练。改进后的Resnet50网络模型识别准确率可达93.35%,检测速度能满足实际需求,可在复杂环境中全天候识别新鲜度等级。The safety of aquatic products has attracted increasing attention,and in view of the problems of traditional aquatic product freshness detection methods,such as damage to samples,high time-consuming and high cost,and low detection efficiency,the development of cold chain storage and transportation urgently needs a rapid and accurate shrimp freshness detection technology.Based on the Resnet50 network model,a freshness detection model of chilled red shrimp is constructed,and the ECA attention mechanism is introduced to give higher weight to target features to improve target recognition accuracy,and transfer learning is used to better train.The recognition accuracy of the improved Resnet50 network model can reach 93.35%,and the detection speed can meet the actual needs,and the freshness level can be identified around the clock in complex environments.

关 键 词：计算机视觉 Resnet50 轻量化模型 深度学习 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] S126[自动化与计算机技术—控制科学与工程]