基于PaddleX的计量器具智能识别及信息检索系统研究  

Research on Intelligent Recognition and Information Retrieval System for Measuring Instruments Based on PaddleX

作　　者：顾方 刘鹏 Gu Fang;Liu Peng(Guangzhou Institute of Measurement and Testing Technology,Guangzhou Guangdong 510633)

机构地区：[1]广州计量检测技术研究院,广东广州510633

出　　处：《中国仪器仪表》2025年第4期36-40,共5页China Instrumentation

基　　金：广州市NQI-质量安全科技协同创新中心(2023B04J0407)。

摘　　要：为解决计量器具送检人工依赖性强、易出错等问题,本文用深度学习方法对典型的177类计量器具图像分类识别,共采集了计量器具图像7080张,按7∶2∶1的比例随机划分训练集、验证集和测试集,选用PaddleX平台的ResNet50训练出图像深度学习分类模型,并基于该模型设计出一个计量器具智能识别检索系统装置,其中模型Top1准确率99.86%,Top5准确率100%。经实物测试结果表明,该系统能够对计量器具快速识别并精准检索,达到预期目标。To solve the problems of strong manual dependence and error prone in the submission of measuring instruments for inspection,deep learning methods are used to classify and recognize images of typical 177 types of measuring instruments.A total of 7080 images of measuring instruments were collected and randomly divided into training,validation,and testing sets in a 7:2:1 ratio.We trained an image deep learning classification model using the ResNet50 on the PaddleX platform,and designed an intelligent recognition and retrieval system for measuring instruments based on this model.The Top1 accuracy of the model was 99.86%,and the Top5 accuracy was 100%.The physical test results show that the system can quickly identify and accurately retrieve measuring instruments,achieving the expected goals.

关 键 词：PaddleX 计量器具 智能识别 ResNet50 深度学习 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]