基于深度学习的实验设备识别系统开发设计  

Development and Design of Experimental Equipment Recognition System Based on Deep Learning

作　　者：黄济川 杨雨秋 廖磊[1] HUANG Ji-chuan;YANG Yu-qiu;LIAO Lei(Sichuan Normal University,Chengdu 610101,China)

机构地区：[1]四川师范大学,四川成都610101

出　　处：《工业技术创新》2020年第3期1-6,共6页Industrial Technology Innovation

基　　金：“互联网+”孵化项目(项目编号:S201910636331)

摘　　要：为了让使用者科学规范使用实验设备、教学者提高实验室教学效率,提出了基于深度学习的实验设备识别系统,系统搭载Android平台环境,使用者通过拍照或本地相册选取实验设备图像。Android客户端采集图像并裁剪,使用TCP/IP网络传输协议将图像发送至服务器端;服务器端使用残差网络和改进的YOLO网络模型对图像进行检测,并在数据库中查找图像特征值对应的实验设备;数据库将识别结果和设备使用方法、注意事项等反馈至Android客户端。测试表明,系统对实验设备的识别准确率可达99%以上。实验设备识别系统可为同行提供参考借鉴,提高教学效率和质量。In order to make the users use the experimental equipment scientifically and normatively, and to make the teachers improve the efficiency of laboratory teaching, an experimental equipment recognition system based on deep learning is proposed. The system is equipped with Android platform environment. Users can select images of experimental equipment by taking photos or local albums. The Android client collects and cuts the image, and sends the image to the server using the TCP/IP network transmission protocol;the server uses the residual network and the improved YOLO network model to detect the image, and finds the experimental equipment corresponding to the image characteristic value in the database;the database feeds back the recognition result, the equipment use method and precautions to the Android client. The test shows that the recognition accuracy of the system to the experimental equipment can reach more than 99%. The experimental equipment recognition system can provide references for peers and improve the teaching efficiency and quality.

关 键 词：深度学习 实验设备识别系统 Android客户端 服务器端 TCP/IP网络传输协议 残差网络 改进的YOLO网络模型 

分 类 号：TP311.1[自动化与计算机技术—计算机软件与理论]