Android应用开发病虫害识别,基于Android系统的储粮害虫图像识别软件设计与实现...

Design and Implementation of Android Software for Image Recognition of Stored Grain insects

Zhao Binyu

1

赵彬宇(1993-)，男，硕士研究生在读，主要研究方向：测控技术与智能系统

Zhou Huiling

1

周慧玲(1965-)，女，教授，主要研究方向：嵌入式系统与测控网络

1、Automation schoool, Beijing University of Posts and Telecommunications, Beijing 100876

Abstract：This paper designed and implemented a stored grain insect image recognition mobile phone software based on the Android system. The Resnet model with less network parameters was used to train a stored grain insect image recognition algorithm. The algorithm can identify 10 kinds of stored grain insects in 6 categories with an average accuracy of 0.9. In this paper, the trained algorithm is exported to pb file, and the file was transplanted to Android mobile phone. Then TensorFlow Lite was used to call the algorithm model, which realized the identification of the stored grain insect image on the Android mobile phone. The results show that it is feasible to identify the species of stored grain pests locally on the mobile phone. Compared with the online identification method, the local identification on the mobile phone can be protected from network conditions and server performance. This Android software could be an assistant for grain storage personnel and the majority of grain storage farmers.