Application Reaserch Of Mobile Terminal Based On Tensorflow Image Analysis

In recent years,with the maturity of artificial intelligence technology and related algorithms,related Applications in various industries have also increased.In particular,algorithms that use neural networks for image classification can help people improve their work efficiency and facilitate their daily lives,becoming the mainstream algorithm in today's smart field.Nowadays,in the industry of inspection,the traditional papermaking work has been replaced by intelligent inspection,but there are also some problems in the intelligent inspection system.For example,the images taken during the inspection often have insufficient details.If the brightness is not up to standard or the shooting content does not meet the requirements,this will result in the subsequent task review.Firstly,this paper designs a patrol system architecture for image quality prediction on the mobile end,and proposes a design idea of running the image classification model on the mobile end.The inspection system is divided into two main parts,as well as the mobile end and the background.The requirements and design goals of each part are analyzed separately,and the specific design scheme is given.Then,the focus of this paper is to prepare sufficient training salmples for the classification model.The pre-judgment of the quality of the captured image includes the definition of sharpness,brightness,and shooting content.Among them,judging the ambiguity of the image is more difficult to construct the sample than the other two indicators.Different requirements for the clarity of the captured image are proposed for different inspection tasks.So we need to build several training samples with different ambiguities.In this paper,the algorithm of Gaussian blur is optimized and improved.It is applied to the production samples,and the training samples suitable for judging image ambiguity in the inspection industry are obtained.Finally,the training of the model is completed and transplanted to the mobile terminal for functional development.This includes optimization of classification models for mobile storage and computing capabilities,as well as development of functional modules such as mobile location and database storage.Finally,the whole system was built and the accuracy of the model and the availability of the platform were verified.