Research And Implementation Of Automatic Test Optimization Method For Mobile Phone Pages Based On Image Matching

In recent years,with the development of science and technology,the popularity of the Internet,the speed of the network increased.There are more and more mobile application software and more and more complex functions.Software testing is the guarantee of normal use of mobile application software.Relying solely on manual completion requires too much manpower and material resources,especially because of different parameters such as mobile phone brand,model,resolution,color and screen proportion,which may cause mobile phone interface problems.At present,there are a lot of automated testing tools for mobile phone interface,but all of them are used to check the interface elements and can't really detect the problems such as text coding and image stretching.With the development of machine learning and computer vision technology,test developers have found that the application of image processing technology to automatic test of mobile phone interface,processing and detecting mobile phone interface as one picture after another can not only improve efficiency,but also accurately detect interface problems.The main results of this thesis are as follows:(1)A semantic segmentation model is proposed to process mobile phone interface images.At present,the automated testing framework based on image processing does not take image segmentation as the method of image processing in mobile application software interface.The first annotation data set of mobile phone interface images was made through screenshots of the mobile phone application interface,and various image enhancement processing was carried out on the data set.In this thesis,data is trained by the pre-training model of Deep Lab V3+,a semantic segmentation model.Because Deep Lab V3+ model adopts the mainstream encoder-decoder structure,backbone network(backbone)is used to extract features in the process of encoder,and the computing cost of extracting features of different backbones varies greatly.In this thesis,Mobile Net V2 is used as the backbone to extract the characteristics,which can ensure the precision and improve the speed.(2)An automatic testing framework for mobile phone interface based on image matching is proposed.The application of image processing technology is an attempt in the field of testing.The image matching technology used by the existing automated test platform is based on SIFT model,with low matching accuracy and lower matching degree under different resolutions.In this thesis,image matching was carried out based on SIFT enhanced version SURF(Speeded-Up Robust Features)and false matching feature points were removed by RANSAC(Random Sample Consensus)algorithm.On this basis,aiming at the commonality between the layout and functions of mobile application interface,the matching effect is optimized through the peak signal to noise ratio.With the development of science and technology,the popularity of the Internet.there are more and more mobile application software and complex functions.Software testing is the guarantee for the normal use of mobile applications.It is time-consuming and laborious to solely rely on manual testing,especially in the aspect of mobile interface testing.mobile phone have different parameters ?brands and models such as resolution,color and screen ratio.Therefore,when testing the compatibility of mobile phones,it is necessary to test the functions of mobile applications on different models as much as possible.Automated test has the characteristics of high efficiency and high reliability,and has gradually become the mainstream test method in the process of mobile application development.However,due to the continuous improvement of mobile development technology and the change of page rendering technology,the test script construction and maintenance cost of traditional automated test framework is high.(3)Mobile Net v2 was used to optimize the image feature extraction,and image blurring and multi-resolution images in the image enhancement technology were trained to extract more precise features and improve the matching degree between the benchmark image and the image to be tested.