Realization And Application Of Image Classification Using OpenCV

With the rapid development of internet technology,widely use of high-resolution cameras on smart phones,and the universal coverage of communication network,our lives are accompanied with a large number of images.However,it is difficult for us to search for an image with nominated content over the vast of images.Therefore,we need an application of classifying images in mobile terminals.In this thesis,we design and implement a model that is used to automatically classify images.First,we study the background of image classification and decide which technology we will use in this paper.Second,we research the technology of face detection and highlight the theory of the Haar classifier.At the same time,we introduce how to train a Haar classifier.Then,we expound the background of scene classification.In our research,we mainly use the HOG feature and the theory of SVM to implement the scene classification.From our experimental procedure and results,we conclude that classification produces the best results for street and city scenes while not so well in ocean images.In the last part of this thesis,we describe in details how to implement image classification in Android systems.First of all,we present the development of the Android system and the system framework.Then,the tool of NDK provided by Android system is described in details.It can not only increase the speed of image classification but also avoid decompile.Simultaneously,the interface of JNI offered by NDK could make us to use the Java and C/C++ to execute together.It can also generate the dynamic library.so file.NDK is able to make the dynamic library file and the code of the Java to an installation package of Android system.From the effect of our result,this application performs relatively well.