Design And Implementation Of Image Scene Classification System Based On Hadoop

Image scene classification plays a very important role in the fields of computer vision,pattern recognition and machine learning.Image scene classification is widely used in many fields,such as target recognition and behavior detection.However,for a single object instance,the accuracy of image classification is usually reduced due to different illumination conditions,shooting angle and image acquisition distance.At the same time,the non-rigid deformation of the object itself and the partial occlusion of other objects will also lead to great changes in the performance characteristics of the object instance.At the same time,with the rapid development of multimedia and Internet technology,the scale of data generated by these technologies continues to expand.Most of the data is presented in the form of images and videos.Fast and accurate image classification has become one of the hot spots in the field of computer vision.Hadoop is one of the mainstream distributed computing platforms,which has a great advantage in dealing with massive data.Therefore,the powerful computing and storage functions on Hadoop platform are used to improve the efficiency of image classification.The content of this thesis mainly includes the following parts:(1)This thesis briefly introduces the architecture of Hadoop platform and the basic methods of image classification,with emphasis on the SIFT algorithm of low-level feature extraction and the basic theoretical knowledge of BoW model of middle-level feature modeling.(2)The problems caused by SIFT algorithm and BoW model are thoroughly analyzed.In this thesis,some optimization algorithms are proposed from two aspects: low-level feature extraction of SIFT algorithm and middle-level feature modeling of BoW algorithm.These optimization algorithms are used for image scene classification to improve the accuracy of image scene classification.(3)This thesis uses the improved classification algorithm to classify images on Hadoop platform.Experiments show that using the improved correlation classification algorithm on Hadoop platform not only improves the accuracy of classification,but also improves the speed of classification.(4)The image scene classification algorithm based on Hadoop is summarized and the prospect of future image scene classification is given.