Research And Implementation Of Image Forensic And Match Based On Mobile Device

This topic applies feature extraction algorithms to image matching, andcombines mobile Internet applications with the electronic evidencecollection, and develops Image Forensics System based on mobile deviceplatforms.The paper studies methods of image capture, locating and networkestablishing of mobile device. The application implements camera featuresin Windows Mobile smart phone based on DirecShow and managed toaccess the detailed location of the mobile device through GPS, by means ofthird-party Web Service, and create a stable and reliable GPRS connectionon the basis of automatic distinction between mobile service providers, andmade the picture transmission.Menwhile, the paper researches and compares a variety of main imagefeature extraction algorithms, and selects the one called SIFT with the bestcombined extraction results as the basis of image matching, and usingadaptive Gaussian kernel scale and feature redundancy method to improvethe SIFT algorithm, which has been improved to speed up the creation process of the Gaussian pyramid, and removes many-to-one match in thefeature point matching process. Testing shows that the improved SIFTfeature extraction algorithm success in enhancing the efficiency of featureextraction on the basis of maintaining the original matching accuracy. Theimproved SIFT feature extraction algorithm has been used in the matchingof image evidence, and the image feature extraction and matching is carriedout on the server side to make sure whether the image exists in the evidenceimage library. It also designs exception capture module which deals with theexception capture and treatment during image file transmission process. Atthe same time, the logging module managed to deal with the subsequentprocessing of the unfinished tasks, which improves the performance of thesystem’s fault tolerance.Finally, several experimental results show that, the mobile device as theimage acquisitor is capable of transmitting images in all kinds of networkconditions. The server’s extraction of image feature meets real-timerequirements, and the matching results are good, and this system has apractical value.