Design and development of a graphic user interface for the SURF sensitivity integrated study depending on filming conditions

Currently, computer vision is used in many applied areas. It is becoming more and more important to solve problems of automatic processing and subsequent analysis of visual information. One of the computer vision central problems is the difficulty in extracting structured target information from photographic or video filming. Most of the existing techniques and approaches proved their applicability only in solving a limited number of problems. In addition, video analytics systems should remain operational under varying illumination conditions, when changing an object scale in the frame, with perspective, aspect and other geometric distortions, scene blurring, or exposed to various noises. The paper describes an algorithm designed to detect key points of the analyzed images using the deterministic Speeded Up Robust Features (SURF) method. Besides, part of the work is devoted to image preprocessing techniques. Using interactive software and tools of the Matlab environment, application was designed and developed to study stability of the method under consideration to alterations in filming conditions and evaluating the prospects of its application in cases, where it is impossible to match pixel-by-pixel the frames and the required reliability level. The application functionality makes it possible to ensure qualitative and quantitative comparative analysis of the efficiency in detecting key points by the SURF algorithm having distortions of different nature. Demonstrative visual part provides any user with a possibility to track the method critical indicators for images with different background and to evaluate results of the found points filtering with the stable RANdom SAmple Consensus (RANSAC) algorithm. Implemented application with image processing procedures and pattern recognition by comparing local features demonstrates invariance of the SURF algorithm to scaling, image rotation, changing the lighting conditions, blurring and noise in fairly wide ranges.