Uma abordagem otimizada para reconhecimento de logomarcas de veículos usando SURF

Abstract

Perform fast and accurate recognition of objects is still a challenging task in many applications. In this sense the techniques based on detection and description of image features and in correspondence (matching) has been set up as an important technological evolution in the development of object recognition systems. The most common procedure for object recognition systems is that, given some knowledge about the appearance of some objects, one or more images are examined to determine whether the object is present and the location thereof. Recognition of flat objects, such as images and logos is part of that context today and the approaches proposed for vehicle logos recognition problem. This problem has been one of the domain issues in automatic identification solutions. But realize the recognition of logos vehicles in quick way to images while maintaining the precision of the method employed is still a problem. On the other hand there is also the problem of enabling the application of the method speeded Up Robust Features (SURF) in recognition solutions objects. This method makes it possible to identify points of interest in images quickly and robustly. This paper proposes the development of an approach based on image processing and computer vision techniques to carry out reconnaissance vehicles of logos. The proposed method makes use of the SURF algorithm to perform recognition logos vehicles, combining algorithm with other techniques, such as the Canny method, Conci, Carvalho and Rauber, ROI, KNN and RANSAC to solve the problem, performing recognition the logos of vehicles from 45% to 100% of the images analyzed in a good runtime. The development of this research demonstrate a possible alternative choice for the logos of recognition problem of vehicles through efficient use of a model SURF method.