Vehicle Logo Recognition Based On Video Analysis

This paper made a systematic analysis on vehicle recognition. It focused on the accurate positioning of the vehicle logo and the analysis of vehicle recognition. It proposed the theory methods to solve the relative questions in the above technology. It also verified the validity through the experiment. The main contents of the research in this paper:(1) The research of the technical methods about the logo positioning:logo positioning is a typical kind of target detection and positioning technology. The level of the target’s accurate positioning is the key and premise for its correct recognition in the automatic target recognition system. The vehicle license plate has a distinct texture features and the regular shape, but the texture features, shapes and sizes of various logos are different. So, it may take a large amount of computing time by using the conventional matching positioning algorithm. Based on the above analysis we can find that the positioning of the vehicle logo has become the application bottleneck of the logo recognition technology in the real-time system. This paper proposed the algorithm for logo positioning from coarse to fine:the vehicle logo is mostly in the central axis of the car, based on this characteristic, this paper proposed the methods for logo positioning from coarse to fine, including the head positioning, central axis positioning, the coarse positioning of the logo and the accurate positioning of the logo.(2) The research of the technical methods about the logo recognition:At present, the scholars at home and abroad mainly focused on the recognition of the vehicle license plate, and tried to achieve a rough classification of the vehicle type by analyzing the vehicle’s shape, size and color. So, the research of the vehicle branch is relatively small. The key to the recognition of the logo lies to find its main features. and select the proper classification method for the classification and identification. By the influence of the light, noise and shooting angle, the conventional recognition algorithm is difficult to achieve the satisfying results. This paper proposed the logo recognition algorithm which based on the SIFT features, and combined the template matching and feature matching. It greatly improved the recognition accuracy rate of the vehicle logo. SIFT (Scale Invariant Feature Transform) is one of the most active algorithms in the field of image matching. SIFT features has a favorable invariance for the image scaling, translation and rotation. It also has certain robustness to illumination changes, affine transformation or three-dimensional projection. Due to the invariance advantage of the SIFT features, SIFT algorithm is widely applied to the field of image matching. The experimental results show that this algorithm’s positioning rate is96%for the actual shooting of the video image, the techniques in this paper is reasonable, and it has greater theoretical significance and practical value.