Visual saliency-based vehicle logo region detection

Abstract

Vehicle logo detection (VLD) is one of the crucial parts of intelligent transportation system (ITS).VLD methods are mostly based on learning progress or the relative position of vehicle logo and license plate.However, the learning progress is time-consuming, and the relative position above limits the application of VLD, especially when the license plate is removed.In this paper, a novel VLD method, based on the features of vehicle logo using saliency detection is proposed and solved the two problems above.Three dataset containing totally 3000 images is generated to assess the accuracy of this system.A detection rate of 92.27% is finally obtained, demonstrating the robustness and efficiency of our method.