Vehicle Identification in Haze Based on V-System and Dense Convolution Network

Abstract

Haze increases the possibility of accidents. Therefore, it is important to improve the accuracy of the vehicle identification system in such a situation. This research proposes an image pre-processing algorithm for the system. Firstly, the accuracy is enhanced by the Single Scale Retinex algorithm. Secondly, the pictures are processed by V-system processing. Finally, the pre-processed pictures are identified by dense convolution network. The results show that the algorithm’s accuracy is 0.93% higher than the traditional method.