A Statistical Approach for the Automatic Recognition of Traffic Sign Deterioration

摘要

This paper describes a software application based on statistical methods for the automatic recognition of traffic sign deterioration. The evaluation of traffic sign degradation is usually performed by devices applied on top of the road sign surface, measuring color parameters such as chromatic coordinates and the luminance factor. Moreover, the devices can only check a small fraction of the traffic sign surface at a time, requiring several acquisitions on the same traffic sign. In order to reduce the costs related to monitoring and have a periodic control of the traffic sign status, we propose a fast automatic method based on video acquisition and processing that can be easily operated in patrolling vehicles provided with a camera. A pattern detection algorithm based on color and texture features is applied to the images extracted from the acquired videos in order to detect the traffic signs ROIs, which are analyzed using a statistical approach based on the Kullback-Leibler divergence and Kolmogorov-Smirnov test. Making use of a control sample of not deteriorated traffic sign images, a comparison between the acquired and the reference images is performed. Both statistical methods have been used to compare 150 pairs of traffic signs, achieving high precision and recall, proving that the proposed approach can be a good candidate solution for automatic traffic sign deterioration analysis.