Traffic sign array decomposition using support vector machines

Generally, in most traffic sign recognition systems based on image processing, the recognition process is performed individually, which means that every traffic sign must be isolated from the background, and from other traffic signs, before the object goes into the recognition process. When it comes to a traffic sign array, sometimes the previous blocks do not succeed in the separation of the array signs, and the recognition of the signs is bound to fail. In this work, we have developed an algorithm based on Support Vector Machines and the structural information of traffic sign arrays to separate the signs of those arrays which were not detected isolated, as it would be the desired case. The algorithm has been tested over a set of real outdoor images which contain traffic sign arrays. The experimental results show good performance in the detection and decomposition of arrays that would, otherwise, be missed.