Shape searching in real world images: a CNN-based approach

One of the major problems in computer vision is to build systems with the ability of detecting shapes in arbitrary "real world" images. The target application of our work is the correct identification of road traffic signs in images taken by a car mounted camera. The basic technique used in this kind of situation is to compare each portion of an image with a set of known models. The approach taken in our work is to implement this comparison with cellular neural networks, making it possible to efficiently use a massively parallel architecture. In order to reduce the response time of the system, our approach also includes data reduction techniques. The results of several tests, in different conditions, are reported in the paper. The system correctly detects a test shape in almost all the experiments performed. The paper also contains a detailed description of the system architecture and of the processing steps.