Traffic sign shape classification based on correlation techniques A. VÁZQUEZ-REINA, S. LAFUENTE-ARROYO, P. SIEGMANN, S. MALDONADO-BASCÓN, F. J. ACEVEDO-RODRÍGUEZ Departamento de Teoría de la Señal y Comunicaciones Universidad de Alcalá Escuela Politécnica Superior. Campus Universitario.28871, Alcalá de Henares, Madrid SPAIN Abstract: - In this paper we present a correlation-based matching method for traffic sign shape classification. Our purpose is to offer a robust and reliable framework which can be used in numerous applications like driver assistance systems. The shape classification is scale, translation and rotation invariant. The process involves obtaining essential features (e.g. edges, ridges, corners) from each area, and comparing it to the stored templates of known patterns. The algorithm is very flexible, easy to reconfigure for many different shapes and the results we obtained show the success rate. Key-Words: - Advance driver-assistance systems (ADASs), intelligent vehicles, road sign detection and classification. 1 Introduction The information provided by either traffic signs or route-guidance signs is extremely important for safe and successful driving. An ADAS with the automatic ability to extract and identify these signs would help human drivers to avoid traffic hazards. Such systems would indicate the presence of a sign in advance so that human incorrect decisions could be avoided, and could assist drivers on signs they didn’t notice. There are many examples of ADAS which aid piloting and navigation, many of them include road lane detection, obstacles recognition and avoidance in the vehicle’s path [1,2,3], traffic lights and marks on roads detection [4] etc. Traffic sign classification techniques can be seen as pattern recognition systems which aim to classify data based on either a priori knowledge or on statistical information extracted from observations. In most of these systems, the patterns to be classified are usually groups of measurements defining points in an appropriate multidimensional space. The classification or description scheme is usually based on the availability of a set of patterns that have already been classified or described. This set of patterns is termed the training set and the resulting learning strategy is characterized as supervised. Learning can also be unsupervised, in the sense that the system is not given an a-priori labeling of patterns, instead it establishes the classes itself based on the statistical regularities of the patterns. Pattern recognition is typically an intermediate step in a longer process. These steps generally are acquisition of the data (image, sound, text, etc.) to be classified, preprocessing to remove noise or normalize the data in some way (image processing, stemming text, etc.), computing features, classification and finally post-processing based upon the recognized class and the confidence level. The classification or description scheme usually uses one of the following approaches: statistical (or decision theoretic), syntactic (or structural), or neural. Statistical pattern recognition is based on statistical characterizations of patterns, assuming that the patterns are generated by a probabilistic system. Structural pattern recognition is based on the structural interrelationships of features, and neural pattern recognition employs the neural computing paradigm that has emerged with neural networks. A typical approach for detecting and recognizing road sign shapes consists of 2 stages. First, color segmentation or color thresholding is applied to distinguish possible signs in the image, then template matching is applied for actual shape detection [5]. Wei et al [6] proposed a method based on color filter, boundary smoothing by close and open morphological operations, and a shape analysis which evaluates the similarity between a region and a given shape, it includes major and minor axis as well as comparing area. Jim Torresen. Jorgen W. Bakke and Lukas Sekanina [7] proposed an automatic Proceedings of the 5th WSEAS Int. Conf. on Signal Processing, Computational Geometry & Artificial Vision, Malta, September 15-17, 2005 (pp149-154)