Combining Region-Based and Point-Based Algorithm to Detect Vehicle Plate Location

Electronic payment applications (such as toll roads and parking), toll road monitoring applications, and traffic monitoring applications are examples of applications supporting smart infrastructure systems. One of the most critical aspects of such smart infrastructure support systems is how to recognize a vehicle. There are two main problems in vehicle license detection: the plates and how big they are. Usually, the candidate character's position on the plate is first identified, and the square area of the plate is determined later. According to the research, it is well known that Maximally Stable Extremal Regions (MSER) feature detector can find the text area well, and Speed-Up Robust Features (SURF) and neural networks are highly capable of character recognition. This paper combines a region-based detection algorithm and a point-based algorithm at the feature extraction stage to detect vehicle number plates' location. Based on experimental results, combining MSER and SURF methods could detect vehicle number plates' location well. The results show that the proposed method has achieved 97.63% accuracy, 69.17 precision, and a recall value of 66.14 with an average computation time of 32.92 seconds.