Motorcycle and License Plate Detection Using Fixed-Size Vertical Projection and Multi-Part Mean Analysis

This paper proposed methods for motorcycle and license plate detection using fixed-size vertical projection and multi-part mean analysis. The method consisted of two parts: a motorcycle detection and a license plate detection. Fixed-size vertical projection and neural network were used in the motorcycle detection process. Vertical and horizontal scanning was used to extract features for the neural network. Later, a rough filter and a thorough filter were used to approximately locate and finally confirm a license plate, respectively. These methods were tested on both motorcycles with license plates and motorcycles without license plates. The experimental results gave accuracy of 93.20% for motorcycle detection, 93.33% for motorcycles with license plates, and 87.23% for motorcycles without license plates.