License Plate Recognition for Moving Motorcycles Using Industrial Computer Vision Library

License plate recognition technique has been widely used in daily life. It mainly detects the license plate characters by computer vision technology. Common application scenarios such as parking lot management, traffic management, and so on. The target of many license plate recognition systems is usually vehicles. These systems usually operate in a fixed environment, such as the angle of the camera or environmental brightness. On the other hand, motorcycles in Taiwan occupy a relatively high proportion of transportation. Motorcycle stability is far less than that of vehicles when moving. It causes the angle of the motorcycle license plate in the camera's image is dynamic. The above condition increases the difficulty of motorcycle license plates recognition. In order to solve the above problem, a license plate recognition system for moving motorcycles are proposed in this paper. The system is divided into three stages: the first stage is license plate detection, that is, locates the license plate in the image. The second stage is license plate calibration. The license plate is transformed to a fixed front view without influenced by the actual angle in the camera image. The third stage of optical character recognition (OCR) is to extract the license plate characters. In order to achieve high accuracy and stability of the system, Euresys Open eVision, an industrial computer vision library was integrated. In the experimental study, the accuracy of the system reaches above 95.5% with an average recognition time of 73.4ms. The proposed system achieves the purpose of license plate recognition of moving motorcycles.