all in one license plate recognition machine < / a > manufacturer's brief introduction to the process of license plate recognition

In the process of license plate recognition, the recognition of license plate color is based on different algorithms and may be realized in the above different steps. It usually cooperates and verifies with license plate recognition

　　1. License plate positioning

In the natural environment, the background of automobile image is complex and the illumination is uneven. How to accurately determine the license plate area in the natural background is the key of the whole recognition process. Firstly, a large-scale correlation search is carried out on the collected video image, and several regions in line with the characteristics of the vehicle license plate are found as candidate regions. Then these candidate regions are further analyzed and evaluated. Finally, a best region is selected as the license plate region and separated from the image

　　2. License plate character segmentation

After the location of the license plate area is completed, the license plate area is divided into a single character, and then recognized. Character segmentation generally adopts vertical projection method. Because the projection of characters in the vertical direction must be near the local minimum value at the gap between or within characters, and this position should meet the character writing format, characters, size restrictions and some other conditions of the license plate. The vertical projection method has a good effect on character segmentation in automobile image in complex environment

　　3. License plate character recognition

The main methods are template matching algorithm and artificial neural network algorithm. Based on the template matching algorithm, firstly, the segmented character is binarized and its size is scaled to the size of the template in the character database, then it is matched with all templates, and the best matching is selected as the result. There are two algorithms based on artificial neural network: one is to extract the features of characters, and then use the obtained features to train the neural network distributor; Another method is to input the image directly into the network, and the network automatically extracts the features until the recognition result is obtained

In practical application, the recognition rate of license plate recognition system is also closely related to the quality of license plate and shooting. The quality of license plate will be affected by various factors, such as rust, stain, paint peeling, font fading, license plate shielding, license plate inclination, highlight and reflection, multiple license plates, false license plates, etc; The actual shooting process will also be affected by environmental brightness, shooting mode, vehicle speed and other factors. These factors reduce the recognition rate of license plate recognition to varying degrees, which is the difficulty and challenge of license plate recognition system. In order to improve the recognition rate, in addition to constantly improving the recognition algorithm, we should also find ways to overcome various lighting conditions to make the collected images most conducive to recognition