Research on Methods of Truck Plate Detection in Expressway Scene Based on Machine Learning

Due to the long shooting time span of truck images and the drastic changes in lighting conditions during daytime and night time, phenomena such as large differences in image brightness, overexposure, blurred truck plate writing and edges of characters will occur. Therefore, in order to accurately locate the position of the truck plate in the image, this paper studies the truck plate location method in the expressway scene based on machine learning, and compares and analyzes six different truck plate location methods such as deformable part model, edge-information-based model, color-information-based model, template matching model, character-feature-based model, and HOG-feature-based support vector machine model. Then, six different experiments based on different methods by actual truck image dataset were implemented. The experimental results show that the truck plate location method based on deformable part model is better than the other five truck plate location methods, and its location accuracy reaches 99.34%.