Automatic recognition of car models - based on Matlab

Abstract

Based on vehicle classification standards, the construction of autonomous driving modes requires collecting relevant information, including parameters related to vehicle types. Subsequently, this mode can automatically classify vehicles (i.e., vehicle type identification), serving as the foundation for autonomous driving while providing data support for subsequent manual inspections and statistical analysis. The system achieves self-identification of vehicle categories through applications of image processing, image interpretation, and pattern recognition. The main research focuses are on three aspects: image preprocessing, feature extraction of vehicle characteristics, and vehicle type recognition based on BP neural networks. To enhance system accuracy and stability, we propose the following strategies: image enhancement, filtering, and edge detection. Features such as roof length ratio, roof height ratio, and front-rear ratio are extracted. A BP network model is constructed and BP algorithm is implemented. Through training and testing of eight samples from two categories, we have essentially completed preliminary classification of common vehicle types. Systematic testing confirms identification accuracy of 90% for both buses and sedans, demonstrating that our technical capabilities meet practical application requirements across all aspects.