Research & Implementation Of Statistical-Feature-Based Vehicle Recognition Algorithm

With the economical development and the speedup of urbanization and motorization, modern management methods are needed to administer traffic, which begets the research on intelligent transportation system (ITS). Driver assistance system (DAS) is an important component of ITS, and has a bright prospect owing to its intrinsic pre-accident warning mechanism, especially in high quality active automobile safety and accident avoidance measurement. DAS based on computer vision technology is one of the booming research areas, due to their far signal detection range, detection integrity and excellent performance-cost ratio.Vision-based vehicle detection algorithm has been used in driver assistance system over the past decades of years. Robust and reliable vehicle detection is an important component of driver assistance system. On-road vehicle detection is a very challenging task. Vehicles, for example, come into view with different speed and may vary in shape, size, and color. Vehicle appearance depends on its pose and is affected by nearby objects. In-class variability, occlusion, and lighting conditions also change the overall appearance of vehicles.This paper summarized the vehicle detection algorithms used in recent years, and analyzed their characteristics, then designed and implemented a vehicle detection algorithm which contains two main steps: hypothesis generation (HG) step and hypothesis verification (HV) step. In the hypothesis generation step, possible image locations where vehicles might be present are hypothesized by using vehicle pre-knowledge. Hypothesis verification step verifies those hypotheses using statistical pattern recognition method.Processing flow of vehicle recognition using statistical pattern recognition in this paper composed of three parts: feature extraction, feature selection, classification decision. Use the feature extraction method (such as Gabor, PCA, and Wavelet) to extract feature on the vehicle and background training or recognition image, and the moment feature is extrcated also. Use the GA (genetic algorithm) to implement feature selection on the feature extraction result, then...