Research on Vehicle Identification Method Based on Computer Vision

Abstract

Identifying the vehicle in front of road is an important research topic for active safety and intelligent driving of vehicles. A vehicle identification algorithm is proposed based on computer vision using supervised machine learning algorithm AdaBoost and Haar-like features. Firstly, in terms of feature selection, dimension reduction processing is performed from two aspects of feature type and feature size, and integral graph is applied to accelerate the calculation of Haar-like eigenvalues. Secondly, a more efficient classifier is constructed based on a small number of effective features, and a single strong classifier is used to identify and verify the vehicle in front. Finally, the whole vehicle identification algorithm is tested with the test data including 350 frames captured from the highway video set and 450 frames captured from the urban road video set. The result shows that the vehicle identification algorithm have a high detection rate and Lower detection error rate.