一种树形结构的实时道路车辆识别方法

针对实时道路车辆识别是车辆计算机辅助驾驶、自主导航以及主动安全的关键技术，给出了一种基于树形结构的车辆识别方法。该方法采用Haar—like特征来表达车辆特征，选择GentleAda Boost算法训练出强分类器，最后将多个强分类器组合成树形结构。通过建立样本库，对车辆识别分类器进行了训练，并对分类器的性能进行了实验。结果表明：对750个目标车辆进行识别，识别率为74．4％，识别时间为34．4ms。说明树形结构的车辆识别方法具有较好的实时性和一定的鲁棒性。

Realtime On-Road Vehicle Detection Approach Based on Tree Structure

Aiming at the importance of realtime on-road vehicle detection in driver assistance systems,autonomous driving and active safety,a vehicle detection algorithm based on tree structure is introduced.Haar-like features are used in this application,and Gentle AdaBoost algorithm is chosen to train the strong classifiers wich form the tree structure vehicle classifier.The data sets are collected,and vehicle detection classifier is trained and evaluated.The experiments on test data sets wich contain 750 target vehicles show that the detection rate is 74.4% and the processing time is about 34.4 ms.It can be seen that the tree structure classifier is significantly efficient and robust.