基于ABT-SVM算法行人检测系统设计

Flow and Implementation of Pedestrian Detection System Based on AdaBoost SVM Algorithm

行人检测方法在智能监控系统、智能辅助驾驶、人机交互等方面有着广泛的应用价值,针对传统基于AdaBoost(ABT)和SVM分类器存在的问题,采用AdaBoost和RBFSVM相结合的算法设计分类器进行行人检测,在特征提取上采用OpenCV提供HOGDescriptor类进行HOG特征提取,最后通过实验分析,结果表明,本文设计的AdaBoost-RBFSVM级联的分类器在分类分类准确率和误报率等方面都优于传统的AdaBoost级联分类器,同时还可以保证分类器的训练速度和检测速度。

Absrtact:Pedestrian detection method has a wide range of application in intelligent monitoring system,intelligent assistant driving,human-computer interaction and so on.Aiming at the problems of traditional classifier based on AdaBoost and SVM,this paper uses AdaBoost and rbfsvm algorithm to design classifier for pedestrian detection,and uses opencv in feature extraction.Finally,through the experimental analysis,the results show that the AdaBoost rbfsvm cascade classifier designed in this paper is superior to the traditional AdaBoost cascade classifier in terms of classification accuracy and false alarm rate,and can also ensure the training speed and detection speed of the classifier.