基于决策导向非循环图SVM的汽车车型识别

Vehicle image recognition based on DDAG SVM

用决策导向非循环图支持向量机研究了汽车车型的识别。运用图像像素相减的差分方法去除背景,获取车对象;然后对图像进行均值滤波,以除去噪声干扰,再对图像用"分水岭"变化的阈值方法,获取车的二值图像;最后提取车的几何形状特征,并将其输入DDAG支持向量机进行训练和识别,以实现对车型分类的目的。

This paper studies a vehicle image recognition based on DDAG SVM. Firstly, image pixel subtraction theory is adopted to extract vehicle. Secondly, it uses average filter to remove noise. Thirdly, threshold is segmented by using the ＂watershed＂ changes to get a two-image car. At last, the feature vectors of the object is extracted and a DDAG SVM is trained to classify vehicle, in order to realize the classification of vehicle type.