基于稀疏编码的车型图像分类研究

实现车型分类是智能交通系统中的重要技术,为了克服传统车型分类方法测量困难的弱点,研究基于车型图像的车型分类方法.图像特征采用了基于图像局部特征稀疏编码的直方图表示,稀疏编码量化误差较小,更能精确地捕捉图像最突出的信息,使用了支持向量机(SVM)作为分类器对车型图像进行分类.实验结果表明本算法能使6类车型图像分类正确率达到90％以上,相对于现有方法有一定的提升.

Vehicle Image Classification Based on Sparse Coding

Vehicle classification is an important technology in intelligent transportation systems. In order to overcome the hard-to-measure weaknesses of traditional classification method, this study employes the automatic classification method based on vehicle image. The image features are based on the histogram of the image local feature sparse coding. The sparse coding can capture the most prominent information in an image accurately and causes less quantization error. The experiment uses support vector machine (SVM) as classifier for vehicle image classification. The experimental results show that the algorithm for image classification can achieve the accuracy of more than 90% of the six categories of vehicle with a certain increase relative to existing methods.