基于MapReduce的公路视频图像车型分类研究*

分析公路视频图像,从而对经过的车辆进行较高精度的分类,是一个颇且实用价值的课题.如何在保证分类精度的同时提高系统性能,无疑是一个具有挑战性的任务.提出了一个多特征融合的分类框架,结合车辆的全局几何特征、SIFT局部特征,以及Gabor纹理特征对车辆进行分类,提高了分类精度;为了提高系统的性能,设计了基于MapReduce的并行算法,通过对图像分块,实现数据并行.实验结果表明,该方案能够在提高分类精度的基础上仍然保持较高的系统性能.

Vehicle Classification based on Highway Video Using MapReduce Programming Model

Vehicle classification with a higher accuracy based on highway video, using image analysis technology is a very valuable subject. However, how to improve system performance while ensuring the classification accuracy is undoubtedly a challenging task. Since both global features and local features are essential to classification, this paper presents a multi-feature fusion classification framework, which combines with global geometric features, SIFT, and gabor local features to improve the classification accuracy. In order to improve system performance, a parallel algorithm based on MapReduce programming model is designed. Experimental results show that this scheme can improve the classification accuracy and still maintain a high system performance.