基于HOG和SVM的图像识别方法
首发时间:2013-12-23
摘要:图像识别技术目前已广泛应用于人们的工作与生活中,不同的应用场景对图像识别技术提出了不同的应用要求。本文提出一种基于HOG和SVM的图像识别方法。首先对输入图像进行校正与归一化等预处理,去除背景噪声的干扰,然后利用梯度方向直方图(HOG)提取图像特征,建立对图像的鲁棒表示,最后使用SVM分类器对送入的图像特征模式进行分类。为了降低计算复杂度,采用了一种积分图方法对HOG特征提取过程进行了优化。实验结果表明,本文提出的方法能够较好地对图像进行分类,具备较高的识别准确率。
For information in English, please click here
Image recognition method based on HOG and SVM
Abstract:Nowadays image recognition technology is widely employed in our work and daily life, however different application scenarios put forward different requirements. An image recognition method based on HOG and SVM is proposed. Firstly, in order to remove background noise, an image correction and normalization is required. Secondly, we establish a robust representation of image by using HOG feature extraction. Finally we classify different input feature vectors by using SVM. We take an integral image approach, which is applied to optimize HOG feature extraction, to reduced the computational complexity. The experiment results show that our method achieve a good performance on image recognition.
Keywords: Image Recognition Projection Transformation HOG SVM Integral Image
基金:
论文图表:
引用
No.****
同行评议
勘误表
基于HOG和SVM的图像识别方法
评论
全部评论0/1000