基于SVM的室外场景图像的景物识别  

Recognition of Outdoor Scene Image Features Based on SVM

作　　者：郑梦兴[1] 薛静[1] 张通[1] 

机构地区：[1]西北工业大学自动化学院,陕西西安710072

出　　处：《计算机仿真》2011年第4期250-253,339,共5页Computer Simulation

基　　金：西北工业大学科技创新基金;航天支撑基金项目(N9XW001)

摘　　要：关于图像识别优化问题,针对在图像的处理中颜色、纹理等单一特征不能全面描述图像的问题,为了精确识别目标图像,提出一种颜色特征和纹理特征相结合的方法。采用基于RGB空间的颜色直方图和基于多通道Gabor滤波器分别对颜色特征和纹理特征进行提取,并对颜色与纹理特征进行外部归一化得到组合特征,并应用支持向量机法(SVM)对组合特征样本训练分类。实验结果表明,改进方法可以克服用单一特征分析图像的片面性,同时提取了理想的图像结构和光谱特性,可以在图像目标的识别和分类中取得较好的效果。Because color texture and other individual characteristics can not fully describe the image problem in the image,the paper puts forward a new method which combines the advantages of the color features and texture features.The color histogram based on the RGB color space is extracted as color feature;the texture feature is extracted based on multi-channel Gabor filters,then combination of color and texture feature vectors is obtained after external normalization,and the support vector machine method（SVM） is used on the combination of characteristics of samples for training and classification.Experimental results show that the method can overcome the one-sidedness of single feature in image analysis,extract the structure and spectral characteristics of the image,and achieve better results in target recognition and classification.

关 键 词：颜色矩 纹理特征 伽柏滤波器 支持向量机 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]