彭娟
摘 要: 针对当前图像分类模型无法满足实际应用要求的难题,为了得到更优的图像分类效果,提出特征选择和聚类分析相融合的图像分类模型。首先提取图像的原始特征,采用主成分分析对图像特征进行选择,然后采用聚类分析算法对图像样本进行处理,选择与待分类图像相关的样本,减少训练样本的规模,最后采用支持向量机建立图像分类器,对标准图像库中的图像进行分类实验。实验结果表明,该模型减少了图像分类的特征和图像分类的训练样本,加快了图像分类建模的速度,同时图像分类正确率明显高于其他图像分类模型。
关键词: 图像处理; 原始特征; 聚类分析算法; 图像分类器
中图分类号: TN911.73?34; TP181 文献标识码: A 文章编号: 1004?373X(2017)19?0079?04
Image classification model based on feature selection and clustering analysis
PENG Juan
(Chongqing Institute of Engineering, Chongqing 400056, China)
Abstract: Aiming at the problem that the current image classification model cannot meet the requirements of practical application, an image classification model based on feature selection and clustering analysis is proposed to obtain better results of image classification. First, the original image features are extracted and the principal component analysis (PCA) is used to select the image features. Then, the clustering analysis algorithm is adopted to process image samples, select samples relevant to the images waiting for classification, and reduce the scale of training samples. Finally, the support vector machine is used to construct the image classifier and classification experiments are carried out for images in the standard image database. The results show that, by using this model, the features and training samples of image classification are reduced, the image classification modeling is sped up, and the correctness of image classification is obviously higher than that of other image classification models.
Keywords: image processing; original feature; clustering analysis algorithm; image classifier
0 引 言
图像分类是图像处理研究领域中一个重要方向,广泛应用于军事、工业生产、图像分析、医学以及环境保护等领域[1]。图像分类过程十分复杂,到目前为止,图像分类过程还存在一些待解决的难题,获得理想的图像分类效果面临巨大挑战[2?3]。
特征提取和选择是获得理想图像分类结果的基础[4]。当前图像分类特征很多,如纹理特征、颜色特征等,为了更好地描述图像类别信息,人们尽可能多的提取图像特征,使得图像的特征数量急剧增加[5]。图像特征数量大给图像分类带来两个方面的难题:一是图像分类器的输入维数高,结构十分复杂,设计合理结构的图像分类器难度增加,使图像分类速度急剧变慢;……