遗传算法确定特征权重值的图像分类  

Image classification to determine feature weight values by genetic algorithm

作　　者：唐彩红 TANG Caihong(Qilu Medical University,Zibo 255300,China)

机构地区：[1]齐鲁医药学院

出　　处：《现代电子技术》2020年第3期58-61,65,共5页Modern Electronics Technique

摘　　要：针对多种特征权重值无法合理设置,得到的检索效果无法达到用户要求的问题,将多种特征融合在一起,对图像进行分类处理,可以弥补单一特征检索带来的局限性问题,并采用遗传算法对图像特征进行提取并进行优化,以获取最优的特征权重值,将获取的最优权重值用于图像检索与分类中。运用加权融合方法对图像的颜色、纹理等特征进行加权处理,可以有效实现多种特征融合的目的。通过实验证明,基于遗传算法确定特征权重值的图像检索与分类方法具有较强的学习效果,可以自动为特征权值进行赋值,大大提高了图像分类的简洁性,提升了图像检索的效果。In order to deal with the fact that it is difficult to set multiple feature weight values reasonably and the retrieval results fail to satisfy user′s requirements,multiple features are fused to classify images,which can make up the limitation of single feature retrieval.The genetic algorithm is used to extract and optimize the image features to obtain the optimal feature weight values for image retrieval and classification.The image features like color and texture can be weighted by the weighted fusion method,which can effectively achieve the purpose of multi⁃feature fusion.The experiment results show that the method of image retrieval and classification to determine feature weight values by genetic algorithm has obvious learning effect and can automatically assign values to feature weight,which greatly improves the simplicity of image classification and the effect of image retrieval.

关 键 词：遗传算法 特征权重值 图像分类 卷积神经网络 初始种群 检索率 

分 类 号：TN911.73-34[电子电信—通信与信息系统]