一种改进的SVM的昆虫图像检索算法及仿真

An improved SVM insect image retrieval algorithm and simulation

物种的变异性导致种群数量繁多,如何快速、准确地从海量的昆虫图像数据库查询吻合用户意图的图像成为棘手问题。基于内容的图像检索从图像本身出发,提取图像的底层特征与语义特征,提高了检索结果准确性。提取图像颜色与纹理综合特征,运用SVM建立训练模型,使得预测图像样本逼近训练样本,实现了基于SVM的图像检索仿真。在此基础上,提出颜色与纹理特征结合图像分块特征,并借鉴Bag of Words模型,弥补了图像空间分布信息,更加全面地描述了图像内容。实验表明,全面的特征提取提高了检索精度。

The variability of species results in various populations.How to demand images which match user intent fast and accurately from the mass of insects image database becomes a thorny issue.Content-based image retrieval is based on the image itself,and extracts the image features and semantic features of the bottom of society,which improves the accuracy of the retrieval results.By comprehensive features extraction of image color and texture,this paper uses the SVM training model to forecast the image sample close to the training sample,and realizes the simulation of image retrieval based on SVM.On this basis,the color and texture feature combing with image block features,and drawing lessons from the Bag of Words model,this paper makes up the image spatial distribution information,and describes the image content more fully.Experiments show that the comprehensive feature extraction improves the retrieval accuracy.