摘要
为解决动物图像分类识别问题,提出了一种基于颜色特征的动物图像分类识别方法。该方法很好地利用了颜色直方图和低阶统计矩的属性。首先确定选择HSV颜色空间,依据图像的颜色直方图特性调整低阶统计矩数值作为特征描述量,再通过K近邻方法判断图像所属类别,并构建仿真系统。实验另外对比了组合类与单一类特征在识别准确率上的差异。实验表明,该方法能较有效地识别出不同种类的动物,平均正确识别率可达89%。仿真系统具有实用性,组合类特征可在一定程度上提高识别准确率并降低识别时间。
To solve the problem of animals’images recognition,a kind of classified method based on their color features is put forward.The method takes the advantage of nature of color histogram and lower statistical moment.First,the HSV color space is chosen and the lower statistical moment value is modified based on the nature of color histogram as the characteristic statistic variables.Then the images category is judged through k-nn search method and a simulation system is built.The experiment also compares the classified accuracy of combined class with that of single class.The experiment shows that the method can classify the animals effectively and averaged class recognition rate reaches89%.The simulation system is practical and combined class features can improve recognition accuracy and reduce its time to some extent.
作者
张公伯
谷昱良
朱和贵
ZHANG Gongbo;GU Yuliang;ZHU Hegui(Institutions of Mathematic,Northeastern University,Shenyang 110004)
出处
《舰船电子工程》
2017年第5期81-85,共5页
Ship Electronic Engineering
基金
国家级大学生创新创业训练计划资助项目(编号:201610145014)
中国博士后科学基金资助项目(编号:2016M591446)
中央高校基本科研业务费(编号:N140503004)
国家自然科学基金青年科学基金资助项目(编号:61402097)资助
作者简介
张公伯，男，研究方向：应用统计、图像处理。;谷昱良，男，研究方向：图像处理。;朱和贵，男，博士，副教授，研究方向：机器学习、统计建模、图像处理、多媒体信息安全。
