照片分类法与图像识别技术相结合的森林可燃物分类  

Classifying Forest Fuel Types by Photo Classification and Image Recognition

作　　者：刘方策 张运林[1] 满子源[1] 孙龙[1] Liu Fangce;Zhang Yunlin;Man Ziyuan;Sun Long(Northeast Forestry University,Harbin 150000,P.R.China)

机构地区：[1]东北林业大学,哈尔滨150040

出　　处：《东北林业大学学报》2018年第11期51-57,共7页Journal of Northeast Forestry University

基　　金：国家自然科学基金项目(31470658);国家林业局"948"项目(2014-4-06)

摘　　要：以大兴安岭地区51块样地为研究对象,通过聚类分析划分5种可燃物类型,运用森林可燃物类型照片分类法与图像识别技术对森林可燃物进行分类。结果表明:照片分类法中,每层输入3张照片的情况下,输出结果无显著差异,随着照片数量增加,输出结果的差异越来越小;每块样地每个层次随机组合输入3张照片,采用Top3结果输出法准确率较高。因此,该方法在不需要高精度参数的情况下,可以快速划分可燃物类型,降低野外调查成本,及时为森林防火工作提供可燃物类型和调查参数。With 51 plots in the Daxing’an Mountains,the forest fuels were classified into five types by cluster analysis.Five fuel types were classified by cluster analysis.Classification of forest fuel using forest fuel type photo classification and image recognition techniques.In the photo classification method,when three photos are input on each layer,the output results are not significantly different,and as the number of photos increases,the difference in output results becomes smaller.Each piece of sample plot randomly inputs three photos at each layer,and the accuracy of the Top3 result output method is higher.Therefore,the method can quickly classify forest fuel without reducing the high-precision parameters,reduce the cost of field survey,and provide fuel types and survey parameters for forest fire prevention work in a timely manner.

关 键 词：森林可燃物 可燃物类型 照片分类法 图像识别 

分 类 号：S715.3[农业科学—林学]