摘要
以五营森林公园中不同腐朽程度的红松为研究对象,选取无损、易获取指标(阻力损失及心材、边材异常面积比)为模型自变量,腐朽等级为因变量,利用无损检测并结合随机森林对活立木腐朽程度进行分级。使用重复分层K折交叉验证保证样本数据在训练和测试时分类均衡,并建立随机森林立木腐朽分级模型。结果表明:随机森林腐朽分级模型训练、测试精度分别为100%、90%,说明其拟合效果好、泛化能力强。随机森林对腐朽等级(Ⅰ、Ⅱ、Ⅳ、Ⅴ)的辨识均正确,召回率都为100%,腐朽等级Ⅲ错判1例,对腐朽立木进行分级的准确率高达96.67%,随机森林的微观、宏观平均AUC值分别为0.979 2、0.970 8。
Taking the Korean pine with different decay degrees in Wuying Forest Park as the research object, the non-destructive and easy-to-obtain indicators(resistance loss, heartwood and sapwood abnormal area ratio) were selected as the independent variables of the model, the decay grade was the dependent variable. The decay degree of living trees was classified by nondestructive testing combined with random forest. The repeated stratified K-fold cross validation was used to ensure the classification equilibrium of sample data in training and testing, and the random forest tree decay classification model was established. The training and testing accuracy of random forest decay classification model were 100% and 90%, respectively, indicating that the model has good fitting effect and strong generalization ability.Random forest could correctly identify all decay grades(Ⅰ, Ⅱ, Ⅳ, Ⅴ), the recall rate was 100%, and one case was wrong in decay grade III. The overall classification accuracy of decayed living trees was as high as 96.67%. The micro and macro average AUC values of random forests were 0.979 2 and 0.970 8, respectively.
作者
谢军明
王立海
林文树
郝泉龄
解光强
孟庆凯
李怡娜
阚相成
Xie Junming;Wang Lihai;Lin Wenshu;Hao Quanling;Xie Guangqiang;Meng Qingkai;Li Yina;Kan Xiangcheng(Northeast Forestry University,Harbin 150040,P.R.China)
出处
《东北林业大学学报》
CAS
CSCD
北大核心
2022年第4期99-103,110,共6页
Journal of Northeast Forestry University
基金
国家自然科学基金项目(31570547)。
关键词
随机森林
活立木
腐朽
分级
Random forest
Living tree
Decay
Classification