基于无人机影像的天山云杉林冠幅提取及蓄积量反演

Crown extraction and volume inversion of Picea Schrenkiana var tianshanica forest based on UAV image

以无人机影像为源数据,天山云杉(Picea Schrenkiana var tianshanica)为研究对象,根据样方实测数据,提取冠幅信息与胸径树高参数进行拟合并反演蓄积量,为天然林保护工程实施后天山云杉生态恢复与科学管理提供技术手段。基于新疆农业大学实习林场的无人机影像数据和60个样方的每木检尺数据,通过多尺度分割技术对天山云杉林冠幅进行分割提取,根据实测数据建立冠幅-胸径模型和冠幅-胸径-树高模型,反演天山云杉林蓄积量。结果表明:无人机影像分割效果较好,根据多尺度分割技术得到的最合优尺度参数为60,形状因子为0.5,紧密度因子为0.5。通过构建天山云杉冠幅-胸径模型,得幂函数模型拟合度最高,R2为0.718,而冠幅-胸径-树高二元一次模型R2为0.817。将提取参数代入模型反演蓄积量与样方蓄积量对比后完成T检验,检验结果sig=0.051,说明模型拟合度较好。研究得出采用面向对象方法对天山云杉林冠幅信息的提取效果较好,基于模拟数据和实测数据构建模型得到胸径参数进而反演林分蓄积量的方法可降低人力调查成本并满足天山云杉林经营管理的需要。

The crown information of Picea Schrenkiana var tianshanica forest was extracted by UAV image and based on the measured data of the sample site,and the parameters of diameter at breast height(DBH)and tree height were proposed to be combined to invert the accumulations,providing technical means for the scientific management of forest after the implementation of the natural forest protection project.On the basis of the UAV image from the practice forest farm of Xinjiang Agricultural University and the data of 60 sample plots,the crown of forest was segmented and extracted by multi-scale segmentation technology.The crown-DBH model and crown-DBH-tree height model were established based on the actual measurement data to invert the stock of forest.The image segmentation effect of UAV is better.According to the multi-scale segmentation technology,the optimal scale parameter is 60,the shape factor is 0.5,and the compactness factor is 0.5.By constructing the DBH-crown model of forest,the fitting degree of the power function model is the highest,R^2 is 0.71,while the R^2 of the DBH-crown-tree height binary primary model is 0.81.The T test was carried out after the image extraction value was substituted into the model and compared with the actual measured value.The test result indicates a good model fitting degree.The study showed that the object-oriented method is better to extract the crown information of P.schrenkiana var tianshanica forest;the method that constructing model based on simulated data and measured data to obtain DBH and then retrieve forest storage can reduce the cost of manpower investigation and meet the needs of management of P.schrenkiana var tianshanica forest.