Study On The Method Of Extracting Forest Information From 3D UAV/RS

The forest in China has the characteristics of rich ecological resources, large area and so on, which have increased the difficulty for the investigation of forest resources. With the development of remote sensing technology and unmanned aerial vehicle technology, the investigation method of forest resources has changed from the traditional wood seized feet fieldwork to the global, indoor, large scale survey. This observation method can not only save resources and provides a reference for the national forest management planning formulation.In this paper, the Beijing Yanqing Songshan Nature Reserve is taken as the research region. The UAV data and remote sensing data are preprocessed respectively. Based on UAV data, sub compartment division of the forest is completed and the basicsed information of sub compartment tree measurement is extracted. The extracted basic information is used to establish the single tree model and stand forest-mensuration factors model. Based on the remote sensing image data, the forest volume inversion model is established. The precision analysis of two kinds of model with the experimental data is carried out. The results meet the accuracy requirements. The comparing analysis of two kinds of model shows that the accuracy of the UAV data model is obviously better than that of the remote sensing images data model. The results of the study can provide a reference for future research.The main research contents and conclusions are as follows:1、The UAV image is preprocessed by the SVS UAV image processing software. The high resolution DSM is obtained through the initial geometric correction and the aero triangulation process. The high resolution DEM can be got by the processing of DSM. The high resolution DOM is obtained through image mosaic and fusion.2、The obtained DEM and DSM data is used to extract slope, aspect and slope position information with ArcGIS software. Combined with information of forest type, sub compartment division of forest compartment is completed.3、The Ecognition software and ArcGIS software are used for the extraction of small class based information (tree crown and tree height) in UAV image respectively. Multi scale segmentation method based on Ecognition and the nearest neighbor classification method are used to extract canopy information. Using the spatial analysis tool of ArcGIS, the canopy model is established and the tree height information is extracted. The results are compared with the data from the survey of the field operation. The results of the accuracy verification meet the requirements.4、Uses multiple linear regression method to establish the single tree models of coniferous forest and broadleaf forest on the data from field investigation. Based on the correlation and error examination, the most reasonable model is selected. The diameter at breast height (DBH) information is obtained based on the model. Through using nine-tree sample plot investigation method, stand description factors model is established, including the average DBH and average height and stand density and forest stock volume model. The precision analysis shows the results meet the requirements of forest resource survey.5、The remote sensing image of satellite ZY-3 is used to extract the stock volume of Song Shan Nature Reserve. Using remote sensing image band information and topographic information to obtain modeling factors with high correlation. The coniferous and broadleaved forest volume models are established. And the models meet the precision requirements. The model is compared with the previous UAV model. Analysis shows that, compared with the measured data, the result through UAV model and nine-tree sample survey method has better accuracy than that of the result by using remote sensing image data.