Forest Information Extraction And Model Building Based On UAV Image

Survey and monitoring of forest resources is to understand the quantities, quality, distribution and changes of forest resources, consisted of office operation and field work. With the rapid development of communication technology, computer technology, remote sensing technology, the time and work intensity of field work is decreased and the automation and informatization of office operation is improved, the time-effectiveness and accuracy is also significantly improved. UAV (unmanned aerial vehicle) as an important platform for remote sensing data collection, application in investigation and monitoring of forest resources are increasingly being widely used.In this study, the UAV image as remote sensing data, produce high precision digital surface model (DSM) and digital orthophoto map (DOM), using object oriented image information extraction technology and GIS spatial analysis technology, and the tree crown, tree height, canopy density, stand density information is extracted, combined with field survey data, the extraction results are compared and analyzed, the establishment of aerial standing timber volume table. The main research contents and conclusions are as follows:(1)The production of digital surface model and high precision digital orthophoto mapThe selection of PixelGrid-UAV aerial image processing platform, processing of UAV original images with GPS/IMU data, after distortion correction, connection point extraction and adjustment, empty three encryption, image matching, orthophoto correction step, produce digital surface model and digital orthophoto map with 0.1m accuracy.(2)Using the object oriented classification method and multi-scale segmentation technique to extract the canopy and canopy density informationThe use of high precision digital orthophoto map, using object oriented classification method and multi-scale segmentation technology, selection of suitable segmentation scale parameters and features, first extract to obtain the target stand, then two times for segmentation and classification of crown and canopy information obtained. In order to estimating single tree canopy, canopy density and number density parameter. The use of object-oriented classification method and multi-scale segmentation technology, single tree average crown extraction precision is about 0.70; the extraction of canopy density value and human-computer interaction interpretation results are compared, the accuracy can reach more than 0.90, and in line with the basic field visual results; the overall accuracy of number density estimation for 0.81.(3) Using GIS spatial analysis technology to estimate individual tree height and stem density parameterUsing DSM and DOM interactive editing mode, ground elevation point, using the Kriging interpolation method, to produce the digital elevation model, then obtain a digital canopy model (CHM). The use of GIS analysis of neighborhood analysis and topological screened the treetops point, finally get the single tree height and stem density. In the non-closed forest, individual tree height estimation accuracy can reach more than 0.86; in the high, medium, low canopy density of stand, the overall accuracy of number density estimation for 0.67.(4)The establishment of aerial DBH-crown model, DBH-crown-tree height model and aerial one variable or two variable standing timber volume tableThe use of UAV image extraction of crown diameter and tree height parameter information, combined with field survey data, establish high model aerial DBH-crown model and the DBH-crown-tree height model, R2 up to 0.67.Comparative study of the one variable or two variable standing timber volume table of study area, aerial one variable or two variable standing timber volume table are established respectively.To sum up, by the use of high precision digital surface model and digital orthophoto map produced by UAV image, combined with object oriented classification method, the segmentation technology and spatial analysis technology of multi scale, in the extraction of tree crown width, tree height, canopy density, number density of forest parameter information, the overall accuracy could more than 0.70. The application of the research on the UAV image in forest resources investigation made some attempts, discusses the production of forest, DOM, DEM and DSM information extraction method, and provides a way for other forest information extraction.