Research Of Forestry Geometrical Parameter Extraction With Light And Small Airborne Remote Sensing System

As a new remote sensing platform, Combining Laser scanning rangefinder, digital camera and POS (Posture Observation System) together, light and small airborne remote sensing system can obtain high density LiDAR point cloud data and high spatial resolution aerial remote sensing image of study area with laser detection and ranging technology and digital photogrammetry technology. The purpose of this study is to model and estimate forestry parameters by building aerial stem volume table with ground survey data, processing and analyzing point cloud data and aerial image of light and small airborne remote sensing system. The main conclusions include:(1) The DBH(diameter at breast height) estimated model of Chinese red pine (Pinus massoniana) base on filed survey data has high precision.Base on sample plot survey data,5candidate DBH-K(Crown diameter) and10candidated DBH-H(tree height) were chosen to build the optimal DBH-K, DBH-H, and DBH-K, H model to estimated DBH from crown diameter and tree height of Chinese red pine, the highest determination coefficient was0.615. In addition, one and two variable timber volume table of Chinese red pine in Shangcheng area were referenced to build one and two variable timber volume aerial table, which will help to estimate forestry parameters base on LiDAR data and high spatial resolution aerial images from light and small airborne remote sensing.(2) The light and small airborne LiDAR data can extract forestry parameters such as average stand high, individual tree high and stand density effectively.With the spatial analysis function of GIS, airborne LiDAR point cloud data were used in extracting the average stand high, individual tree high and stand density, and compare with filed survey data, the estimated accuracy of average stand high and individual tree high are83%and88%respectively, and compared with visual interpretation the accuracy of tree number extracted from LiDAR data is84%.(3) The high spatial resolution aerial image of light and small airborne remote sensing can extract forestry parameters such as crown density, stand average tree crown and individual tree crown effectively.In the full study of spectral and texture features of high-resolution aerial remote sensing image of light and small airborne remote sensing, base on object-oriented classification method and multi-scale segmentation, forest and non-forest area can be extracted firstly, and then forestry parameters such as crown density, stand average tree crown and individual tree crown can be extracted effectively. The accuracy of individual tree crown is72%, while the average tree crown reaches87%. (4) Estimating forest parameters by integrated application of the LiDAR data and high spatial resolution aerial images data.Integrated the advantages of airborne LiDAR data and its synchronization high spatial resolution aerial image, tree high and forest density can be extracted from LiDAR data which contain the three-dimensional coordinates, tree crown diameter, canopy density can be extracted from aerial image with a0.05m high spatial resolution, then individual tree volume and forestry stand volume can be estimated and modeled with the parameters come from light and small airborne remote sensing system. The estimated accuracy of individual tree volume base on one variable timber volume aerial table is51%, and57%while base on two variables timber volume aerial table. A multiple linear regression model estimated forestry stand volume with average tree high, forest density, average crown and canopy density as dependent variables has a correlation coefficient of0.583, and the determination coefficient is0.34.Above all, the accuracy of forestry stand and individual tree parameters such as average tree height, average crown diameter, individual tree high and crown diameter estimated from airborne LiDAR data and high spatical resolution aerial images are between72%-88%, the accuracy of individual stem volume are51%and57%respectively base on one and two variable timber volume aerial tables. The multiple linear regression model estimated forestry stand volume base on the forestry parameters estimated from light and small airborne remote sensing data has a correlation coefficient of0.583. As the limit of modeling data, more filed survey and verification test needed to find the application area. Further research need to be done to find out how to make full use of the light and small airborne remote sensing data, dig the inner knowledge of LiDAR data and high spatial resolution aerial images, and improve the estimated accuracy of individual tree and forestry stand parameters.