Forest Parameters Inversion Using TerraScan Based On Multi-source Airborne Data

The forest is ecological system, which has the most widely distributed, largest area of land, the most complicated composition structure and the most abundant material resources. It is not only related to the development of social economy, but also has a tremendous influence to the ecological environment. In the early forest resources monitoring and forest parameters extraction, the single remote sensing image data provided by MSS, TM and other optical wide band passive remote sensing sensor was the most be application. However, in recent years, with the improvement of human environmental protection consciousness, the importance of forest resources to get more and more extensive considers. At the same time with the deepening of the research of forest resources, the remote sensing data, which obtained by traditional ways is not enough to meet the many forestry production and research needs. And the forest resources monitoring and survey urgent need to remote sensing technology to provide both qualitative and quantitative data. Hyperspectral remote sensing technology and laser radar remote sensing" technology came into being in this context. However the current domestic research is focused on a single Hyperspectral data, or is large footprints laser radar data, and as a single data sources, big laser flare and low sampling density causes study has not been high precisionIn this paper the cloud data of2009years of small multi-beam airborne laser radar airborne laser radar and its synchronization of the gain of Hyperspectral data in Heilongjiang province Yinchuan city cold water nature reserve was be used as data sources. And the high precision of forest parameter information was extracted (including single wood parameters). Specific content is listed below.Use of the TerraScan by the methods of setting threshold, irregular TIN algorithms, dynamic surface fitting algorithm and manual editing of LiDAR data filtering and classification of the height anomaly removed strips overlapping redundant data is removed, the point cloud data, theeffectively separated from the ground point and ground point; generate accurate digital surface model DSM and digital elevation model of the DEM, both for poor forest canopy height model CHM; and experience linear method to eliminate the atmospheric effects.Hyperspectral data filtering, calibration, and basic processing, after mutual alignment with the radar data, and use the the CHM mask processing of Hyperspectral data to eliminate the impact of non-woodland echo, the eventual adoption of SVM (support Vector Machinetheory) and SAM (spectral angle mapping) species classification, based on forest biomass and volume and other stand parameters (including the individual tree parameters) estimates, the accuracy reached above80%able to meet the needs of actual production. Of which:tree height estimation of the highest accuracy,97%, showed that the combination of a small spot laser radar data with Hyperspectral data, two remote sensing data to high-efficiency, high-precision offers a variety of stand parameters for forestry produce timely and accurate provide the necessary data to support.