Forest Parameters Estimation Using TLS And Airborne Image Lidar

Ouantitative description of forest structure to help the investigation, management and planning of forest resource, and contribute to the assessment of forest biomass. LiDAR’s imaging mechanism is different from that of passive optical remote sensing, it can provide high-precision and three-dimensional structural information, and with the strong ability to detect vegetation spatial structure and topography. In the quantitative measurement of forest parameters, using TLS to measure forest parameters not only saves manpower, but also improves work efficiency, which has become an effective way method for quick access to the tree geometry parameters now.This paper tries to study the extraction of forest parameters based on Terrestrial Laser Scanner (TLS) and Airborne Image LiDAR. The main results are as follows:1. Registration of TLS data and extraction of DEM. Based on the plot reflectors from total station to register multi-stations scan data into the global coordinate system, the RMSE (Root Mean Square Error) of registration is in the range of±2mm. To distinct the ground points and none ones from points cloud in TerraSolid, and generate the DEM of5cmx5cm with those ground points. The results show that it is effect to extract high-precision DEM of forest areas from TLS data.2. Extraction of a single tree’s parameters based on TLS. To slice points cloud horizontally and generate multi-bands grid image of horizontal stratifications. By setting three different detection thresholds to run the Hough Transform of grid image to get the initial results, and then detecting their points cloud amount in vertical0.3m-0.6m above the ground respectively. To extract locations and DBHs (Diameter at Breast Height) from the results of Hough Transform and points cloud amount detection, and extract height using the difference value of the highest and lowest points cloud in vertical cylinder. The accuracy of single trees extraction is up to74.71%, the R2and RMSE of determination for DBHs are0.6648and1.217cm, and those for heights are0.782and0.604m respectively.3. Extraction of canopy densitys or coverages based on TLS. To extract DSM (Digital Surface Model) from all points cloud and combine DSM with DEM to generate CHM (Canopy Height Model), and then extract canopy densitys or coverages by setting different height thresholds of ROIs (Regions of Interest) in CHM. The R2and RMSE of six tree plots are0.8966and0.073, and those of twenty-four other plots are0.5226and0.16respectively.4. Extraction of canopy densitys or coverages based on Image LiDAR. Based on the DEM from TLS to complete geography calibration of the DSM obtained by distance solver from Image LiDAR, and integrate DSM and DEM from TLS to generate CHM. The extraction methods of canopy densitys or coverages are same with those of TLS. The R2and RMSE of five tree plots are0.7409and0.1269, and those of fourteen other plots are0.1602and0.2735respectively. The geometric distortion of Image LiDAR’s images is one of the reasons result in the lower extraction precision.The innovations of the thesis are as follows:1. The pre voxelization processing of raster image before Hough transform detection.2. Joining TLS and Image LiDAR for the extraction of forest parameters.