| Forests are the main body of terrestrial ecosystem.They can not only provide the materials for human survival and development,but also have a variety of ecological functions.They play an irreplaceable role in maintaining balance between the land and the global ecological and biodiversity.It’s of great importance for monitoring the dynamics of forestry resources information.The traditional methods of field investigation of forest resource can obtain the information data,but be a time-consuming and costly process.With the characters of macroscopic,effective,dynamic,low cost,periodic short,informative,the remote sensing become the main means of forest resources survey.Lighting Detection And Ranging(LIDAR)as an advanced active remote sensing technology,which can directly acquire three-dimensional coordinate information of earth target objects,so the application of Li DAR datas have a distinctive advantage in the field of forestry.In this paper,we used the LiDAR point cloud data as the data source.Filtering and classification of the LiDAR data is one of the focus in this research.Based the ArboLiDAR software,creating the forest height and density raster,which need filter and composite bands.Forest stand automatic segmentation and calculating different LiDAR variables had been done with the tools of the ArboLi DAR before creating the regression model to estimate the forest stand variables from the Li DAR variables.In the end,used the estimation results and the biomass growth equation to estimate the stand biomass,and evaluated the estimation accuracy of the results.Research conclusions are as following:(1)The filtering and classification accuracy of the LiDAR point cloud data as a key to the quality of the DTM,also directly determined the elevation normalized results of LiDAR point cloud.(2)Before doing the automatic forest stand segmentation with the ArboLiDAR software,we have filtered the rasters with the median filter method and mean-shift filter method to smoother the raster appearance,which lay a good foundation for the follow-up forest stand segmentation.Selecting the appropriate segmentation parameters and merge parameters values can be more desirable to get the satisfactory stand segmentation results.(3)The regression analysis of LiDAR variables with the measured parameters of field reference plots was performed.The estimation results showed that: the estimation accuracy ofaverage HGW of stand is the highest,with R2 of 0.860 and the average estimate accuracy of90%;The estimation accuracy of average DBH was only next to average HGW;Estimation accuracies of stand volume and stand basal area were greatly influenced by the forest stand density and age structure and the average estimate accuracy both were 80%;Estimation accuracy of the stand density was lowest,with R2 of 0.708,an average estimate of the accuracy of 74%.(4)Forest stand biomass was estimated with the allometric growth equation and estimated stand density and stand volume combining.The average estimation accuracy was90% with R2 of 0.724 and easily influenced by stem density and stand volume but higher than both them. |
