Research On The Method Of Forest Canopy Height Retrieval Based On ICESat-2 Lidar Photon Cloud Data

The ice,cloud and land elevation satellite-2(icesat-2)has been launched in September 2019.The satellite is equipped with advanced terrain altimetry system and a new micro pulse multi beam photon counting lidar,which can widely collect the elevation data of global ice cover,grassland,land water and forest.Icesat-2 can measure the global canopy height,but due to the characteristics of photon counting lidar,its photon cloud data is greatly affected by noise photons,the signal-to-noise ratio is related to scanning time,and the photon distribution density is not uniform,which needs to be used after the precise denoising and classification of the data.At present,most of the de-noising and classification algorithms developed are based on high-density airborne lidar data,which can not be well applied to icesat-2 data.Based on the above problems,this paper proposes an improved adaptive denoising algorithm to achieve the removal of noise photons,and through the peak detection based photon classification algorithm to achieve the classification of ground and canopy photons,and finally achieve the extraction of forest canopy height.In this paper,two ecological regions in Alaska are used as research areas,and icesat-2 Level 2 data is used as test data to complete the denoising and classification algorithm experiments of photon cloud data,and retrieve the forest canopy height information according to the classified data.At the same time,high-precision g-liht(Goddard's lidar,Hyperspectral and thermal imager)airborne lidar data are used as validation data to evaluate the effectiveness of the algorithm and the accuracy of canopy height inversion.The experimental results show that the denoising algorithm in this paper can remove most of the noise photons.The average accuracy of the denoising algorithm is 0.93,and the average F1 score is 0.92.Through the photon classification algorithm,the ground photon and canopy photon can be classified,and the noise photon can be further removed in the classification algorithm stage.The evaluation results show that the RMSE of ground line and airborne verification data extracted by the photon cloud classification algorithm is in the range of 1-2m.After the classification of signal photons is obtained,the relative canopy height is extracted according to the ground line fitted by ground photons,and the maximum canopy height,average canopy height,rh90,rh75,Rh50 are obtained,The results show that the distribution of the maximum canopy height and the average canopy height extracted by lidar is very close to the actual distribution.The average canopy height RMSE is 2.8 m,the maximum canopy height RMSE is 7 m,rh90,rh75,Rh50,RMSE of four percentage canopy height parameters of RH25 were 3.2m,2.6m,2.3m and 1.9m respectively.The highest correlation between the percentage canopy height parameter rh75 and the same percentage canopy height parameter on board is 0.64.In general,through the denoising and classification algorithm in this paper,the correct recognition of signal photons and the determination of photon category can be realized,and the canopy height parameters of forest can be accurately extracted from the photon cloud data.The extracted forest parameter results also show that the space-based lidar can realize the estimation of regional forest parameters.