Tree Height Estimation Based On Analyzing Factors For Dense Matching Of Aerial Photographs

This study integrated aerial photographs and airborne LiDAR data to calculate and monitor the forest canopy height,and study the influencing factors regarding to the dense matching.First of all,understory high accuracy digital elevation model and digital surface model were constructed based on LiDAR point cloud data.Digital surface models were then created by applying an automated stereo-matching algorithm utilizing different combination of aerial photographs.Then spatial resolution,heading angle,image coverage,terrain and shadow can be test how influence the dense matching.The canopy height was obtained by subtracting the LiDAR ground elevations from the DSM.Using historical aerial photographs of 1996,2004 and digital aerial photographs,LiDAR data of 2014,multi-temporal CHMs were reconstructed within a period of 18 years.(1)The R squared between the three types of canopy height extracted by SGM,PMVS,tSGM and corresponding LiDAR CHM is 0.71,0.76,0.64.(2)The R squared between the canopy height models acquired by LiDAR data and corresponding different combination of digital aerial photograph is 0.52,0.53,0.62,0.59,0.65 and the root mean square error is 1.3m,0.23 m,1.96 m,1.62 m,1.62 m respectively.(3)Compared with the actual data acquired from field plots,our data showed a measurement accuracy of 85.73% with maximum and mean absolute error of 3.53 m and 1.5 m(4)Combined with the aerial photos of year 1996,2004 and 2014,these multi-temporal canopy height models have a similar growth trend to the growth curve of Chinese fir.Different dense matching algorithm has similarcapability to measure the tree height using same aerial photographs.Based on the result,utilizing aerial photographs can reflect the variation of canopy height in the sunny slope of mountainous terrain.However,for forests located in the valley bottom,the canopy height would be under estimated by aerial photographs.Our quantitative analysis reflecting the variation of the canopy height can provide the possibility for monitoring forest growth and potentially model forest biomass and production using photogrammetric point cloud data.