Abstract: [Objective] According to the characteristics of the discrete point cloud in forest area, canopy height model(CHM) was built through different interpolation methods. The results of the different interpolation methods were compared, analyzed and evaluated in order to provide the reference for choice of interpolation methods. [Method] In this study, the discrete point cloud data in plots(30 m×30 m) were used as the experimental data. CHMs were generated by B-Spline, triangulation with linear interpolation(TLI), ordinary Kriging(OK) and inverse distance weighted(IDW) interpolation methods, respectively, through the open source software SAGS-GIS. 2D views, 3D views, profiles and pixel statistics of CHMs with different interpolation methods in plots were compared and analyzed. At the same time, the search radius parameters of IDW interpolation were discussed, compared and analyzed.[Result] The spatial distribution was uniform and there was height mutation for the discrete point cloud in forest area. For B-Spline interpolation, zero value(no data) region was filled, canopy gap was grossly filled and pixel maximum of the CHM deviated significantly from the height value of the original data. For TLI interpolation, the CHM appeared to be more fragmentation. For OK interpolation, the image of CHM was not clear duo to excessive smoothing applied. And for the IDW interpolation, the CHM on the top of canopy was properly filled and smoothed, but canopy edge was not excessive smoothing and retained elevation mutation, meanwhile, canopy gap still retained not be over filled. The results showed that the most suitable search radius of IDW interpolation method was about 1.5±2.5 times of the mean interval of original point cloud in forest canopy. [Conclusion] The IDW interpolation was better than B-spline, TLI, OK interpolation for generating CHM from discrete point cloud data. The formation of CHM with IDW interpolation could accurately reflect the truly natural form of forest canopy. So it was good for the extraction of forest parameters.

Key words: light detection and ranging(LiDAR), canopy height model(CHM), inverse distance weighted(IDW), B-spline, triangulation with linear interpolation(TLI), ordinary Kriging(OK), discrete point cloud