	US 12,276,731 B2
	Method of individual tree crown segmentation from airborne LiDAR data using novel gaussian filter and energy function minimization
Ting Yun, Nanjing (CN); Kang Jiang, Nanjing (CN); Guangchao Li, Nanjing (CN); Yiduo Li, Nanjing (CN); and Lin Cao, Nanjing (CN)
Assigned to NANJING MAOTING INFORMATION TECHNOLOGY CO., LTD., (CN)
Appl. No. 17/780,538
Filed by NANJING MAOTING INFORMATION TECHNOLOGY CO., LTD., Nanjing (CN)
PCT Filed Sep. 30, 2020, PCT No. PCT/CN2020/119150 § 371(c)(1), (2) Date May 27, 2022, PCT Pub. No. WO2022/067598, PCT Pub. Date Apr. 7, 2022.
Prior Publication US 2023/0350065 A1, Nov. 2, 2023
Int. Cl. G01S 17/89 (2020.01); G06V 20/10 (2022.01)

CPC G01S 17/89 (2013.01) [G06V 20/182 (2022.01)]

4 Claims

1. A method of individual tree crown segmentation from airborne LiDAR data using a novel Gaussian filter and energy function minimization, comprises the following steps of:

1): generating a DSM from airborne LiDAR data for a forest plot by a rasterization method;

2): filtering the DSM based on a dual Gaussian filter;

3): detecting candidate treetops from the filtered DSM using a local-maximum-finding method, and employing a screening strategy based on an angle threshold to exclude false treetops; and

4): conducting an accurate delineation of tree crowns, after the false treetops are excluded, by a water expansion model controlled by an energy function, along with an estimation of crown width in two perpendicular directions, wherein the energy function is as follows:

the DSM of each forest plot is vertically and evenly stratified into many height interval h_j, j=1, 2, 3 . . . in top-down order based on a fixed spacing; in equation (2), c_i,d^brepresents a boundary cell of water expansion in b th triangle of a distance d to boundary cells of a previous height interval h_j-1; cells c_d, d=1,2,3 . . . together form a concentric contour structure through water expansion with the convergent centre of treetop t_k, satisfying a connectivity principle; c_i,d^b,zrepresents height of the cell c_i,d^b;

represents the vector of a maximum gradient of the boundary cell c_i,d^bon a DSM C^zand

represents the vector from a corresponding treetop location t_kto the boundary cell c_i,d^b; arccos

represents an included angle between a gradient vector of the boundary cell C_i,d^band the vector from a starting point of water expansion with the convergent centre of treetop t_kto the boundary cell c_i,d^b; Q_b^krepresents a number of boundary cells c_i,d^bin the b th triangle expanded from the convergent centre of treetop t_kat the distance d;

represents height differences between an average height value of all water expanded boundary cells c_i,d^band three surrounding treetops t_c_{_i}_b,lthree vertices of the b th triangle; c_d^b,zrepresents the average height value of the boundary cell c_i,d^b, t_c_{_i}_^b_,l^zrepresents heights of three surrounding treetops of the boundary cell c_i,d^b; and the coefficients α≈0.1 and

where c_d^b,zdenote a monotonic decrease in the average height value of all expanded boundary cells c_i^bin each height interval with increasing distance d.