Method, system and equipment for vegetation restoration or rehabilitation of simulating natural ecosystem based on machine learning

Patent metadata

Field	Value
Publication number	US-2023123790-A1
Application number	US-202217967907-A
Country	US
Kind code	A1
Filing date	Oct 18, 2022
Priority date	Oct 20, 2021
Publication date	Apr 20, 2023
Grant date	—

Abstract

A method for vegetation restoration or rehabilitation of simulating a natural ecosystem based on machine learning (ML) includes: acquiring historical growth environment data of a vegetation community; extracting a site condition feature and a growth condition feature of each vegetation species; restoring and rehabilitating a vegetation community structure selection model; and selecting, based on the vegetation community structure selection model, an optimal vegetation species according to current growth environment data of the vegetation community, and restoring and rehabilitating a vegetation community simulating a natural ecosystem. The method comprehensively considers various factors affecting vegetation restoration based on the site condition and the growth condition. The method has the advantages of simple operation, fast modeling speed, high calculation efficiency, and high screening accuracy and can realize accurate and effective vegetation restoration to improve the quality of the ecological environment.

First claim

1 . A method for a vegetation restoration or rehabilitation, the method being based on machine learning (ML) and comprising:
acquiring historical growth environment data of a vegetation community; extracting a site condition feature and a growth condition feature of each vegetation species from the historical growth environment data; restoring and rehabilitating a vegetation community structure selection model based on the site condition feature and the growth condition feature of the vegetation species, specifically comprising the following sub-steps:
acquiring growth condition features of a meteorological environment where the vegetation species is located and selecting a most correlated growth condition feature of the vegetation species;
acquiring site condition features of a geographical environment where the vegetation species is located and selecting a most correlated site condition feature of the vegetation species;
establishing, based on the selected most correlated growth condition feature and site condition feature of the vegetation species, a sample database of a one-to-one correspondence; and
training, based on the sample database, the vegetation community structure selection model by an ML method, and determining a correspondence between the growth condition feature and the vegetation species and a correspondence between the site condition feature and the vegetation species; and
selecting, based on the vegetation community structure selection model, an optimal vegetation species according to current growth environment data of the vegetation community, and restoring and rehabilitating the vegetation community to simulate a natural ecosystem.

2 . The method according to claim 1 , wherein the site condition feature of the vegetation species comprises a slope aspect, a slope gradient, a slope position, and a micro-topography of the geographical environment where the vegetation species is located;
the slope aspect of the geographical environment where the vegetation species is located specifically comprises one of sunny slope, semi-sunny slope, semi-shady slope, and shady slope; the slope gradient of the geographical environment where the vegetation species is located specifically comprises one of slight slope, gentle slope, abrupt slope, steep slope, sharp slope, and dangerous slope; the slope position of the geographical environment where the vegetation species is located specifically comprises one of ridge slope, upper slope, middle slope, lower slope, and valley slope; and the micro-topography of the geographical environment where the vegetation species is located specifically comprises one of gentle platform, subsidence, long gully, shallow gully, and scarp.

3 . The method according to claim 2 , wherein the growth condition feature of the vegetation species comprises sunshine hours, soil moisture, air temperature, air pressure, and relative humidity of the meteorological environment where the vegetation species is located.

4 . The method according to claim 1 , wherein the step of acquiring the growth condition features of the meteorological environment where the vegetation species is located and selecting the most correlated growth condition feature of the vegetation species specifically comprises the following sub-steps:
acquiring the growth condition feature of the meteorological environment where the vegetation species is located and constructing a growth condition feature matrix; setting a growth condition threshold and constructing a growth condition threshold matrix; calculating a correlation coefficient between each growth condition feature and the corresponding growth condition threshold; calculating, based on the calculated correlation coefficient, a correlation order of each growth condition feature; and sorting by the calculated correlation order and selecting a growth condition feature with a maximum correlation order.

5 . The method according to claim 4 , wherein the step of acquiring the site condition features of the geographical environment where the vegetation species is located, and selecting the most correlated site condition feature of the vegetation species specifically comprises the following sub-steps:
acquiring the site condition feature of the geographical environment where the vegetation species is located and constructing a site condition feature matrix; setting a site condition threshold and constructing a site condition threshold matrix; calculating a correlation coefficient between each site condition feature and the corresponding site condition threshold; calculating, based on the calculated correlation coefficient, a correlation order of each site condition feature; and sorting by the calculated correlation order and selecting a site condition feature with a maximum correlation order.

6 . The method according to claim 5 , wherein the step of training, based on the sample database, the vegetation community structure selection model by the ML method, and determining the correspondence between the growth condition feature and the vegetation species and the correspondence between the site condition feature and the vegetation species specifically comprises the following sub-steps:
acquiring sample data in the sample database as a training sample; normalizing the sample data; setting a number of hidden layers; initializing an evolution count, a population size, a crossover probability, and a mutation probability; encoding a population in a real number, and taking an error between predicted data and expected data as a fitness function; cycling through a selection, a crossover, a mutation, and a fitness calculation until the evolution count is reached, to acquire an optimal initial weight and an optimal threshold; taking the acquired optimal initial weight and the optimal threshold as a weight and a threshold between initial neurons of a neural network, and restoring and rehabilitating the vegetation community structure selection model; and training, by the training sample, the vegetation community structure selection model, and constructing the correspondence between the growth condition feature and the vegetation species and the correspondence between the site condition feature and the vegetation species.

7 . A system for a vegetation restoration or rehabilitation, the system being based on machine learning (ML) and comprising:
a data acquisition module configured to acquire historical growth environment data of a vegetation community; a data extraction module configured to extract a site condition feature and a growth condition feature of each vegetation species from the historical growth environment data; a model constructing module configured to restore and rehabilitate a vegetation community structure selection model based on the site condition feature and the growth condition feature of the vegetation species, specifically to
acquire growth condition features of a meteorological environment where the vegetation species is located and select a most correlated growth condition feature of the vegetation species;
acquire site condition features of a geographical environment where the vegetation species is located and select a most correlated site condition feature of the vegetation species;
establish, based on the selected most correlated growth condition feature and site condition feature of the vegetation species, a sample database of a one-to-one correspondence; and
train, based on the sample database, the vegetation community structure selection model by an ML method, and determine a correspondence between the growth condition feature and the vegetation species and a correspondence between the site condition feature and the vegetation species; and
a vegetation community constructing module

Assignees

• China Inst Water Resources & Hydropower Res

Inventors

• Liu Guangquan
• Shi Wanli
• Liu Liu
• Qi Chunyu
• Bai Yingfei
• Li Puhang
• Zhang Qiang
• Chen Yonggui
• Fu Heng

Classifications

• Y02A40/22

  Improving land use; Improving water use or availability; Controlling erosion · CPC title

• G06F30/13 Primary

  Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads · CPC title

• G06F17/16

  Matrix or vector computation {, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization (matrix transposition G06F7/78)} · CPC title

• G06N3/08

  Learning methods · CPC title

• G06F30/27 Primary

  using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model · CPC title