CN-121982566-A - Time sequence remote sensing index-based hemp bamboo forest spatial distribution identification method

CN121982566ACN 121982566 ACN121982566 ACN 121982566ACN-121982566-A

Abstract

The invention relates to the technical field of the spatial distribution identification of the bamboos, and discloses a method for identifying the spatial distribution of the bamboos based on a time sequence remote sensing index, which comprises the steps of constructing a time sequence remote sensing data set of the growth of the bamboos based on a high-resolution remote sensing image; the method comprises the steps of obtaining a standardized remote sensing image sequence capable of being used for time sequence analysis, respectively extracting spectral reflectivity and vegetation index values of the hemp-bamboo forest land sampling points and similar forest land sampling points in a plurality of growth stages based on the standardized remote sensing image sequence, constructing time sequence remote sensing indexes reflecting seasonal variation amplitude of the hemp-bamboo growth based on all characteristic values, inputting the time sequence remote sensing indexes and basic spectral characteristics as classification characteristics into a classification model, and identifying and extracting the spatial distribution of the hemp-bamboo forest. The invention obviously improves the distinguishing capability of the hemp bamboo forest and similar forest lands.

Inventors

SHU HUILIN
DING CHEN
QI LIANGHUA
LI XIAOMIN
LEI LI
LI XIANWEI
SU CHUNMEI
SONG HANQING
WU JIAJIE
LI YIYANG

Assignees

国际竹藤中心

Dates

Publication Date: 20260505
Application Date: 20260122

Claims (10)

1. A method for identifying the spatial distribution of a hemp bamboo forest based on a time sequence remote sensing index is characterized by comprising the following steps: Acquiring a high-resolution remote sensing image of a research area in a complete growth period, and constructing a time sequence remote sensing data set of the growth of the phyllostachys praecox based on the high-resolution remote sensing image; performing terrain correction, cloud mask processing and spatial resolution unification on the time sequence remote sensing dataset to obtain a standardized remote sensing image sequence capable of being used for time sequence analysis; based on the standardized remote sensing image sequence, respectively extracting spectral reflectivity and vegetation index values of the hemp and bamboo forest land sampling points and similar forest land sampling points in a plurality of growth stages; screening the base remote sensing characteristics capable of representing the characteristics of the bamboos by comparing the spectrum change difference and the index change difference of the bamboos plot sample points and the similar forest plot sample points in different growth stages; Based on the base remote sensing characteristics, respectively calculating characteristic values of the base remote sensing characteristics corresponding to the relative dormancy periods in the growing season of the bamboos, and constructing a time sequence remote sensing index reflecting the seasonal variation amplitude of the growing of the bamboos based on all the characteristic values; And inputting the time sequence remote sensing index and the basic spectral features as classification features into a classification model, and identifying and extracting the spatial distribution of the hemp and bamboo forest.
2. The method for identifying the spatial distribution of the basswood forest based on the time sequence remote sensing index according to claim 1, wherein when the time sequence remote sensing data set is subjected to terrain correction, cloud mask processing and spatial resolution unification, the method comprises the following steps: The method comprises the steps of performing terrain correction on a time sequence remote sensing data set by adopting an SCS+C terrain correction model, generating a cloud mask based on a quality control wave band of a remote sensing image, removing cloud coverage pixels, resampling wave bands with different spatial resolutions, and unifying the wave bands to the same spatial resolution.
3. The method for identifying the spatial distribution of the basketry based on the time-series remote sensing index according to claim 2, wherein when determining the basketry site sampling points and the similar forest site sampling points, the method comprises the following steps: Determining a research area, collecting a national forest resource checking distribution map and the high-resolution remote sensing image, and determining a bamboo forest distribution area in the research area and a similar forest land area with confusion with the hemp bamboo forest in remote sensing characteristics based on the national forest resource checking distribution map and the high-resolution remote sensing image; According to a statistical sampling principle, taking a hemp-bamboo forest distribution area and a similar forest land distribution area in the research area as a sampling overall respectively, and determining the number of sampling points according to the corresponding distribution area; Distributing sampling points in each sampling assembly in a random or layered random mode; And randomly dividing the hemp bamboo forest sampling points and similar forest land sampling points respectively, wherein 70% of the sampling points in each type of sampling points are used as training sampling points, and the rest 30% of the sampling points are used as verification sampling points.
4. The method for identifying the spatial distribution of the bastose based on the time sequence remote sensing index as claimed in claim 3, wherein when determining the number of the sample points according to the corresponding distribution area, the method comprises the following steps: Analyzing the nationwide forest resource checking distribution map and the high-resolution remote sensing image to obtain the land distribution complexity corresponding to the bamboo forest distribution area and the similar forest land area; Constructing a sample point feature vector according to the distribution area and the distribution complexity of the land class; Comparing the sample characteristic vector with a historical sample number set, and determining the sample number according to a comparison result; When the historical sample point feature vector which is the same as the sample point feature vector exists in the historical sample point number set, taking the historical sample point number corresponding to the historical sample point feature vector as the sample point number; And when the historical sample point feature vector which is the same as the sample point feature vector does not exist in the historical sample point number set, determining the sample point number according to the sample point feature vector.
5. The method for identifying the spatial distribution of the bamboos based on the time-series remote sensing index according to claim 4, wherein when determining the number of the sample points according to the feature vector of the sample points, the method comprises the following steps: comparing the distribution area with a distribution area threshold, comparing the ground distribution complexity with a ground distribution complexity threshold, and determining the number of the sample points according to a comparison result; when the distribution area is greater than or equal to the distribution area threshold and the ground type distribution complexity is greater than or equal to the ground type distribution complexity threshold, determining the number of the sampling points as a first number; when the distribution area is greater than or equal to the distribution area threshold and the ground type distribution complexity is less than the ground type distribution complexity threshold, determining the number of the sampling points as a second number; When the distribution area is smaller than the distribution area threshold and the ground type distribution complexity is greater than or equal to the ground type distribution complexity threshold, determining the number of the sampling points as a third number; And when the distribution area is smaller than the distribution area threshold and the ground type distribution complexity is smaller than the ground type distribution complexity threshold, determining the number of the sampling points to be a fourth number.
6. The method for identifying the spatial distribution of the bastose based on the time sequence remote sensing index according to claim 5, wherein when the spectral reflectances and vegetation index values of the bastose land sample points and the similar land sample points in a plurality of growth stages are respectively extracted based on the standardized remote sensing image sequence, the method comprises the following steps: Acquiring remote sensing image data covering a plurality of time phases of the research area in the standardized remote sensing image sequence; respectively extracting pixel reflectivity values of the tingling forest land sample points and the similar forest land sample points on a preset spectrum band in remote sensing image data corresponding to each time phase, and carrying out statistical processing on the pixel reflectivity values to obtain spectrum reflectivity of each sample point on a corresponding spectrum band; Calculating vegetation index values of each sample point in corresponding time phases based on the spectral reflectivity; Dividing a plurality of growth stages according to the growth process of the bamboos, carrying out statistics and summarization on the spectral reflectivity and vegetation index values in the same growth stage, and forming a spectral reflectivity sequence and a vegetation index sequence of the bamboos forest land sampling points and similar forest land sampling points in the plurality of growth stages.
7. The method for identifying the spatial distribution of the bamboos based on the time sequence remote sensing index according to claim 6, wherein the method for screening the base remote sensing features capable of representing the characteristics of the bamboos by comparing the spectrum change difference and the index change difference of the local points of the bamboos with those of the similar woods in different growth stages comprises the following steps: Respectively constructing time sequence change curves corresponding to the tingling bamboo forest land sampling points and similar forest land sampling points in a plurality of growth stages based on the spectral reflectivity sequence and the vegetation index sequence; Based on the time sequence change curve, extracting change amplitude parameters, change trend parameters and peak occurrence time corresponding to each spectrum band and vegetation index value in different growth stages; according to the variation amplitude parameter, the variation trend parameter and the peak occurrence time, comparing and analyzing the corresponding characteristics of the hemp forest land sampling points and the similar forest land sampling points, and determining a candidate spectrum band or a candidate vegetation index; Aiming at the candidate spectrum band or the candidate vegetation index, respectively calculating class separation indexes corresponding to the tingling bamboo forest land sampling points and similar forest land sampling points in each growth stage; Comparing class separation indexes corresponding to each spectral band or vegetation index value in each growth stage, and determining the spectral band or vegetation index value meeting the screening conditions according to a preset characteristic screening rule; And determining a spectral band or vegetation index value meeting the screening condition as a base remote sensing characteristic for representing the characteristics of the dendrocalamus latiflorus weathers.
8. The method for identifying the spatial distribution of the hemp and bamboo forest based on the time sequence remote sensing index according to claim 7, wherein the comparing and analyzing the corresponding characteristics of the hemp and bamboo forest land sample points and similar forest land sample points according to the variation amplitude parameter, the variation trend parameter and the peak occurrence time comprises the following steps: summarizing the variation amplitude, variation trend and peak occurrence time corresponding to each spectral band or vegetation index in each growth stage to form a feature set; Respectively calculating the average value and variance of the corresponding variation amplitude, variation trend and peak value appearance time of the hemp-bamboo forest sample points and similar forest land sample points in each growth stage; Comparing the average value difference and variance range of the hemp bamboo forest sample points and the similar forest land sample points in each growth stage for the same spectral band or vegetation index; the characteristic that the difference of the average value of the dendrocalamus latiflorus sampling points and the similar forest land sampling points exceeds the sum of the respective variances in at least A growing stages is marked as a candidate spectrum band or a candidate vegetation index.
9. The method for identifying the spatial distribution of the bastose based on the time sequence remote sensing index according to claim 8, wherein the time sequence remote sensing index is obtained by the following formula: wherein, TSI represents the time sequence remote sensing index; xs represents the characteristic value corresponding to the growing season of the bamboos, and Xw represents the characteristic value corresponding to the growth relative to the dormancy period of the bamboos.
10. The method for identifying the spatial distribution of the bastose based on the time sequence remote sensing index according to claim 9, which is characterized in that the time sequence remote sensing index and the basic spectrum characteristic are used as classification characteristics to be input into a classification model, and when the identifying and the extracting of the spatial distribution of the bastose are carried out, the method comprises the following steps: acquiring basic feature combinations, and introducing red edge index features and time sequence remote sensing indexes to form a plurality of groups of feature combination schemes for comparison analysis; respectively taking each characteristic combination scheme as an input characteristic, combining the constructed training points, inputting a classification model to perform model training, and generating a corresponding bast fiber forest space distribution result based on the trained classification model; performing accuracy verification on the spatial distribution results of the hemp and bamboo forests corresponding to each characteristic combination scheme by utilizing the reserved verification points, and obtaining classification accuracy evaluation indexes corresponding to each characteristic combination scheme; comparing and analyzing the classification precision evaluation indexes of different characteristic combination schemes, determining the characteristic combination scheme with the highest classification precision in the spatial distribution identification of the bast fiber and marking the characteristic combination scheme as an optimal characteristic combination scheme; Based on the optimal feature combination scheme, analyzing the contribution degree of each feature factor in the hemp bamboo forest identification process by using a feature importance evaluation result of the classification model, and finishing final identification and extraction of the hemp bamboo forest spatial distribution according to the contribution degree.

Description

Time sequence remote sensing index-based hemp bamboo forest spatial distribution identification method Technical Field The invention relates to the technical field of hemp and bamboo forest spatial distribution identification, in particular to a hemp and bamboo forest spatial distribution identification method based on a time sequence remote sensing index. Background The bamboo forest is used as an important economic bamboo species and ecological resource, and the space distribution information of the bamboo forest has important significance for forestry resource management, ecological environment protection, regional economy planning, carbon circulation research and the like. The traditional method for identifying the spatial distribution of the bamboos relies on manual field investigation, and is time-consuming, labor-consuming, high in cost and difficult to realize large-scale, rapid and dynamic monitoring. Along with the development of remote sensing technology, the identification of vegetation types by using remote sensing images has become an important means. However, the remote sensing data of a single or a few time phases often has difficulty in capturing the physical characteristics of the bamboos in different growth periods, is easily influenced by cloud cover, atmospheric scattering, homoiotic spectrum, foreign matter homospectrum and other phenomena, and has low identification precision. In particular, the bastose and other bamboo species or broad-leaved tree species have certain similarity in spectral characteristics, and are difficult to effectively distinguish by only relying on spectral information of a single time phase. Therefore, it is necessary to design a method for identifying the spatial distribution of the hemp and bamboo forest based on the time sequence remote sensing index to solve the problems in the prior art. Disclosure of Invention In view of the above, the invention provides a method for identifying the spatial distribution of the bastose based on time sequence remote sensing indexes, which aims to improve the accuracy and efficiency of identifying the spatial distribution of the bastose by integrating the physical characteristics and the spectral information in multi-time sequence remote sensing data and realize the rapid response to the dynamic monitoring of the bastose in a large range. The invention provides a method for identifying the spatial distribution of a hemp bamboo forest based on a time sequence remote sensing index, which comprises the following steps: Acquiring a high-resolution remote sensing image of a research area in a complete growth period, and constructing a time sequence remote sensing data set of the growth of the phyllostachys praecox based on the high-resolution remote sensing image; performing terrain correction, cloud mask processing and spatial resolution unification on the time sequence remote sensing dataset to obtain a standardized remote sensing image sequence capable of being used for time sequence analysis; based on the standardized remote sensing image sequence, respectively extracting spectral reflectivity and vegetation index values of the hemp and bamboo forest land sampling points and similar forest land sampling points in a plurality of growth stages; screening the base remote sensing characteristics capable of representing the characteristics of the bamboos by comparing the spectrum change difference and the index change difference of the bamboos plot sample points and the similar forest plot sample points in different growth stages; Based on the base remote sensing characteristics, respectively calculating characteristic values of the base remote sensing characteristics corresponding to the relative dormancy periods in the growing season of the bamboos, and constructing a time sequence remote sensing index reflecting the seasonal variation amplitude of the growing of the bamboos based on all the characteristic values; And inputting the time sequence remote sensing index and the basic spectral features as classification features into a classification model, and identifying and extracting the spatial distribution of the hemp and bamboo forest. Further, when performing terrain correction, cloud mask processing and spatial resolution unification on the time-series remote sensing dataset, the method comprises the following steps: The method comprises the steps of performing terrain correction on a time sequence remote sensing data set by adopting an SCS+C terrain correction model, generating a cloud mask based on a quality control wave band of a remote sensing image, removing cloud coverage pixels, resampling wave bands with different spatial resolutions, and unifying the wave bands to the same spatial resolution. Further, in determining the bamboos plot and similar plots, the method comprises: Determining a research area, collecting a national forest resource checking distribution map and the high-resolution remote sensing image, and de