A method for extraction of coniferous forest information

Abstract

The application relates to remote sensing image processing technology and a spatio-temporal data analysis method, and discloses an extraction method for coniferous forest information. By analyzing the best distinguishing phase of normalized difference vegetation index (NDVI), difference vegetation index (DVI), ratio vegetation index (RVI), soil-adjusted vegetation index (SAVI) and land surface water index (LSWI) of the coniferous forest, different vegetation indexes of different time phases are combined, a normalized forest index (NDFI) for distinguishing the coniferous forest and other forest lands is constructed, monthly average spectral curves are obtained by using time sequence data of the vegetation index, threshold values for extracting coniferous forest area information are obtained by performing probability density curve analysis on the normalized vegetation index NDFI, and high-precision coniferous forest information extraction and spatial distribution mapping based on satellite-borne satellite remote sensing data are realized. The new vegetation index extracted by the application enhances the difference between the vegetation, is beneficial to the distinction of the vegetation, can be used for extracting coniferous forest distribution information at a regional scale, and has higher accuracy and practicability.

Description

I. TECHNICAL FIELD

[0001] The present application relates to remote sensing image processing technology and spatio-temporal data analysis method, and particularly to a method for extracting coniferous forest based on a new vegetation index, belonging to the technical field of remote sensing image processing and forestry application. II. TECHNICAL BACKGROUND

[0002] Coniferous forests are widely distributed in the world, mainly in the northern hemisphere. In China, according to the distribution environment and flora composition, they can be divided into cold temperate, temperate, subtropical and tropical coniferous forests. They are mainly distributed in Northeast China, North China, Northwest China, Southwest China, tropical rainforests in the south, and Taiwan, etc.

[0003] Under natural conditions, various types of coniferous forests are quite stable and have strong natural regulation ability, especially good at resisting drought, strong winds and cold climate. They are the main component of China's forest resources, accounting for 65% of the total forest stock volume in China. Wild plants for food and medicine also grow in coniferous forests. Coniferous forests also play a very important role in carbon cycling. Extracting coniferous forest information at the regional scale plays a very important role in monitoring, protecting and preventing pests and diseases of coniferous forests. Compared with large-scale coniferous forest information extraction, regional-scale coniferous forest information extraction can exclude other interference factors, more accurately and comprehensively reflect more relevant information, better protect coniferous forests, and provide relevant basis for relevant departments to make decisions.

[0004] Accurate monitoring of vegetation structure and spatial distribution information is of great practical significance for changes in forest ecosystems and policy making. Traditional census and survey methods are time-consuming and difficult to accurately reflect the spatial distribution of vegetation. Satellite remote sensing technology, with its advantages of rapid and large-scale repeated data acquisition, has become an effective means of monitoring vegetation information, and many monitoring methods have been developed. Multi-temporal remote sensing analysis methods are widely used. Different vegetation has different growth cycles, so the phenological characteristics play an important role in vegetation identification. Vegetation index reflects the response characteristics of vegetation to spectral bands. The vegetation index of crops is different at different growth stages due to different pigments and structures; different vegetation indices have different emphases and reflect different vegetation; the vegetation index of different vegetation changes throughout the growth period. Therefore, the combination of multiple vegetation indices with time series information can distinguish the growth period and growth state of vegetation, and then distinguish different vegetation, which lays a theoretical foundation for using vegetation index time series data to extract coniferous forest information.

[0005] Currently, there are many global land cover data products publicly released, most of which only perform first-level classification of land use and do not distinguish forest types. Professor GONG Peng's team of the Department of Earth System Science of Tsinghua University and researcher LIU Liangyun's team of the Innovation Institute of Aerospace Information of the Chinese Academy of Sciences have respectively developed FROM_GLC10-2017 and GLC_FCS30-2020 land cover fine classification products, which can obtain different forest type data, but the accuracy of global classification products will decrease when they are applied to regional scales. III. SUMMARY

[0006] The purpose of the present application is to provide a method for extracting coniferous forests by using a new vegetation index. It combines different vegetation indices at different time phases by analyzing the optimal classification time phases of coniferous forests and other forest vegetation indices to construct a new vegetation index to extract coniferous forest distribution information, which has higher accuracy. Therefore, the present application uses the time series variation characteristics of each vegetation index to construct a normalized forest index (NDFI) to distinguish coniferous forests and other forests.

[0007] The purpose of the present application is achieved as follows: According to the characteristics of the time series differences of different vegetation indices, based on the forest range in the land cover data, the Landsat time series data are used to analyze the vegetation index variation, find the characteristic time points, construct the normalized forest vegetation index NDFI, and realize the extraction and distribution mapping of coniferous forest information.

[0008] The present application has the following advantages compared with ordinary vegetation indices:

[0009] (1) According to the results of annual variation analysis of different vegetation indices, the present research combines different vegetation indices at different time phases by analyzing the optimal classification time phases of coniferous forests and other forest vegetation indices to construct a new vegetation index to extract coniferous forest distribution information.

[0010] (2) The annual variation curves of different forest types vegetation indices have similar trends, but there are still some differences. The participation of the comprehensive normalized vegetation index in constructing the new vegetation index can highlight the vegetation information, enhance the difference between vegetation, and be beneficial to the differentiation of vegetation.

[0011] (3) The variation trends of DVI vegetation index reflectance from May to June, NDVI and SAVI vegetation index reflectance from June to July, RVI from June to August, and LSWI vegetation index reflectance from July to August can provide good discrimination between coniferous forests and other forests. IV. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure One Distribution map of coniferous forests in Anhui Province

[0013] Figure TwoMap of pine forest extent based on Class II survey data V. Detailed Implementation Methods

[0014] This invention: A method for extracting information about coniferous forests, the method comprising the following steps:

[0015] Step 1: This invention classifies regional forest types into coniferous forests and other types. Forest land, a primary type in the GlobeLand30 land cover data from GEE, is used as the basis for extracting coniferous forest information. Vegetation sample points are selected using high-resolution satellite imagery from Google Earth Pro software and uploaded to the GEE platform. Monthly average values ​​of vegetation indices for different forest types in the same year are extracted based on Landsat8 imagery.

[0016] Step 2: This invention selects five indices—Normalized Difference Vegetation Index (NDVI), Difference Vegetation Index (DVI), Ratio Vegetation Index (RVI), Soil-Adjusted Vegetation Index (SAVI), and Land Surface Water Index (LSWI)—to construct the coniferous forest extraction index. Detailed information is shown in Table 1.

[0017] Table 1. Vegetation indices used for extracting coniferous forest vegetation.

[0018]

[0019] Note: NIR, RED, SWIR1 and SWIR2 represent the reflectance of the near-infrared band, the red band, the short-wave infrared band 1, and the short-wave infrared band 2, respectively.

[0020] Step 3: Using the GEE cloud platform, based on the forest area, calculate and extract five vegetation indices (NDVI, DVI, RVI, SAVI, and LSWI) for sample points of coniferous forests and two other forest types within the same year. Then, average the monthly vegetation indices of the same type for each sample to obtain the monthly average spectral curves of NDVI, DVI, RVI, SAVI, and LSWI for the same year for coniferous forests and other forest types, thus identifying the optimal time point for extracting information on coniferous forests. Experimental verification shows that the optimal time point for effectively distinguishing the vegetation indices of coniferous forests from other forest types is from May to August.

[0021] Step four: According to the results of the annual change analysis of different vegetation indexes, the present application combines different vegetation indexes at different time phases to construct a new vegetation index to extract the distribution information of coniferous forest by analyzing the optimal classification time phase of each vegetation index of coniferous forest and other forest land. Since the normalized vegetation index can highlight the vegetation information and can enhance the difference between the vegetation, which is conducive to the differentiation of the vegetation, the present application selects the normalized vegetation index to participate in the construction of the new vegetation index. The normalized forest index (NDFI) for distinguishing coniferous forest and other forest land is constructed by using the time series change characteristics of each vegetation index, and the formula is as follows

[0022]

[0023] In the formula, NDFI represents the constructed normalized forest index, NDVI6, NDVI7, DVI5, DVI6, RVI6, RVI8, SAVI6, SAVI7, LSWI7 and LSWI8 represent the monthly average values of the spectral reflectance of the five vegetation indexes NDVI, DVI, RVI, SAVI and LSWI of different forest types in May, June, July and August, respectively.

[0024] Step five: The probability density curve analysis of the normalized forest index NDFI corresponding to different forest types is performed, that is, the distribution of all sample points of each forest type in the normalized forest index NDFI. According to the probability density curves of different forest types, the intersection value of the curve of coniferous forest and other forest land is calculated as the threshold value for extracting the information of coniferous forest, and the calculation formula is as follows

[0025] X = (σ1μ1+σ2μ2) / (σ1+σ2) (2)

[0026] In the formula, μ1, μ2 and σ1, σ2 represent the spectral mean and standard deviation of the NDFI index of two types of samples, respectively.

[0027] Step six: Based on the calculated vegetation index NDFI, the NDFI index value greater than the threshold value is the coniferous forest, so as to complete the extraction of the information of coniferous forest.

[0028] In order to verify the effectiveness and the extraction accuracy of coniferous forest of the present application, the applicant uses Anhui Province as the experimental area, uses the new vegetation index NDFI constructed based on the time series index change analysis, and uses the constructed new vegetation index NDFI to obtain the extraction result of the coniferous forest information of Anhui Province in 2020 by threshold discrimination. Table 2 is the distribution information of the coniferous forest of Anhui Province in 2020, Figure 1The coniferous forest distribution map in 2020 is obtained. 2000 sample points are selected in the range by using high-definition satellite image Google earth pro software, 1000 sample points of coniferous forest and other forests are taken as the verification sample of the classification result, the extraction result in 2020 is verified, and the total accuracy reaches 87.75%, and the Kappa coefficient is 0.755.

[0029] Table 2 is the accuracy evaluation of the extraction result of the coniferous forest.

[0030]

[0031] Table 3 is the extracted coniferous forest area and the statistical area of the small class with pine forest as the dominant tree species and the relative error of the two. Figure 2 The small class statistical area distribution range map with pine forest as the dominant tree species is obtained according to the second class survey data. According to the area statistical table and the distribution range comparison chart, it can be obtained that the extraction result of the coniferous forest is basically consistent with the distribution range of the small class statistical area. The small class statistical area is relatively larger than the classification result area, and the distribution range of some areas is different, and the relative error is 7.93%. It is considered that the small class area is the pine forest area with pine forest as the dominant tree species, which is larger than the actual pine forest area. Therefore, it is proved that the method of the present application meets the accuracy requirement for the extraction range of the coniferous forest information in Anhui Province, has high accuracy, and can be used for the information extraction of the coniferous forest.

[0032] Table 3 is the accuracy evaluation of the classification area.

[0033]

[0034] It can be seen that the method of the present application has good effect for the extraction of the coniferous forest, can realize the accurate extraction of the coniferous forest, and has great advantages for excluding interference and monitoring diseases and pests.

[0035] The above is only the preferred embodiment of the present application, and those skilled in the art can make some improvements and refinements without departing from the principle of the present application, and these improvements and refinements are also regarded as the protection scope of the present application.

Claims

1. A method for extracting coniferous forest information, characterized in that: firstly, five indexes of normalized difference vegetation index (NDVI), difference vegetation index (DVI), ratio vegetation index (RVI), soil-adjusted vegetation index (SAVI) and land surface water index (LSWI) are used as the basis for constructing coniferous forest extraction index, and the monthly average data of NDVI, DVI, RVI, SAVI and LSWI of coniferous forest and other forest land are extracted in the forest land range, to obtain the time series monthly average spectral curve of NDVI, DVI, RVI, SAVI and LSWI, and the time point with the largest difference between coniferous forest and other forest land in the curve is selected as the best time point for extracting coniferous forest information; secondly, the normalized forest index (NDFI) for distinguishing coniferous forest and other forest land is constructed by using the time sequence variation characteristics of each vegetation index, and the calculation formula is as follows: NDFI = (NDVI6-NDVI7) / (σ1+σ2) (1) wherein, NDFI represents the constructed normalized forest index, NDVI6, NDVI7, DVI5, DVI6, RVI6, RVI8, SAVI6, SAVI7, LSWI7 and LSWI8 represent the monthly average of NDVI, DVI, RVI, SAVI and LSWI spectral reflectance of different forest types in May, June, July and August; finally, the threshold value for extracting coniferous forest information is calculated by establishing the probability density curve of different forest types, and the intersection value of the curve of coniferous forest and other forest land is taken as the threshold value, and the NDFI index value greater than the threshold value is coniferous forest, so as to complete the extraction of coniferous forest information. The NDFI index makes full use of the time sequence variation characteristics of vegetation index of coniferous forest and other forest types, which can highlight the vegetation information, enhance the difference between vegetation and be conducive to the differentiation of vegetation. In order to better extract coniferous forest information, the probability density curve of normalized forest index NDFI corresponding to different forest types is calculated, the intersection value of the curve of coniferous forest and other forest land is calculated according to the probability density curve, which is taken as the threshold value for extracting coniferous forest information, and the formula is as follows: X = (σ1μ1+σ2μ2) / (σ1+σ2) (2) wherein, μ1, μ2 and σ1, σ2 represent the spectral mean and standard deviation of NDFI index of two types of samples. ​ ​ 2. The method for extraction of information of coniferous forest according to claim 1, characterized in that: ​ 3. The method for extraction of information of coniferous forest according to claim 1, characterized in that: ​ ​ ​

Images