Remote sensing image stony desertification grading method, device and equipment based on feature space

Abstract

The invention belongs to the field of space earth observation technology and application, and discloses a remote sensing image stony desertification grading method, device and equipment based on a feature space, and the method comprises the steps: obtaining a remote sensing image, elevation data DEM and land utilization data of a research area, and carrying out the preprocessing; terrain correction is carried out on the area with the gradient larger than 25 degrees through an SCS + C terrain correction model; constructing a rock index RCRI and a vegetation index NDRE, and performing normalization; an RCRI-NDRE feature space is constructed; constructing a stony desertification index according to a distance formula from a point to a straight line; high-resolution remote sensing images are used for assisting visual interpretation of stony desertification sample points of different degrees, and grading thresholds between stony desertification of different degrees in the RCRI-NDRE feature space are calculated; and stony desertification grading is carried out based on the grading threshold value, and distribution conditions of stony desertification of different degrees are obtained. According to the method, remote sensing image data are extracted, and the RCRI-NDRE feature space of the rock index RCRI and the vegetation index NDRE is constructed, so that different degrees of stony desertification grading of the target area can be realized.

Description

Technical Field

[0001] The invention belongs to the field of space earth observation technology and application, and relates to a remote sensing image rocky desertification classification method, device and equipment based on feature space. Background Art

[0002] Karst Rock Desertification (KRD) is a serious environmental problem. It refers to the situation in which the soil is severely eroded, the bedrock is exposed on a large scale, the land production capacity is sharply reduced, and the surface presents a land degradation phenomenon similar to desertification due to the interference of unreasonable human activities in the fragile karst geological background in the subtropical region. In order to effectively deal with the problem of rocky desertification, it is very important to accurately and efficiently grasp the distribution of rocky desertification, because this will directly affect the targeted governance measures to deal with the problem of rocky desertification, thereby improving the efficiency of rocky desertification governance and reducing costs.

[0003] Traditional rocky desertification surveys mainly rely on field observations. Although this method is direct, it consumes huge manpower and material resources, and is time-consuming and difficult to meet the needs of large-scale, rapid monitoring. In contrast, remote sensing earth observation technology has become an ideal choice for monitoring rocky desertification with its significant advantages such as wide observation range, high data acquisition efficiency, and rich spectral information. Through remote sensing technology, surface information can be quickly obtained, providing strong support for the monitoring and classification of rocky desertification.

[0004] In remote sensing, there are many methods that can be used for the classification of rocky desertification, including classification index method, machine learning method, feature space method, etc. Among them, the classification index method is a method of constructing a decision tree based on the existing rocky desertification classification standard. It can quickly and easily obtain the distribution of rocky desertification of different levels in the study area. However, this method has certain limitations in adapting to the requirements of rocky desertification classification under different environmental conditions. The machine learning method obtains prior knowledge by annotating a large number of samples for training models, and the trained model can be applied to the classification of rocky desertification. The feature space rule constructs a rocky desertification index for classification by exploring the linear relationship between different rocky desertification characterization indices. This method can not only better reflect the interaction between the characterization indices, but also does not require the acquisition of prior knowledge in advance. Therefore, it has great application potential and flexibility in the classification of rocky desertification. Summary of the invention

[0005] The purpose of the present invention is to overcome the existing problems and propose a remote sensing image rocky desertification classification method, device and equipment based on feature space, which can effectively realize the rocky desertification classification of rocky desertification areas of different degrees in the target area.

[0006] To solve the above problems, the present invention provides the following technical solutions:

[0007] The remote sensing image rocky desertification classification method based on feature space includes:

[0008] S1. Obtain remote sensing images, elevation data DEM and land use data of the rocky desertification coverage study area and perform preprocessing;

[0009] S2. Use the elevation data DEM to calculate the slope, set the area with a slope greater than 25° as the steep slope area, and use the SCS+C terrain correction model to perform terrain correction on the steep slope area to eliminate the possible impact of terrain factors on subsequent index inversion;

[0010] S3. Use remote sensing images to obtain spectral information of relevant areas, extract specific characteristics of rocks and vegetation, and obtain corresponding spectral characteristics; construct rock index RCRI based on the spectral characteristics of rocks, and construct vegetation index NDRE based on the spectral characteristics of vegetation;

[0011] S4. Construct RCRI-NDRE feature space based on rock index RCRI and vegetation index NDRE;

[0012] S5. According to the distribution law of rock index RCRI and vegetation index NDRE in RCRI-NDRE feature space and the distance formula from point to line, the rocky desertification index is constructed;

[0013] S6. Use high-resolution remote sensing images to assist visual interpretation of sample points of rocky desertification at different levels, calculate sample points of rocky desertification at different levels in the RCRI-NDRE feature space, and determine the classification thresholds of rocky desertification at different levels;

[0014] S7. Classify rocky desertification based on the classification threshold and obtain the distribution of rocky desertification of different degrees.

[0015] According to one aspect of the present invention, in step S1, the preprocessing includes: performing radiation correction, atmospheric correction, mosaicking, and cropping on the remote sensing image; splicing and cropping the elevation data; extracting water bodies, buildings and other non-desertification areas that are unlikely to develop desertification from the land use data, and masking the remote sensing image based on the extracted non-desertification areas to improve the accuracy and efficiency of the subsequent desertification classification.

[0016] According to one aspect of the present invention, in the step S3, in the step of constructing the rock index RCRI and the vegetation index NDRE, full consideration is given to the strong absorption of vegetation in the red light band and the high reflection in the red edge band, and the reflectivity of rock in the red edge band and the near infrared band is lower than that of vegetation, while the reflectivity of other bands is higher than that of vegetation. The carbonate rock index RCRI is constructed using the red light band RED, the infrared band NIR and the blue light band BLUE, and the calculation formula is as follows:

[0017] RCRI=(BLUE+RED-NIR) / (BLUE+NIR)

[0018] The vegetation index NDRE is constructed using the red light band RED and the red edge band RE. The calculation formula is as follows:

[0019] NDRE=(RE-RED) / (RE+RED)

[0020] According to one aspect of the present invention, in step S4, the rock index RCRI and the vegetation index NDRE are normalized, and the normalized rock index RCRI and the vegetation index NDRE are used as the ordinate and the abscissa, respectively, to construct a RCRI-NDRE feature space; in the RCRI-NDRE feature space, the lower right area has a higher vegetation index NDRE and a lower rock index RCRI, as an area with a lighter level of desertification; on the contrary, the upper left area has a higher rock index RCRI and a lower vegetation index NDRE, as an area with a more serious level of desertification.

[0021] According to one aspect of the present invention, in step S5, the rock index RCRI and the vegetation index NDRE are linearly fitted based on the least squares principle to obtain the fitting straight line of the rock index RCRI and the vegetation index NDRE, and a perpendicular line y is made at the intersection of the fitting straight line and the x-axis. The perpendicular line y is used as the baseline to calculate the distance from the point in the RCRI-NDRE feature space to the straight line y; the point farther away from the straight line y in the RCRI-NDRE feature space is regarded as a severe desertification level, and conversely, the point closer to the straight line y in the RCRI-NDRE feature space is regarded as a lighter desertification level, thereby constructing a desertification index.

[0022] According to one aspect of the present invention, in step S6, in the step of using high-resolution remote sensing images to assist visual interpretation to evenly select rocky desertification sample points of different levels, the means of the rocky desertification sample points of different levels are calculated, and then the median of the means of adjacent levels of rocky desertification is calculated as the grading threshold for distinguishing different levels of rocky desertification.

[0023] According to one aspect of the present invention, in the step S7, in the step of performing rocky desertification classification based on the classification threshold, the rocky desertification index of each pixel of the remote sensing image of the target area is calculated, and the rocky desertification classification is performed using the classification threshold.

[0024] According to one aspect of the present invention, the remote sensing image rocky desertification classification device based on RCRI-NDRE feature space comprises:

[0025] Preprocessing module, used to preprocess remote sensing images, elevation data and land use data of the target area;

[0026] The terrain correction module is used to correct the reflectivity error caused by terrain factors in remote sensing images and improve the accuracy of subsequent desertification characterization index inversion;

[0027] Index calculation module, used to invert rock index and vegetation index using remote sensing images and perform normalization;

[0028] The feature space module is used to construct the RCRI-NDRE feature space and rocky desertification index, and calculate the classification thresholds of different levels of rocky desertification;

[0029] The classification module is used to classify the rocky desertification in the target area according to the calculated rocky desertification classification threshold using the RCRI-NDRE feature space.

[0030] According to one aspect of the present invention, the present invention also provides a remote sensing image rocky desertification classification device based on the RCRI-NDRE feature space, wherein the device is used to store computer instructions, and when the computer instructions are executed by a processor, a remote sensing image rocky desertification classification method based on the RCRI-NDRE feature space can be implemented.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] According to the scheme of the present invention, the red edge band that can sensitively reflect the vegetation growth conditions is used to invert and characterize the vegetation index NDRE in the study area, thereby effectively distinguishing rocks from vegetation; in addition, there are significant differences between rocks and vegetation in the red band, and the effect of effectively distinguishing rocks from vegetation can be achieved by adding the red light band to the carbonate rock index RCRI; the rocky desertification characterization index constructed in this way has advantages in rocky desertification classification.

[0033] According to the solution of the present invention, the adopted feature space model can effectively reflect the interaction between the two characterization indices (rock index RCRI and vegetation index NDRE), is more sensitive to different levels of rocky desertification, and can effectively improve the accuracy of rocky desertification classification; in addition, affected by the different geological, topographical, environmental and climatic conditions in different research areas, the rocky desertification classification standard is not static. Through the high-resolution remote sensing image-assisted visual interpretation of rocky desertification sample points of different degrees, the classification thresholds of different degrees of rocky desertification are calculated, which can better adapt to the requirements of rocky desertification classification under different environmental conditions, be more in line with the actual situation of the target area, and achieve better rocky desertification classification effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0035] Figure 1 A schematic diagram schematically shows a flow chart of a remote sensing image rocky desertification classification method based on feature space according to an embodiment of the present invention;

[0036] Figure 2 A schematic diagram schematically showing a spectrum curve of rocks and vegetation according to an embodiment of the present invention;

[0037] Figure 3 A schematic diagram schematically showing a RCRI-NDRE feature space model according to an embodiment of the present invention;

[0038] Figure 4 A schematic diagram schematically shows the RCRI-NDRE feature space according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0040] For easier understanding, see Figure 1 The present invention provides a remote sensing image rocky desertification classification method based on feature space. This embodiment takes the Sustainable Development Goals Science Satellite 1 (SDGSAT-1) remote sensing image as an example to classify the rocky desertification of the remote sensing image, including the following steps:

[0041] Step 1: Obtain remote sensing images, elevation data DEM and land use data of the desertification coverage study area and perform preprocessing.

[0042] In one embodiment, the specific implementation process of the preprocessing in step 1 includes: performing radiation correction, atmospheric correction, mosaicking, and cropping on the remote sensing image; splicing and cropping the elevation data DEM; extracting water bodies, buildings, and other non-desertification areas that are unlikely to develop desertification from the land use data, and masking the remote sensing image based on the extracted non-desertification areas to improve the accuracy and efficiency of the subsequent desertification classification.

[0043] Specifically, obtain multi-view SDGSTA-1 remote sensing images covering the target area for absolute radiation correction (Formula (1)), atmospheric correction, mosaicking, and cropping; obtain DEM elevation data covering the target area for mosaicking and cropping; obtain land use data of the target area, extract water bodies, buildings and other areas where desertification is unlikely to develop, and mask the remote sensing image based on the extracted non-desertification area.

[0044] L=DN×Gain+Bias (1)

[0045] Where L represents the brightness of the sensor entrance pupil, DN represents the digital brightness value after relative radiometric correction, Gain and Bias represent gain and bias, respectively.

[0046] Step 2: Use the elevation data DEM to calculate the slope, and use the SCS+C terrain correction model to perform terrain correction on areas with slopes greater than 25°.

[0047] In one embodiment, an area with a slope greater than 25° is set as a steep slope area, and the steep slope area is subjected to terrain correction using the SCS+C terrain correction model to eliminate the possible influence of terrain factors on subsequent exponential inversion.

[0048] Specifically, the slope of the study area is calculated using the elevation data DEM data, and the SCS+C terrain correction model (Formula (2)) is used to correct the terrain for steep slopes with a slope greater than 25°. T and cosγ i The linear correlation between the semi-empirical parameter C λ (Formula (4)).

[0049] ρ H =ρ T (cosθ P *cosθ Z +C λ ) / (cosγ i +C λ ) (2)

[0050] cosγ i = cosθ P *cosθ Z +sinθ P *sinθ Z *cos(φ a -φ o ) (3)

[0051] ρ T =α λ +βλ * cosγ i(4)

[0052] ρ H is the corrected DN value, ρ T is the original DN value, θ P is the slope angle, θ Z is the solar zenith angle, C λ is the semi-empirical parameter of the λ band, cosγ i is the solar radiation coefficient, φ a is the solar azimuth, φ o is the slope direction, α λ , β λ are the empirical coefficients of the λ band, C λ =α λ / β λ .

[0053] Step 3: construct the rock index RCRI based on the spectral characteristics of rocks, and construct the vegetation index NDRE based on the spectral characteristics of vegetation.

[0054] In one embodiment, in the process of constructing the rock index RCRI and the vegetation index NDRE according to the spectral characteristics of rocks and vegetation, full consideration is given to the strong absorption of vegetation in the red light band and the high reflection in the red edge band, and the reflectivity of rocks in the red edge band and the near-infrared band is lower than that of vegetation, while the reflectivity of other bands is higher than that of vegetation. The carbonate rock index RCRI is constructed using the red light band RED, the infrared band NIR and the blue light band BLUE, and the vegetation index NDRE is constructed using the red light band RED and the red edge band RE.

[0055] Specifically, the spectral characteristics of rocks and vegetation in SDGSAT-1 remote sensing images are as follows: Figure 2 As shown. Vegetation strongly absorbs electromagnetic radiation in the red light band, resulting in a reflection trough, while the reflectivity increases rapidly in the red edge band, resulting in a reflection peak; the reflectivity of rock also shows an upward trend between the red light and red edge bands, but the slope of the increase is smaller than that of vegetation. In addition, the reflectivity of rock in the red edge band and near-infrared band is lower than that of vegetation, while the opposite is true for other bands. The improved carbonate rock index RCRI is constructed based on the spectral characteristics of rock, and the calculation formula is shown in formula (5).

[0056] RCRI=(BLUE+RED-NIR) / (BLUE+NIR) (5)

[0057] The improved vegetation index NDRE is constructed based on the spectral characteristics of vegetation, and the calculation formula is as shown in formula (6).

[0058] NDRE=(RE-RED) / (RE+RED) (6)

[0059] Among them, BLUE is the spectral reflectance of the blue light band; RED is the spectral reflectance of the red light band; RE is the spectral reflectance of the red edge band; NIR is the spectral reflectance of the near infrared band.

[0060] Step 4: Construct the RCRI-NDRE feature space based on the rock index RCRI and the vegetation index NDRE.

[0061] In one embodiment, the rock index RCRI and the vegetation index NDRE are normalized, and the normalized rock index RCRI and vegetation index NDRE are used as the ordinate and abscissa, respectively, to construct a RCRI-NDRE feature space; in the RCRI-NDRE feature space, the lower right area has a higher vegetation index NDRE and a lower rock index RCRI, as an area with a lighter level of desertification; on the contrary, the upper left area has a higher rock index RCRI and a lower vegetation index NDRE, as an area with a more serious level of desertification.

[0062] Specifically, Figure 3 As shown in the figure, the RCRI-NDRE feature space is a two-dimensional plane composed of two indexes. The rock index RCRI and the vegetation index NDRE are normalized and used as the ordinate and abscissa of the RCRI-NDRE feature space, respectively, which can better reflect the interaction between the rocky desertification characterization indexes. The vegetation index NDRE is closely related to the vegetation growth status. The larger the vegetation index NDRE, the higher the vegetation coverage, the less exposed rock and soil, and the lighter the rocky desertification. As the degree of rocky desertification intensifies and the rock exposure increases, the rock index RCRI shows an increasing trend. Therefore, the change process of rocky desertification from light to heavy from the lower right to the upper left in the RCRI-NDRE feature space. The rocky desertification level can be confirmed by calculating the position of the point in the RCRI-NDRE feature space. The RCRI-NDRE feature space is constructed based on this principle.

[0063] Step 5: construct a rocky desertification index according to the distribution law of the rock index RCRI and the vegetation index NDRE in the RCRI-NDRE feature space and the distance formula from a point to a straight line.

[0064] In one embodiment, the rock index RCRI and the vegetation index NDRE are linearly fitted based on the least squares principle to obtain the fitting straight line of the rock index RCRI and the vegetation index NDRE, and a perpendicular line y is drawn at the intersection of the fitting straight line and the x-axis. The perpendicular line y is used as the reference line to calculate the distance from the point in the RCRI-NDRE feature space to the straight line y; the points in the RCRI-NDRE feature space that are farther away from the straight line y are regarded as severe desertification levels, and conversely, the points in the RCRI-NDRE feature space that are closer to the straight line y are regarded as lighter desertification levels, thereby constructing a desertification index.

[0065] Specifically, Figure 4 As shown in the figure, a univariate linear fitting is performed according to the least squares principle to obtain the fitting straight lines of the rock index RCRI and the vegetation index NDRE. The perpendicular line of the intersection of the fitting straight line and the x-axis is y=kx-b. As the degree of desertification intensifies, the RCRI-NDRE feature space points to the direction of increasing rock index RCRI and decreasing vegetation index NDRE, and vice versa, the degree of desertification decreases. Therefore, the elliptical section perpendicular to the straight line y is used as the boundary to distinguish different levels of desertification. The points of the same level of desertification are between two adjacent elliptical sections, from right to left, they are no, potential, mild, moderate, and severe desertification. Substituting the parameters of the straight line y into the distance formula from the point to the straight line, the desertification index of the RCRI-NDRE feature space is constructed (Formula (7)).

[0066]

[0067] Among them, KRDI is the rocky desertification index, NDRE is the normalized difference red edge index, and RCRI is the improved carbonate rock index.

[0068] Step 6: Use high-resolution remote sensing images to assist in visual interpretation of sample points of rocky desertification of different degrees, calculate sample points of rocky desertification of different degrees in the RCRI-NDRE feature space, and determine the classification thresholds of rocky desertification of different degrees.

[0069] Specifically, sample points of different degrees of desertification were selected through high-resolution images, the mean of KRDI for each level of desertification was calculated, and then the median of the mean of the desertification index of two adjacent desertification levels was taken, and this median was used as the threshold for dividing different levels of desertification.

[0070] Step 7: classify rocky desertification based on the classification threshold to obtain the distribution of rocky desertification of different degrees.

[0071] Specifically, in the process of rocky desertification classification based on the classification threshold, the rocky desertification index of each pixel of the remote sensing image of the target area is calculated, the rocky desertification classification is performed according to the classification threshold, and then the distribution of rocky desertification of different degrees in the target area is obtained by reclassifying the image.

[0072] The embodiment of the present invention provides a remote sensing image rocky desertification degree classification device based on feature space, comprising:

[0073] Preprocessing module, used to preprocess remote sensing images, elevation data and land use data of the target area;

[0074] The terrain correction module is used to correct the reflectivity error caused by terrain factors in remote sensing images and improve the accuracy of subsequent desertification characterization index inversion;

[0075] Index calculation module, used to invert rock and vegetation indices using remote sensing images and normalize the indices;

[0076] The feature space module is used to construct the RCRI-NDRE feature space and rocky desertification index, and calculate the classification thresholds of different levels of rocky desertification;

[0077] The classification module is used to classify the rocky desertification in the target area according to the calculated rocky desertification classification threshold using the RCRI-NDRE feature space.

[0078] The embodiment of the present invention also provides a remote sensing image rocky desertification degree classification device based on feature space, the device includes computer instructions, and when the computer instructions are executed by a processor, a remote sensing image rocky desertification classification method based on RCRI-NDRE feature space can be implemented.

[0079] In summary, according to the concept of the present invention, the above scheme of the embodiment of the present invention designs a remote sensing image rocky desertification classification method based on feature space. In this method, the red edge band can effectively monitor the growth of vegetation, thereby effectively distinguishing rocks from vegetation. In addition, there are significant differences between rocks and vegetation in the red band. By adding the red light band to the carbonate rock index RCRI, the effect of effectively distinguishing rocks from vegetation can also be achieved. The rock index RCRI and the vegetation index NDRE in the rocky desertification characterization index constructed in this way have advantages in rocky desertification classification. In addition, the RCRI-NDRE feature space method we use can effectively reflect the interaction between the two surface parameters, is more sensitive to rocky desertification, and can effectively improve the accuracy of rocky desertification classification. Secondly, under different geological, topographical characteristics and environmental and climatic conditions in different regions, rocky desertification classification will appear more complicated. Compared with the use of inherent rocky desertification classification standards, we select sample points of different degrees of rocky desertification through high-resolution images, and then calculate the thresholds between different degrees of rocky desertification. Through this custom threshold method, it can well adapt to the requirements of rocky desertification classification under different environmental conditions. It is more in line with the actual situation of the target area and achieves a better rocky desertification classification effect. This method provides an operational and simple rocky desertification classification method, which can provide support for rocky desertification monitoring and effective control, and is of great significance to rocky desertification control in southwest my country.

[0080] In addition, it should be stated that the above-described implementation is only a specific form of expression of the present invention and should not be regarded as limiting the present invention. Although the embodiments of the present invention have been shown and described, for those skilled in the art, once they understand the core innovative concepts of the present invention, they can make various modifications, substitutions, variations and improvements to these embodiments without departing from the principles and spirit of the present invention, and these attached claims and their modifications, substitutions, etc. are all deemed to be within the scope of protection of the present invention.

Claims

1. A remote sensing image rocky desertification classification method based on feature space, characterized in that: include: S1. Obtain remote sensing images, elevation data DEM and land use data of the rocky desertification coverage study area and perform preprocessing; S2. Calculate the slope using the elevation data DEM, set the area with a slope greater than 25° as a steep slope area, and perform terrain correction on the steep slope area using the SCS+C terrain correction model; S3. Use remote sensing images to obtain spectral information of relevant areas, extract specific characteristics of rocks and vegetation, and obtain corresponding spectral characteristics; construct rock index RCRI based on the spectral characteristics of rocks, and construct vegetation index NDRE based on the spectral characteristics of vegetation; S4. Construct RCRI-NDRE feature space based on rock index RCRI and vegetation index NDRE; S5. According to the distribution law of rock index RCRI and vegetation index NDRE in RCRI-NDRE feature space and the distance formula from point to line, the rocky desertification index is constructed; S6. Use high-resolution remote sensing images to assist visual interpretation of sample points of rocky desertification at different levels, calculate sample points of rocky desertification at different levels in the RCRI-NDRE feature space, and determine the classification thresholds of rocky desertification at different levels; S7. Classify rocky desertification based on the classification threshold and obtain the distribution of rocky desertification of different degrees.

2. The method according to claim 1, characterized in that: In step S1, the preprocessing includes: performing radiation correction, atmospheric correction, mosaicking, and cropping on the remote sensing image; splicing and cropping the elevation data; extracting water bodies, buildings, and other non-desertification areas that are unlikely to develop desertification from the land use data, and masking the remote sensing image based on the extracted non-desertification areas.

3. The method according to claim 1, characterized in that In the step S3, in the step of constructing the rock index RCRI and the vegetation index NDRE, the rock index RCRI is constructed using the red light band RED, the infrared band NIR and the blue light band BLUE, and the calculation formula is as follows: RCRI=(BLUE+RED-NIR) / (BLUE+NIR) The vegetation index NDRE is constructed using the red light band RED and the red edge band RE. The calculation formula is as follows: NDRE=(RE-RED) / (RE+RED) 4. The method according to claim 1, characterized in that: In the step S4, the rock index RCRI and the vegetation index NDRE are normalized, and the normalized rock index RCRI and the vegetation index NDRE are used as the ordinate and the abscissa respectively to construct the RCRI-NDRE feature space; In the RCRI-NDRE feature space, the lower right area has a higher vegetation index NDRE and a lower rock index RCRI, which is an area with a lighter desertification level. On the contrary, the upper left area has a higher rock index RCRI and a lower vegetation index NDRE, which is an area with a more serious desertification level.

5. The method according to claim 1, characterized in that In the step S5, the rock index RCRI and the vegetation index NDRE are linearly fitted based on the least squares principle to obtain the fitting straight line of the rock index RCRI and the vegetation index NDRE, and a perpendicular line y is drawn at the intersection of the fitting straight line and the x-axis. The perpendicular line y is used as the reference line to calculate the distance from the point in the RCRI-NDRE feature space to the straight line y; the point farther away from the straight line y in the RCRI-NDRE feature space is regarded as a severe desertification level, and conversely, the point closer to the straight line y in the RCRI-NDRE feature space is regarded as a lighter desertification level, thereby constructing a desertification index.

6. The method according to claim 1, characterized in that In step S6, in the step of using high-resolution remote sensing images to assist visual interpretation to evenly select rocky desertification sample points of different levels, the means of rocky desertification sample points of different levels are calculated, and then the median of the means of adjacent levels of rocky desertification is calculated as the grading threshold for distinguishing different levels of rocky desertification.

7. The method according to claim 1, characterized in that In the step S7, in the step of performing rocky desertification classification based on the classification threshold, the rocky desertification index of each pixel of the remote sensing image of the target area is calculated, and the rocky desertification classification is performed according to the classification threshold.

8. A remote sensing image rocky desertification classification device using the remote sensing image rocky desertification classification method based on feature space according to any one of claims 1 to 7, characterized in that: include: Preprocessing module, used to preprocess remote sensing images, elevation data and land use data of the target area; Terrain correction module, used to correct reflectivity errors in remote sensing images caused by terrain factors; Index calculation module, used to invert rock index and vegetation index using remote sensing images and perform normalization; The feature space module is used to construct the RCRI-NDRE feature space and rocky desertification index, and calculate the classification thresholds of different levels of rocky desertification; The classification module is used to classify the rocky desertification in the target area according to the calculated rocky desertification classification threshold using the RCRI-NDRE feature space.

9. A remote sensing image rocky desertification classification device using the remote sensing image rocky desertification classification method based on feature space according to any one of claims 1 to 7, characterized in that: The device is used to store computer instructions, and when the computer instructions are executed by a processor, any remote sensing image rocky desertification classification method based on RCRI-NDRE feature space in claims 1-7 is implemented.

Images