A desertification monitoring method based on multi-dimensional remote sensing features and a Gaussian mixture model

Abstract

The application provides a desertification monitoring method based on multi-dimensional remote sensing features and a Gaussian mixture model, and the method comprises the following steps: acquiring remote sensing image data of a region to be monitored, and pre-processing the remote sensing image data; based on the pre-processed remote sensing data, extracting a vegetation cover index, a ground albedo and a soil moisture index as three key feature dimensions; constructing a three-dimensional desertification index by using Euclidean distance, which is used for comprehensively representing the desertification degree; and classifying the three-dimensional desertification index by using a Gaussian mixture model to obtain a desertification distribution map of different grades. By means of multi-dimensional feature fusion and unsupervised classification, the objectivity, accuracy and regional adaptability of desertification monitoring are significantly improved, and the method is suitable for large-scale, long-time desertification dynamic monitoring and evaluation.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing monitoring and land degradation assessment technology, specifically relating to a desertification monitoring method based on multidimensional remote sensing features and a Gaussian mixture model. Background Technology

[0002] Desertification is one of the most serious ecological and environmental problems globally, especially in arid and semi-arid regions, where land degradation leads to the loss of ecosystem functions, biodiversity decline, and hindered socio-economic development. Traditional desertification monitoring relies heavily on field surveys, which, while accurate, suffer from limitations such as high cost, low efficiency, and difficulty in achieving large-scale and continuous monitoring. With the development of remote sensing technology, monitoring methods based on multi-source remote sensing have gradually become a research hotspot. However, existing methods still have shortcomings: one-dimensional indices such as NDVI can reflect vegetation cover but are difficult to characterize soil moisture and surface bareness; two-dimensional indices such as DDI, while taking into account both vegetation and albedo, have limited applicability, strong parameter dependence, and are difficult to generalize. While point surveys and monitoring classification can obtain good results locally, they rely on samples and are difficult to conduct long-term, cross-regional stable monitoring. Furthermore, since desertification is driven by multiple factors such as vegetation degradation, soil moisture loss, and changes in surface energy, existing methods often fail to comprehensively characterize the process, leading to unstable classification results and insufficient comparability between regions.

[0003] Current capabilities in acquiring multi-source remote sensing imagery have significantly improved. MODIS offers high temporal resolution, making it suitable for long-term time-series monitoring; Landsat and Sentinel-2, on the other hand, have advantages in spatial resolution, capturing detailed surface features. However, achieving index construction and result unification across multi-source images remains a key challenge. Furthermore, methods relying on manual thresholding or large training samples suffer from high computational costs and poor generalization ability. Summary of the Invention

[0004] The purpose of this invention is to provide a desertification monitoring method based on multidimensional remote sensing features and a Gaussian mixture model, to solve the problems of single features, reliance on samples, and poor regional adaptability in existing technologies. The desertification monitoring method includes: Acquire optical remote sensing images of the monitored area and preprocess the optical remote sensing images. The optical remote sensing image data includes MODIS images, Landsat images, and Sentinel-2 images.

[0005] The red, green, blue, near-infrared, and short-infrared bands of each pixel are extracted from the preprocessed optical remote sensing image, and the vegetation cover index, surface albedo, and soil moisture index of each pixel in the optical remote sensing image are calculated based on the red, green, blue, near-infrared, and short-infrared bands.

[0006] The spatial distance from each pixel to the preset desertification extreme point is calculated using the Euclidean distance method to obtain the three-dimensional desertification index of each pixel. The formula for calculating the three-dimensional desertification index is as follows: , in, SAVI For vegetation cover index, Albedo For surface albedo, WET Soil moisture index, , , These are the vegetation cover index, surface albedo, and soil moisture index at the preset desertification extreme points, respectively.

[0007] A pre-trained Gaussian mixture model was used to perform unsupervised classification of the three-dimensional desertification index of each pixel. N clusters were set to correspond to N different desertification levels, where N is a natural number greater than 2. The model parameters were estimated using the expectation-maximization algorithm to obtain the desertification level distribution map of the monitored area.

[0008] Furthermore, the preprocessing includes radiometric calibration, atmospheric correction, cloud masking, image registration, and spatial resolution resampling.

[0009] Furthermore, the formulas for calculating the vegetation cover index, surface albedo, and soil moisture index of each pixel are as follows: , Wherein, NIR stands for near-infrared band, Red for red band, Blue for blue band, and Green for green band. SWIR1 For shortwave infrared band 1, SWIR2 It is shortwave infrared band 2.

[0010] Furthermore, the Gaussian mixture model estimates the optimal probability model parameters using the expectation-maximization algorithm.

[0011] Furthermore, the Gaussian mixture model has 5 clusters and 5 levels of desertification, namely non-desertification, mild desertification, moderate desertification, severe desertification and extremely severe desertification.

[0012] Furthermore, the Gaussian mixture model predicts the results using multiple Gaussian probability density functions.

[0013] Furthermore, samples were selected from field survey data, historical remote sensing images, and desertification survey data to train and verify the accuracy of the Gaussian mixture model.

[0014] The present invention also proposes a desertification monitoring system based on multidimensional remote sensing features and a Gaussian mixture model, comprising at least a microprocessor and a memory, wherein the microprocessor is programmed or configured to execute the steps of the above-described desertification monitoring method, or the memory stores a computer program programmed or configured to execute the above-described desertification monitoring method.

[0015] The present invention also proposes a computer-readable storage medium storing a computer program programmed or configured to perform the above-described desertification monitoring method.

[0016] The desertification monitoring method based on multidimensional remote sensing features and Gaussian mixture model provided by this invention constructs a three-dimensional desertification index that can more comprehensively reflect the complex process of desertification by integrating multidimensional key features such as vegetation cover index, surface albedo, and soil moisture index. By adopting an unsupervised classification Gaussian mixture model, the reliance on a large number of manual samples is reduced, and the adaptability and objectivity of the method in different regions are enhanced. This method has high accuracy and stability in large-scale, long-term dynamic monitoring of desertification, and can provide effective technical support for desertification prevention and ecological governance in China and even globally. Attached Figure Description

[0017] Figure 1 This is a flowchart of a desertification classification method based on multidimensional remote sensing features and a Gaussian mixture model.

[0018] Figure 2 This is a schematic diagram of the Gaussian mixture model.

[0019] Figure 3 This is a schematic diagram of the construction of TDDI based on multidimensional remote sensing features in three-dimensional space. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0021] like Figure 1 As shown, the desertification classification method involved in this embodiment includes the following steps: The first step is to acquire optical remote sensing images of the monitored area and preprocess the optical remote sensing images. The optical remote sensing image data includes MODIS images, Landsat images, and Sentinel-2 images.

[0022] In this step, the preprocessing includes radiometric calibration, atmospheric correction, cloud masking, image registration, and spatial resolution resampling to ensure consistency and comparability among multi-source data.

[0023] The second step is to extract the red, green, blue, near-infrared, and short-infrared bands of each pixel from the preprocessed optical remote sensing image, and calculate the vegetation cover index, surface albedo, and soil moisture index of each pixel in the optical remote sensing image based on the red, green, blue, near-infrared, and short-infrared bands.

[0024] Vegetation cover index, surface albedo, and soil moisture index respectively characterize vegetation status, surface reflectance, and soil moisture conditions, and their calculation formulas are as follows: , Wherein, NIR stands for near-infrared band, Red for red band, Blue for blue band, and Green for green band. SWIR1 For shortwave infrared band 1, SWIR2 It is shortwave infrared band 2.

[0025] The third step involves using the Euclidean distance method to calculate the spatial distance from each pixel to the preset desertification extreme point, thereby obtaining the three-dimensional desertification index for each pixel. The formula for calculating the three-dimensional desertification index is as follows: , Among them, SAVI is the vegetation cover index, Albedo is the surface albedo, and WET is the soil moisture index. , , These are the vegetation cover index, surface albedo, and soil moisture index at preset desertification extreme points. TDDI is a three-dimensional desertification index used to comprehensively assess the degree of desertification.

[0026] The fourth step involves using a pre-trained Gaussian mixture model to perform unsupervised classification of the three-dimensional desertification index of each pixel, thereby obtaining a desertification level distribution map of the monitored area.

[0027] The construction diagram of the Gaussian mixture model is shown below. Figure 2As shown, the model employs the Expectation-Maximization (EM) algorithm. It alternately estimates the optimal probabilistic model parameters through two steps (E-step and M-step). The posterior probabilities of latent variables are calculated based on the initial model parameters or the previous iteration values. Using the expected values ​​of the latent variables obtained in the E-step, maximum likelihood estimation is performed on the parameter model to calculate the parameters of the new iteration model, namely the mean and variance. Then, multiple Gaussian probability density functions are used to predict the results. The probability density function clusters are 5, corresponding to 5 different desertification levels: non-desertification, slight desertification, moderate desertification, severe desertification, and extremely severe desertification. The training and accuracy validation samples for the Gaussian mixture model are selected from field survey data, historical remote sensing images, and desertification census data.

[0028] For the desertification monitoring method described in this embodiment, after obtaining the desertification classification results, it is also possible to combine field survey data, historical Google Earth high-definition images and desertification census data released by national authoritative departments to assist in selecting verification samples, and use the classification results obtained by the Gaussian mixture model to verify the accuracy of the desertification classification results.

[0029] To validate the desertification monitoring method used in this implementation, MODISMOD09A1 time-series data from 2000 to 2020 were first acquired, supplemented by Landsat 8 and Sentinel-2 imagery from 2020. After preprocessing including radiometric calibration, atmospheric correction, and spatial registration, three key features were extracted: Soil Adjusted Vegetation Index (TDDI), surface albedo, and soil moisture index. Based on this, a three-dimensional desertification index was constructed using the Euclidean distance algorithm. The distance from each pixel to the desertification extreme value reference point was calculated in the three-dimensional space composed of these three features, generating a TDDI index map, as shown below. Figure 3 As shown in the figure. Subsequently, a Gaussian mixture model was used to perform unsupervised classification of TDDI, setting 5 cluster numbers to correspond to different desertification levels. The expectation-maximization algorithm was used to estimate the model parameters, and finally, a desertification level distribution map was obtained.

[0030] To verify the effectiveness of the method, 1218 field-measured samples and 12915 Google Earth verification points were used for accuracy evaluation. The results showed that the overall accuracy exceeded 82%, and the Kappa coefficient was above 0.67. This indicates that the desertification monitoring method proposed in this invention can integrate multi-dimensional information such as vegetation cover, albedo, and soil moisture. Furthermore, without requiring a large number of samples, it effectively improves monitoring accuracy and applicability by constructing a comprehensive index in a three-dimensional feature space and combining it with a Gaussian mixture model for unsupervised classification, thus overcoming the shortcomings of existing technologies in large-scale, long-term desertification monitoring.

Claims

1. A desertification monitoring method based on multidimensional remote sensing features and a Gaussian mixture model, characterized in that, The method includes: Acquire optical remote sensing images of the monitored area, and preprocess the optical remote sensing images. The optical remote sensing image data includes MODIS images, Landsat and Sentinel-2 images. The red, green, blue, near-infrared and short-infrared bands of each pixel are extracted from the preprocessed optical remote sensing image, and the vegetation cover index, surface albedo and soil moisture index of each pixel in the optical remote sensing image are calculated based on the red, green, blue, near-infrared and short-infrared bands. The spatial distance from each pixel to the preset desertification extreme point is calculated using the Euclidean distance method to obtain the three-dimensional desertification index of each pixel. The formula for calculating the three-dimensional desertification index is as follows: ， in, SAVI For vegetation cover index, Albedo For surface albedo, WET Soil moisture index, , , These are the vegetation cover index, surface albedo, and soil moisture index at the preset desertification extreme points; A pre-trained Gaussian mixture model was used to perform unsupervised classification of the three-dimensional desertification index of each pixel. N clusters were set to correspond to N different desertification levels, where N is a natural number greater than 2. The model parameters were estimated using the expectation-maximization algorithm to obtain the desertification level distribution map of the monitored area.

2. The desertification monitoring method according to claim 1, characterized in that, The preprocessing includes radiometric calibration, atmospheric correction, cloud masking, image registration, and spatial resolution resampling.

3. The desertification monitoring method according to claim 1, characterized in that, The formulas for calculating the vegetation cover index, surface albedo, and soil moisture index of each pixel are as follows: ， in, NIR This refers to the near-infrared band, with Red for red, Blue for blue, and Green for green. SWIR1 For shortwave infrared band 1, SWIR2 It is shortwave infrared band 2.

4. The desertification monitoring method according to claim 1, characterized in that, The Gaussian mixture model estimates the optimal probability model parameters using the expectation-maximization algorithm.

5. The desertification monitoring method according to claim 1, characterized in that, The Gaussian mixture model has 5 clusters and 5 levels of desertification, namely non-desertification, mild desertification, moderate desertification, severe desertification and extremely severe desertification.

6. The desertification monitoring method according to claim 1, characterized in that, The Gaussian mixture model predicts results using multiple Gaussian probability density functions.

7. The desertification monitoring method according to any one of claims 1 to 6, characterized in that, Samples were selected from field survey data, historical remote sensing images, and desertification survey data to train and verify the accuracy of the Gaussian mixture model.

8. A desertification monitoring system based on multidimensional remote sensing features and a Gaussian mixture model, comprising at least a microprocessor and a memory, characterized in that, The microprocessor is programmed or configured to perform the steps of the desertification monitoring method according to any one of claims 1 to 6, or the memory stores a computer program programmed or configured to perform the desertification monitoring method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that is programmed or configured to perform the desertification monitoring method according to any one of claims 1 to 6.

Images