A method and system for remote sensing monitoring of desertification prevention and control engineering

Abstract

This invention provides a remote sensing monitoring method and system for desertification control projects, belonging to the field of desertification monitoring technology. The method includes: acquiring and preprocessing optical and radar remote sensing images based on the spatial heterogeneity of the desert; determining dynamic target control information by coupling desert ecological evolution laws with the growth cycle of psammophytes; obtaining actual control information through multi-source data fusion and inversion, and determining real-time project progress by combining dynamically calculated control effectiveness attenuation coefficients; obtaining simulated control progress with environmental confidence intervals through a wind-sand-precipitation coupled spatiotemporal sequence prediction model; and correcting the real-time progress with dynamic environmental sensitivity coefficients to obtain accurate control progress. This achieves dynamic and accurate monitoring throughout the entire process, and is adapted to the desert environment, improving the matching degree between monitoring results and project progress and environmental changes.

Description

Technical Field

[0001] This invention relates to the field of desertification monitoring technology, and in particular to a remote sensing monitoring method and system for desertification control projects. Background Technology

[0002] Remote sensing monitoring of desertification control projects is a key technology in the field of desertification control, and its monitoring results are an important basis for evaluating the effectiveness of the projects and dynamically adjusting the control plans. At present, remote sensing monitoring technology for desertification control projects still faces many technical problems in practical applications, with insufficient overall accuracy, dynamism, and adaptability: remote sensing image acquisition does not fully incorporate the spatial heterogeneity of desertification areas, and preprocessing methods are mostly generalized, resulting in insufficient effective information and a lot of interference in the acquired desert images. The determination of control targets relies heavily on fixed control plans, exhibiting static characteristics, which is incompatible with the phased advancement requirements of control projects. Furthermore, the failure to consider the dynamic impact of desert environmental factors leads to significant deviations in the calculation of real-time project progress information, resulting in low early warning accuracy. In summary, these factors reduce the overall efficiency of remote sensing monitoring and make it difficult to meet the refined and dynamic monitoring needs of desertification control projects.

[0003] Therefore, this invention proposes a remote sensing monitoring method and system for desertification control projects. Summary of the Invention

[0004] This invention provides a remote sensing monitoring method and system for desertification control projects to solve the aforementioned technical problems.

[0005] This invention proposes a remote sensing monitoring method for desertification control projects, comprising: Step 1: Real-time acquisition of remote sensing images of the target area using remote sensing monitoring equipment based on the spatial heterogeneity characteristics of desertification areas, and preprocessing of the remote sensing images of the target area, wherein the remote sensing images include optical remote sensing images and radar remote sensing images. Step 2: Based on the desertification control project in the target area, and combined with the coupling matching algorithm of desert ecological evolution law and psammophyte growth cycle, determine the dynamic target control information of the target area; Step 3: Analyze the preprocessed remote sensing images based on the desert multi-source data fusion inversion model to obtain the actual prevention and control information of the target area. Based on the actual prevention and control information and the dynamic target prevention and control information, introduce the desert prevention and control effectiveness attenuation coefficient dynamically calculated based on the wind and sand flow intensity and soil moisture to determine the real-time engineering progress information of the target area. Step 4: Based on the engineering operation time period of the target area and the desert multi-dimensional environmental data during the engineering operation time period, and combined with the wind-sand-precipitation coupled spatiotemporal sequence prediction model, obtain the simulated prevention and control progress information of the target area with environmental confidence intervals; Step 5: Based on the simulated prevention and control progress information and combined with the dynamic environmental sensitivity coefficient of the desert prevention and control stage, the real-time engineering progress information is corrected to obtain the accurate prevention and control progress of the target area.

[0006] Preferably, step 1 includes: Differentiated sub-regions are divided based on the location information and desertification level of the target area: severely desertified areas are divided with a higher density according to the first preset scale, and moderately and lightly desertified areas are divided in a regular manner according to the second preset scale. The center point of each sub-region is determined as the collection point by adaptive offsetting the vector offset algorithm based on the slope and orientation of the dune terrain combined with the desert terrain features. Using remote sensing monitoring equipment, with the collection point as the center point, sub-remote sensing images of each sub-region are collected at a resolution matched to the degree of desertification in the sub-region; Each sub-remote sensing image is seamlessly stitched together according to geographic coordinates to obtain the remote sensing image of the target area; The remote sensing images of the target area are sequentially subjected to atmospheric correction, geometric correction, dust removal, and noise removal based on anisotropic filtering of desert sand texture direction to obtain preprocessed remote sensing images.

[0007] Preferably, step 2 includes: According to the coupling matching algorithm, multiple progressive prevention and control stages of the target area are extracted from the prevention and control plan, and the core governance objectives, wind and sand adaptable engineering layout and dynamic control requirements of each prevention and control stage are determined. The dynamic control requirements are adjusted in real time based on soil moisture and monthly average wind and sand activity data. Based on the project operation time, actual construction progress data, and remote sensing monitoring ecological response data of the desertification control project, the current control stage corresponding to the target area is determined, and the dynamic governance objectives, wind and sand adaptable engineering layout, and real-time adjusted control requirements of the current control stage are determined as the dynamic target control information of the target area.

[0008] Preferably, step 3 includes: The preprocessed remote sensing images are input into the desert multi-source data fusion inversion model to obtain the actual prevention and control information of the target area. The actual prevention and control information includes vegetation cover information and soil information. The soil information is fused with multi-dimensional data of desert soil moisture, sand-fixing layer structure and water content and spatial correlation is established. The vegetation cover information of the target area is extracted from the dynamic target prevention and control information, and multiple dynamic threshold-type target prevention and control indicators of the target area are determined based on the vegetation cover information of the target area and the monthly average environmental data of the desert. The actual prevention and control information is compared with the dynamic threshold-type target prevention and control indicators, and combined with the desert prevention and control effectiveness decay coefficient to determine the real-time project progress information.

[0009] Preferably, multiple dynamic threshold-type target prevention and control indicators are determined for the target area, including: The vegetation cover information of the target area is extracted from the dynamic target prevention and control information. The vegetation cover information includes: spatial constraint information of the key zone for windbreak and sand fixation that integrates the movement law of desert wind and sand flow; improved normalized vegetation index for psammophytic plants that introduces canopy porosity and sand burial coefficient; gridded vegetation type constraint parameters and gridded vegetation density constraint parameters that combine the spatial distribution of soil moisture content. Based on the vegetation cover information, regional characteristics of the target area, monthly average environmental data of the desert, and the intensity of wind and sand movement and the thickness of the soil sand-fixing layer, multiple dynamic threshold-type target prevention and control indicators for the target area are determined.

[0010] Preferably, real-time project progress information is determined, including: According to the preset rules for differentiating desertification levels, the target area is divided into multiple grids, and based on the vegetation cover information and soil information, the vegetation richness, vegetation density and soil moisture content of each grid are calculated. Based on the prevention and control plan and the coordinate information of each grid, the topographic features and wind erosion level of each grid are determined, and based on the coordinate information of each grid, the target vegetation richness, target vegetation density and target soil moisture content of each grid are extracted from the dynamic threshold-type target prevention and control indicators. Based on the topographic influence coefficient and wind erosion level coefficient of each grid, an orthogonal two-dimensional weight matrix is ​​constructed. Calculate the first difference between the vegetation richness and the target vegetation richness, the second difference between the vegetation density and the target vegetation density, and the third difference between the soil moisture content and the target soil moisture content for each grid. Based on the orthogonal two-dimensional weight matrix of each grid, the first difference, second difference, and third difference of the grid are weighted and corrected to obtain the grid engineering progress information of each grid. Connecting the center points of every two adjacent grids yields the topology of the project progress network for the target region; The topology, edge features, and node features of the project progress network are input into a pre-trained neural network with a desert spatial heterogeneity attention layer to obtain the first project progress information of the target area. The edge features are the length and angle of the connecting lines in the topology, and the node features are the grid project progress information of each grid. From the optical remote sensing images of the target area, dust interference information is extracted by combining dust concentration inversion with image texture analysis and using a vector analysis algorithm based on dust concentration gradient and image texture direction, and the dust interference weight and interference direction of each project progress item in the first project progress information are determined. Based on the sandstorm interference weight and direction of each project progress, interference correction is performed on each project progress in the first project progress information, and a second correction is performed based on the desert control effectiveness attenuation coefficient to obtain the real-time project progress information of the target area.

[0011] Preferably, step 4 includes: The multi-dimensional environmental data of the desert in the target area during the engineering operation period are divided according to the prevention and control stage. The multi-dimensional environmental data includes the frequency of wind and sand activity, dust concentration, soil moisture, precipitation distribution and phenological data of desert plants. The environmental data of each stage are preprocessed by spatiotemporal fusion. The preprocessed environmental information of each prevention and control stage is input into a pre-trained wind-sand-precipitation coupled spatiotemporal sequence prediction model, and the simulated prevention and control progress information with environmental impact confidence intervals for each prevention and control stage is output. The wind-sand-precipitation coupled spatiotemporal sequence prediction model is implemented by integrating long short-term memory network and geographical weighted regression algorithm.

[0012] Preferably, step 5 includes: The real-time project progress information is compared with the corresponding simulated prevention and control progress information with environmental confidence intervals in multiple dimensions to obtain quantitative progress deviation information covering progress completion rate, ecological effectiveness achievement rate, and project stability. The quantitative progress deviation information is compared with the dynamic deviation range determined based on the characteristics of the prevention and control stage and real-time desert environmental data; When there are deviation data in the quantified schedule deviation information that exceed the corresponding dynamic deviation range, the core stage of the schedule deviation and the source of deviation transmission are located based on the spatial correlation characteristics of the project schedule. Based on the core deviation stage, the source of deviation transmission, and combined with the dynamic environmental sensitivity coefficient and real-time adjustment coefficient of engineering construction at that stage, the real-time engineering progress information is corrected to obtain the precise prevention and control progress of the target area.

[0013] This invention provides a remote sensing monitoring system for desertification control projects, comprising: The image preprocessing module is used to acquire remote sensing images of the target area in real time using remote sensing monitoring equipment and based on the spatial heterogeneity characteristics of desertification areas, and to preprocess the remote sensing images of the target area, wherein the remote sensing images include optical remote sensing images and radar remote sensing images. The dynamic prevention and control module is used to determine the dynamic target prevention and control information of the target area based on the prevention and control plan of the desertification prevention and control project in the target area, combined with the coupling matching algorithm of desert ecological evolution law and psammophyte growth cycle. The real-time engineering determination module is used to analyze the preprocessed remote sensing images based on the desert multi-source data fusion inversion model to obtain the actual prevention and control information of the target area. Based on the actual prevention and control information and the dynamic target prevention and control information, the module introduces the desert prevention and control effectiveness attenuation coefficient dynamically calculated based on the wind and sand flow intensity and soil moisture to determine the real-time engineering progress information of the target area. The simulation prevention and control module is used to obtain the simulation prevention and control progress information of the target area with environmental confidence intervals based on the engineering operation time period of the target area and the desert multi-dimensional environmental data during the engineering operation time period, and combined with the wind-sand-precipitation coupled spatiotemporal sequence prediction model. The progress correction module is used to correct the real-time project progress information based on the simulated prevention and control progress information and combined with the dynamic environmental sensitivity coefficient of the desert prevention and control stage to obtain the accurate prevention and control progress of the target area.

[0014] Compared with the prior art, the beneficial effects of this application are as follows: It effectively realizes the dynamic, precise and desert environment-adapted remote sensing monitoring of the entire process of desertification control projects, and effectively improves the matching degree between monitoring results and actual project progress and desert environment changes.

[0015] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0016] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a remote sensing monitoring method for desertification control engineering in an embodiment of the present invention; Figure 2This is a structural diagram of a remote sensing monitoring system for desertification control engineering in an embodiment of the present invention. Detailed Implementation

[0018] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0019] This invention proposes a remote sensing monitoring method for desertification control projects, such as... Figure 1 As shown, it includes: Step 1: Real-time acquisition of remote sensing images of the target area using remote sensing monitoring equipment based on the spatial heterogeneity characteristics of desertification areas, and preprocessing of the remote sensing images of the target area, wherein the remote sensing images include optical remote sensing images and radar remote sensing images. Step 2: Based on the desertification control project in the target area, and combined with the coupling matching algorithm of desert ecological evolution law and psammophyte growth cycle, determine the dynamic target control information of the target area; Step 3: Analyze the preprocessed remote sensing images based on the desert multi-source data fusion inversion model to obtain the actual prevention and control information of the target area. Based on the actual prevention and control information and the dynamic target prevention and control information, introduce the desert prevention and control effectiveness attenuation coefficient dynamically calculated based on the wind and sand flow intensity and soil moisture to determine the real-time engineering progress information of the target area. Step 4: Based on the engineering operation time period of the target area and the desert multi-dimensional environmental data during the engineering operation time period, and combined with the wind-sand-precipitation coupled spatiotemporal sequence prediction model, obtain the simulated prevention and control progress information of the target area with environmental confidence intervals; Step 5: Based on the simulated prevention and control progress information and combined with the dynamic environmental sensitivity coefficient of the desert prevention and control stage, the real-time engineering progress information is corrected to obtain the accurate prevention and control progress of the target area.

[0020] In this embodiment, spatial heterogeneity refers to the non-uniform and differentiated distribution of geographical elements such as desertification degree, topography, vegetation distribution, and soil properties within the target area for desertification control. This is a typical environmental characteristic of desert regions. For example, selecting a 100-square-kilometer area in a desert... The target area for prevention and control is characterized by the following features: the northern part of the area is a region of severe mobile dunes (with a desertification rate of over 90%), the southern part is a region of mild semi-fixed dunes (with a desertification rate of 30%-50%), the western part is a region of saxaul shrubs and desert plants, and the eastern part is a region of bare sand. The dunes in the region are mainly oriented in a northwest-southeast direction, with slopes ranging from 5° to 20°. These characteristics represent the spatial heterogeneity of the region.

[0021] In this embodiment, remote sensing images of the target area are collected in real time on a monthly basis. The wind-sand-precipitation coupled spatiotemporal sequence prediction model is matched with the real-time project progress on a monthly time scale. It should be noted that the model uses precipitation substitute indicators for calculation in extremely arid areas where the multi-year average monthly precipitation is 0.

[0022] Remote sensing monitoring equipment refers to equipment capable of acquiring remote sensing images of desert areas, including spaceborne remote sensing equipment, airborne remote sensing equipment, and ground-based remote sensing equipment, which respectively acquire optical and radar remote sensing images.

[0023] Optical remote sensing images refer to images formed by remote sensing monitoring equipment through receiving electromagnetic waves in the optical bands such as visible light, near infrared, and short-wave infrared reflected from ground objects. They can clearly reflect the surface features of ground objects such as vegetation cover and soil color. Radar remote sensing images refer to images formed by remote sensing monitoring equipment through actively emitting microwaves and receiving echo signals from ground objects. They have the characteristics of penetrating sand and dust and being unaffected by weather. They can reflect the deep features of ground objects such as topography, soil moisture content, and sand-fixing layer structure.

[0024] The desertification control plan refers to a comprehensive implementation plan formulated for the desertification control target area, which includes the division of control stages, core governance objectives, engineering layout, control requirements, and expected ecological restoration indicators. For example, if the control plan for the target area is a 5-year implementation plan, it is planned to include the sand fixation stage (years 1-2), the vegetation restoration stage (years 3-4), and the ecological stability stage (year 5). The core governance objectives are to reduce the degree of desertification to below 60% in the sand fixation stage, increase the vegetation coverage to above 25% in the vegetation restoration stage, and form a stable psammophytic plant community in the ecological stability stage.

[0025] The ecological evolution pattern of deserts refers to the phased changes in ecological elements such as soil properties, microbial communities, wind and sand activity, and vegetation cover over time in desert areas under the intervention of sand fixation projects. For example, in the target desert area, under the intervention of sand fixation projects, the ecological evolution shows the pattern of bare sand fixation - soil improvement - vegetation settlement - community stability. In the sand fixation stage, the sand fixation capacity of the soil gradually increases and the frequency of wind and sand activity gradually decreases. In the vegetation restoration stage, the soil moisture content and microbial biomass gradually increase, and psammophytic plants gradually settle in.

[0026] The growth cycle of desert plants refers to the staged growth pattern of desert plants from seedling stage to mature stage. The growth cycle of desert plants is closely related to the ecological evolution of deserts, and their growth status is an important indicator of desert ecological restoration. For example, Haloxylon ammodendron, which is planted in the target area, is a typical desert plant. Its growth cycle is seedling stage, mature stage and mature stage. During the seedling stage, Haloxylon ammodendron grows slowly and has shallow roots, so it needs to rely on sand fixation projects. During the mature stage, Haloxylon ammodendron grows rapidly and has deep roots, so it can fix sand and improve the soil on its own.

[0027] Dynamic target prevention and control information refers to prevention and control target information that is dynamically adjusted according to desert ecological evolution and desert plant growth based on prevention and control plans and the calculation results of coupled matching algorithms. It includes the core governance targets, engineering layout, control requirements and quantitative indicators of the current prevention and control stage. For example, if the first year of the target area is the sand fixation stage, coupled with the growth characteristics of Haloxylon ammodendron seedlings, the dynamic target prevention and control information is that the core governance target is to construct a straw checkerboard sand fixation layer, the engineering layout is to densely arrange straw checkerboards in areas with severe mobile dunes, the control requirements are to monitor the integrity rate of the sand fixation layer every month, and the quantitative indicators are that the integrity rate of the sand fixation layer is ≥90% and the frequency of wind and sand activity is reduced by 30%.

[0028] Actual prevention and control information refers to the information on the actual implementation effect of desertification prevention and control projects in the target area obtained by analyzing remote sensing images through a desert multi-source data fusion and inversion model.

[0029] The intensity of wind and sand movement refers to the degree of erosion of desert wind and sand in the target area on prevention and control projects, psammophytes, and soil surface. It is a core environmental factor affecting the effectiveness of desertification control and is characterized by quantitative indicators. For example, the intensity of wind and sand movement in the northern part of the target area with severe mobile dunes is strong, with a corresponding value of 0.8, while the intensity of wind and sand movement in the southern part with mild semi-fixed dunes is weak, with a corresponding value of 0.3.

[0030] Soil moisture refers to the water content of desert soil in the target area, including indicators such as soil volumetric water content and soil capillary water content.

[0031] In this embodiment, the quantitative indicators of actual prevention and control information are compared one by one with the quantitative indicators of dynamic target prevention and control information. The completion rate of each indicator is calculated, the weight of each indicator is determined by the analytic hierarchy process, the comprehensive completion rate is calculated, and then the desert prevention and control effectiveness attenuation coefficient is introduced for correction to obtain real-time project progress information.

[0032] The project operation period refers to the duration of operation of a desertification control project from its inception to the present. It can also refer to the time span of each stage of the planned control project. It serves as the time dimension for progress prediction. For example, if the control project in the target area is launched in January 2025, the project operation period from January to December 2025 is one year, corresponding to the planned sand fixation stage (years 1-2).

[0033] Desert multidimensional environmental data refers to various environmental monitoring data related to desertification control during the project operation period in the target area, covering multiple dimensions such as wind and sand, precipitation, soil, and plant phenology.

[0034] The wind-sand-precipitation coupled spatiotemporal sequence prediction model is based on long short-term memory network and geographically weighted regression algorithm. Historical environmental data and engineering progress data of the target area are selected as training samples to train and optimize the model.

[0035] The environmental confidence interval refers to the range of confidence set for simulating prevention and control progress information, taking into account the volatility and spatial heterogeneity of desert environmental data. It provides the upper and lower limits of the progress value with a confidence level of 95%.

[0036] Simulated prevention and control progress information refers to the project progress prediction information for the future prevention and control stage of the target area obtained by using a wind-sand-precipitation coupled spatiotemporal sequence prediction model, combined with the project operation period and multi-dimensional environmental data. The inclusion of an environmental confidence interval is an important feature. For example, the simulated prevention and control progress information for the second year of sand fixation in the target area, predicted by the model, is as follows: sand fixation layer integrity rate 92%, vegetation coverage 7%, overall progress completion rate 90%, and environmental confidence interval [85%, 95%].

[0037] Desertification control effectiveness attenuation coefficient Calculation method: Where α is the wind and sand erosion influence coefficient, which is calibrated according to desert type: 0.09 for mobile desert, 0.08 for semi-mobile desert, 0.03 for semi-fixed desert, and 0.01 for fixed desert, and is dimensionless; The monthly average wind and sand erosion index is obtained by normalizing the frequency of wind and sand activity and the concentration of dust. It is dimensionless and ranges from 0 to 1. t is the duration of the prevention and control project. β is the time decay index, which is determined according to the degree of desertification, i.e., 0.9 for severe desertification, 0.75 for moderate desertification, and 0.6 for mild desertification. The actual soil moisture in the target area; Soil moisture levels in the target prevention and control area of ​​Haloxylon ammodendron planting area =1.5%, Caragana korshinskii planting area =1.2%, sea buckthorn planting area =2.0%; t0 is the planning baseline duration for the target prevention and control stage, such as sand fixation stage t0=24 months, vegetation restoration stage t0=24 months, and ecological stability stage t0=12 months.

[0038] Dynamic environmental sensitivity coefficient Calculation method: ,in, The coefficients are used as the basic coefficients for each prevention and control stage, such as 1.2 for the sand fixation stage, 0.9 for the vegetation restoration stage, and 0.5 for the ecological stabilization stage. They are dimensionless. The Normalized Index of Monthly Precipitation (NRI) is derived from the ratio of monthly average precipitation to the multi-year average precipitation of the region. It is dimensionless and ranges from 0 to 1. It should be noted that: Sand fixation stage: vegetation coverage < 15%, sand fixation layer integrity rate ≥ 80%; Vegetation recovery stage: 15% ≤ vegetation coverage < 30%, sand fixation layer integrity rate ≥ 90%; Ecological stability stage: vegetation coverage ≥ 30%, forming a stable psammophytic plant community (dominant species account for ≥ 60%). = Actual monthly precipitation / Average monthly precipitation in the target area over many years. If the actual monthly precipitation is greater than the average monthly precipitation over many years, then... =1. It should be noted that soil moisture is used as a substitute for precipitation index for areas with an average monthly precipitation of 0 over many years.

[0039] Precision prevention and control progress The calculation formula is as follows: ,in, For real-time project progress; w0 is the environmental impact weight, which is set to 0.7 for the prevention and control stage, 0.5 for vegetation restoration, and 0.3 for ecological stability, and the weight allocation is determined based on the actual measured data of the project. The real-time adjustment coefficient for engineering construction is obtained by normalizing construction efficiency and material arrival rate. It is dimensionless and ranges from 0.8 to 1.2. The confidence weights for the environmental confidence intervals are obtained by normalizing the confidence weights of the simulated progress confidence intervals. They are dimensionless and range from 0.9 to 1.0.

[0040] It should be noted that, When ≤5%, =1; 5% < When ≤8%, =0.95; 8 < When ≤10%, =0.9; If the rate is greater than 10%, the simulation and prediction of the prevention and control progress should be carried out again.

[0041] The beneficial effects of the above technical solution are as follows: By accurately acquiring and preprocessing remote sensing images based on the spatial heterogeneity of deserts, the problems of poor image acquisition targeting and excessive interference information in existing technologies are solved; by determining dynamic target prevention and control information through the coupling of desert ecological evolution and the growth cycle of psammophytes, the problem of static target information in existing technologies is solved; by acquiring actual prevention and control information through the fusion and inversion of desert multi-source data, and combining the prevention and control effectiveness attenuation coefficient dynamically calculated by wind and sand erosion and soil moisture, the calculation accuracy of real-time project progress information is improved; by integrating the spatiotemporal sequence prediction model of wind and sand-precipitation coupling relationship, simulated prevention and control progress information with environmental confidence intervals is obtained, improving the reference value of prediction results; by combining the progress correction with the dynamic environmental sensitivity coefficient of desert prevention and control stages, real-time progress correction is achieved, and finally, accurate prevention and control progress is obtained. This effectively realizes the dynamic, accurate, and desert environment-adapted remote sensing monitoring of the entire process of desertification prevention and control projects, and effectively improves the matching degree between monitoring results and actual project progress and desert environmental changes.

[0042] This invention proposes a remote sensing monitoring method for desertification control projects, step 1, including: Differentiated sub-regions are divided based on the location information and desertification level of the target area: severely desertified areas are divided with a higher density according to the first preset scale, and moderately and lightly desertified areas are divided in a regular manner according to the second preset scale. The center point of each sub-region is determined as the collection point by adaptive offsetting the vector offset algorithm based on the slope and orientation of the dune terrain combined with the desert terrain features. Using remote sensing monitoring equipment, with the collection point as the center point, sub-remote sensing images of each sub-region are collected at a resolution matched to the degree of desertification in the sub-region; Each sub-remote sensing image is seamlessly stitched together according to geographic coordinates to obtain the remote sensing image of the target area; The remote sensing images of the target area are sequentially subjected to atmospheric correction, geometric correction, dust removal, and noise removal based on anisotropic filtering of desert sand texture direction to obtain preprocessed remote sensing images.

[0043] Location information refers to the geographic coordinates of the target area for desertification control and its internal regions, including latitude and longitude coordinates, Gaussian projection coordinates, etc.

[0044] The degree of desertification is classified into three levels based on indicators such as the proportion of desertified area, dune activity, and vegetation coverage. The industry generally classifies it into three levels: severe, moderate, and mild. For example, a severe desertification area is an area where mobile dunes account for more than 90% and vegetation coverage is less than 5%; a moderate desertification area is an area where semi-mobile dunes account for 50%-90% and vegetation coverage is 5%-15%; and a mild desertification area is an area where semi-fixed dunes account for more than 50% and vegetation coverage is more than 15%.

[0045] Differentiated sub-region division refers to the division of the target area into sub-regions based on its location information and desertification level, using different scales. Severely desertified areas are further subdivided, while moderately and lightly desertified areas are divided using conventional methods. Among these: The first preset scale is for severely desertified areas: it is determined according to the monitoring accuracy level. When the monitoring accuracy is ≤3m, the first preset scale is 300m×300m; when 3m < monitoring accuracy ≤5m, the first preset scale is 500m×500m; when the monitoring accuracy is >5m, the first preset scale is 800m×800m. The criteria for judging severely desertified areas are that the proportion of mobile sand dunes is more than 90% and the vegetation coverage is less than 5%.

[0046] The second preset scale is for areas with moderate to mild desertification: when the monitoring accuracy is ≤5m, the second preset scale is 800m×800m; when the monitoring accuracy is >5m, the second preset scale is 1000m×1000m; the criteria for judging moderate desertification areas are that semi-mobile sand dunes account for 50%-90% and vegetation coverage is 5%-15%, and the criteria for judging mild desertification areas are that semi-fixed sand dunes account for more than 50% and vegetation coverage is more than 15%.

[0047] Scale adjustment principle: When the average height of sand dunes in the target area is >10m, the scale shall be reduced by 20% based on the above; when the average height of sand dunes is ≤10m, the basic scale shall be followed.

[0048] The acquisition point refers to the core location for remote sensing image acquisition determined after correction by the vector offset algorithm. Image acquisition is carried out with the acquisition point as the center point to ensure that the acquired image contains effective prevention and control information.

[0049] Vector offset coordinates of acquisition points : The offset is calculated based on the slope and orientation of the sand dunes to ensure that the data collection points avoid dune ridges / shifting sand areas. The coordinates are planar geographic coordinates. ,in, 1 represents the original coordinates of the center point of the sub-region; d represents the basic offset distance, which is determined according to the dune height. If the dune height is greater than 5m, then d is 20m; if it is less than or equal to 5m, then d is 10m. The offset coordinate calculation avoids the dune ridge and the shifting sand area; θ represents the azimuth angle of the dune direction; γ represents the slope of the windward slope of the dune.

[0050] Anisotropic filter weight coefficients Calculation method: ,in, The angle between the main direction of the sand texture and the direction of the filter window, with a value range of 0° to 90°; The basic filter weights are calibrated by the image noise level: 0.8 for high noise, 0.5 for low noise, and 0.65 for medium noise. They are dimensionless. The image signal-to-noise ratio (SNR) < 20dB is considered high noise, 20dB ≤ SNR ≤ 30dB is considered medium noise, and SNR > 30dB is considered low noise.

[0051] Matching resolution refers to the differentiated resolution set for image acquisition based on the degree of desertification in the sub-region. High resolution is used for severely desertified areas, while conventional resolution is used for moderately and lightly desertified areas.

[0052] Sub-remote sensing images refer to remote sensing images of corresponding sub-regions acquired with each acquisition point as the center point and at a matched resolution. They are the basic units that constitute the overall remote sensing image of the target area.

[0053] Seamless image stitching refers to the process of stitching together sub-regional remote sensing images based on their geographic coordinate information, resulting in a seamless, unified remote sensing image of the target area without overlap or gaps.

[0054] Atmospheric correction refers to the image processing steps that remove the influence of atmospheric scattering, absorption, reflection and other factors on optical remote sensing images and restore the true reflectance of ground objects. Geometric correction refers to the image processing steps that remove geometric distortions caused by factors such as sensor attitude, terrain undulation, and Earth curvature in remote sensing images and make the geographic coordinates of the image consistent with the actual location of ground objects. Dust removal refers to the image processing steps that remove dust interference information from remote sensing images and increase the proportion of effective information in the image. This is an existing technology.

[0055] The spectral characteristics of desert dust refer to the unique reflection and absorption characteristics of dust in desert areas across different spectral bands. Dust has high reflectivity in the visible light band and gradually decreases in the near-infrared band. For example, desert dust has a reflectivity of 0.35 in the blue light band (450nm), 0.4 in the red light band (650nm), and 0.3 in the near-infrared band (850nm).

[0056] The direction of desert sand ripples refers to the regular ripple pattern formed by wind and sand action in desert areas. The direction of desert sand ripples is consistent with the prevailing wind direction and is a typical feature of desert texture. For example, if the prevailing wind direction in the target desert area is northwest, the direction of the sand ripples will be northwest-southeast.

[0057] Anisotropic filtering based on the sand texture direction refers to a filtering method that uses direction-adaptive filtering to denoise remote sensing images according to the texture direction of desert sand ripples. It can remove noise while preserving the texture features of sand ripples. For example, for the sand ripple texture direction in the northwest-southeast direction of the target area, anisotropic filtering is used to denoise radar remote sensing images, removing speckle noise while preserving the texture features of sand ripples without destroying the topographic information of sand dunes.

[0058] Noise removal refers to an image processing step that uses anisotropic filtering based on the sand texture direction of the desert to remove noise information from remote sensing images, such as salt-and-pepper noise in optical images and speckle noise in radar images, thereby improving image clarity. For example, noise removal of Sentinel-1 radar remote sensing images can eliminate speckle noise and make the spatial differences in soil moisture content in the image clearer.

[0059] The beneficial effects of the above technical solution are as follows: By dividing the region into differentiated sub-regions according to the degree of desertification and determining the acquisition points based on the vector offset of the dune terrain, accurate acquisition of remote sensing images is achieved, avoiding areas with no effective acquisition value; by acquiring sub-images with resolution matched according to the degree of desertification and seamlessly stitching them together, both acquisition accuracy and overall integrity are taken into account; through basic processing such as atmospheric correction, geometric correction, and dust removal, combined with anisotropic filtering of the desert sand texture direction, image preprocessing of the desert environment is achieved, improving the targeting and processing quality of remote sensing images, providing a high-quality data source for subsequent inversion of actual prevention and control information, and further improving the accuracy of the monitoring method.

[0060] This invention proposes a remote sensing monitoring method for desertification control projects, step 2 of which includes: According to the coupling matching algorithm, multiple progressive prevention and control stages of the target area are extracted from the prevention and control plan, and the core governance objectives, wind and sand adaptable engineering layout and dynamic control requirements of each prevention and control stage are determined. The dynamic control requirements are adjusted in real time based on soil moisture and monthly average wind and sand activity data. Based on the project operation time, actual construction progress data, and remote sensing monitoring ecological response data of the desertification control project, the current control stage corresponding to the target area is determined, and the dynamic governance objectives, wind and sand adaptable engineering layout, and real-time adjusted control requirements of the current control stage are determined as the dynamic target control information of the target area.

[0061] In this embodiment, the progressive prevention and control stage refers to the desertification prevention and control stage extracted from the prevention and control plan based on the coupling and matching results of the desert ecological evolution law and the growth cycle of desert plants. The stages are progressively advanced in time sequence and the governance goals are progressively advanced. Each stage has continuity and phased characteristics. For example, the progressive prevention and control stages extracted from the 5-year prevention and control plan of the target area are the sand fixation stage, the vegetation restoration stage, and the ecological stability stage. The governance goal is to gradually advance from building a sand fixation layer to forming a stable desert plant community.

[0062] The core governance objective refers to the core work objective of each progressive prevention and control stage. It serves as the implementation guide for the prevention and control project at that stage and has clear quantitative characteristics. For example, the core governance objective of the sand fixation stage is to build a stable sand-fixing layer and reduce the intensity of wind and sand activity; the core governance objective of the vegetation restoration stage is to increase the coverage of psammophytic plants and improve soil properties; and the core governance objective of the ecological stability stage is to form a self-sustaining psammophytic plant community and achieve stable restoration of the desert ecosystem.

[0063] Aeolian-adaptive engineering layout refers to a spatial layout of desertification control projects specifically designed based on the aeolian characteristics of the target area, such as the intensity of sand flow, prevailing wind direction, and dune orientation. This ensures the engineering layout adapts to the aeolian environment and enhances the control effect. In a target area where the prevailing wind direction is northwest, and the intensity of sand flow is stronger in the north and weaker in the south, the aeolian-adaptive engineering layout involves densely arranging straw checkerboard sand-fixing belts along a northwest-southeast direction in the severely desertified northern area, planting sea buckthorn windbreaks along the leading edge of the sand-fixing belts, and scattering straw checkerboards and planting Haloxylon ammodendron and Caragana korshinskii in the lightly desertified southern area.

[0064] Dynamic management and control requirements refer to the management and control requirements for prevention and control projects that are adjusted in real time based on soil moisture and monthly average wind and sand activity data in the target area. These requirements include monitoring frequency, engineering maintenance standards, and requirements for replanting and reconstruction.

[0065] Monthly average wind and sand activity data refers to the monthly monitoring data related to wind and sand activity in the target area, including the frequency of wind and sand activity, dust concentration, and near-surface wind speed.

[0066] Actual construction progress data refers to the actual construction data of desertification control projects in the target area, including project start time, construction area, project completion rate, and material input. It is an important basis for determining the current control stage. For example, the actual construction progress data for the first year of the control project in the target area is as follows: the construction area of ​​the straw checkerboard sand fixation project is 80... The completion rate is 80%, and the planting area of ​​Haloxylon ammodendron is 5. The completion rate is 50%.

[0067] Ecological response data refers to the response data generated by the desert ecosystem in the target area after the construction of the prevention and control project, including vegetation coverage, soil moisture content, frequency of wind and sand activity, etc. It is obtained through remote sensing monitoring and is an important basis for determining the current prevention and control stage. For example, the remote sensing monitoring ecological response data of the target area one year after construction is as follows: vegetation coverage 5%, average soil moisture content 1.0%, and frequency of wind and sand activity reduced by 25% compared with before construction.

[0068] The current prevention and control stage refers to the current desertification prevention and control stage of the target area, which is determined comprehensively based on the project operation time, actual construction progress data, and ecological response data. It is the core basis for determining dynamic target prevention and control information. For example, if the project in the target area has been in operation for 1 year, the actual construction progress completion rate is 80%, and the ecological response data has reached the mid-term indicators of the sand fixation stage, the current prevention and control stage is determined to be the first year of the sand fixation stage.

[0069] In this embodiment, the calculation formula for the coupling matching algorithm is: ,in, The value ranges from 0 to 1, representing the degree of matching between the stages of desert ecological evolution and the growth cycle of psammophytic plants. Let be the normalized value of the i-th ecological evolution index; is the normalized value of the i-th indicator of the growth cycle of psammophytes; n is the number of indicators, ranging from 4 to 6, and the data of each indicator are processed using the extreme value normalization method.

[0070] Control requirements adjustment coefficient ,in, =0.6 is the soil moisture weight; =0.4 represents the weight of wind and sand activity, and its weight allocation is determined based on the analytic hierarchy process. The fuzzy membership degree of soil moisture is calculated by the difference between actual soil moisture and suitable soil moisture through a triangular fuzzy function. It is dimensionless and ranges from 0 to 1.5. The fuzzy membership degree of wind and sand activity is calculated by the difference between the actual wind and sand erosion index and the critical value through a triangular fuzzy function. It is dimensionless and ranges from 0 to 1.5.

[0071] , where x is the actual soil volumetric water content; , where x is the comprehensive index of wind and sand movement intensity.

[0072] In this embodiment, >1.2 indicates stricter control, with monitoring frequency increased by 50% and engineering maintenance cycle shortened by 50%; 0.8≤ ≤1.2 is considered normal control level; A value <0.8 indicates weak control, requiring a 30% reduction in monitoring frequency and a 30% extension in engineering maintenance cycles.

[0073] In this embodiment, the dynamic target prevention and control indicators are updated monthly, and the indicators are adjusted when the wind and sand erosion index or soil moisture deviates from the threshold by ±20%.

[0074] The beneficial effects of the above technical solution are as follows: by extracting progressive prevention and control stages through a coupling matching algorithm, and combining the design of wind and sand-adaptive engineering layouts with the wind and sand environment, the prevention and control stages and engineering layouts are more in line with the coupling characteristics of desert ecology and psammophytic plants; by adjusting the dynamic control requirements based on soil moisture and monthly average wind and sand activity data, the control requirements are updated in real time; by comprehensively judging the engineering operation time, construction progress data and ecological response data, the current prevention and control stage is accurately determined, and targeted dynamic target prevention and control information is extracted, making the determination of dynamic target prevention and control information more scientific, real-time and targeted, and further improving the dynamic adaptability of the monitoring method.

[0075] This invention proposes a remote sensing monitoring method for desertification control projects, step 3 of which includes: The preprocessed remote sensing images are input into the desert multi-source data fusion inversion model to obtain the actual prevention and control information of the target area. The actual prevention and control information includes vegetation cover information and soil information. The soil information is fused with multi-dimensional data of desert soil moisture, sand-fixing layer structure and water content and spatial correlation is established. The vegetation cover information of the target area is extracted from the dynamic target prevention and control information, and multiple dynamic threshold-type target prevention and control indicators of the target area are determined based on the vegetation cover information of the target area and the monthly average environmental data of the desert. The actual prevention and control information is compared with the dynamic threshold-type target prevention and control indicators, and combined with the desert prevention and control effectiveness decay coefficient to determine the real-time project progress information.

[0076] In this embodiment, the basic framework of the desert multi-source data fusion and inversion model is: convolutional neural network (CNN) + fully connected layer, with the input layer consisting of vegetation feature bands (red band, near-infrared band) from optical remote sensing images + soil feature bands (C band, X band) from radar remote sensing images, for a total of 4 input dimensions; Network structure: 2 convolutional layers: the first convolutional layer has 32 convolutional kernels with a kernel size of 3×3, and the second convolutional layer has 64 convolutional kernels with a kernel size of 3×3; the pooling layer uses max pooling with a pooling kernel size of 2×2; 2 fully connected layers: the number of neurons is 128 and 64 respectively, and the output layer has two output dimensions: vegetation cover information and soil information. Sample type: Field monitoring data of the target area + remote sensing image interpretation data. Field monitoring data includes vegetation coverage, soil volumetric water content and sand fixation layer integrity rate. Remote sensing image interpretation data is the characteristic band gray value of optical / radar remote sensing image. Sample size: ≥1000 samples per target area, covering all desertification level sub-areas of the target area, and each sample group contains a one-to-one correspondence between geographic coordinates, remote sensing feature values, and field monitoring values; Sample preprocessing: The field monitoring data were normalized to a value range of 0 to 1. The remote sensing feature values ​​were radiometrically calibrated and atmospherically corrected. Outliers, i.e. samples that deviated from the mean by 3 times the standard deviation, were removed.

[0077] In this embodiment, multidimensional correlation data refers to soil moisture as the water content of desert soil, sand fixation layer structure as the spatial structure, layer thickness, integrity and other characteristics of sand fixation projects such as straw checkerboard and chemical sand fixation layer, and water content as the volumetric water content of soil and sand fixation layer. Multidimensional correlation data refers to the fused data after matching the three types of data according to geographic coordinates and establishing spatial correlation, which can comprehensively reflect the soil sand fixation capacity. For example, the multidimensional correlation data of the severely desertified area in the northern part of the target area is poor soil moisture (volume water content 0.8%), straw checkerboard sand fixation layer structure is 1m×1m checkerboard layout, layer thickness 15cm, integrity rate 88%, sand fixation layer water content 0.9%, and soil water content is positively correlated with sand fixation layer integrity rate.

[0078] The monthly average environmental data of the desert refers to the environmental monitoring data collected monthly in the target area, covering aspects such as wind and sand, precipitation, soil, and vegetation. It is an important basis for determining dynamic threshold-type target prevention and control indicators and is matched with the time dimension of the project progress. For example, the monthly average environmental data of the desert in a certain month during the sand fixation stage of the target area is: monthly average wind and sand erosion index 0.7, monthly average precipitation 4 mm, monthly average soil moisture content 1.0%, and monthly average growth of Haloxylon ammodendron seedling height 0.4 cm.

[0079] Dynamic threshold-based target prevention and control indicators refer to quantitative prevention and control indicators that are adjusted in real time according to environmental changes based on vegetation cover information and monthly average desert environmental data. They are dynamic reference standards for judging the actual prevention and control effectiveness, which are different from fixed threshold indicators. For example, during the sand fixation stage of the target area, the dynamic threshold of the sand fixation layer integrity rate is lowered from 90% to 85% and the dynamic threshold of soil moisture content is lowered from 1.2% to 1.0% based on the increase of the wind and sand erosion index in the current month.

[0080] The beneficial effects of the above technical solution are as follows: By clarifying the two core dimensions of vegetation and soil in actual prevention and control information, and integrating multi-dimensional correlation data of soil moisture, sand-fixing layer structure and water content, the representation of actual prevention and control information is more comprehensive and accurate; by determining dynamic threshold-type target prevention and control indicators based on vegetation cover information and monthly average desert environmental data, the prevention and control indicators are dynamically matched with the desert environment, avoiding the problem of fixed threshold indicators being out of touch with the actual environment; by accurately comparing actual prevention and control information with dynamic threshold indicators, and combining the dynamically calculated prevention and control effectiveness decay coefficient to determine real-time project progress, the calculation accuracy of real-time project progress information and its adaptability to the actual desert environment are further improved.

[0081] This invention proposes a remote sensing monitoring method for desertification control projects, which determines multiple dynamic threshold-type target control indicators for the target area, including: The vegetation cover information of the target area is extracted from the dynamic target prevention and control information. The vegetation cover information includes: spatial constraint information of the key zone for windbreak and sand fixation that integrates the movement law of desert wind and sand flow; improved normalized vegetation index for psammophytic plants that introduces canopy porosity and sand burial coefficient; gridded vegetation type constraint parameters and gridded vegetation density constraint parameters that combine the spatial distribution of soil moisture content. Based on the vegetation cover information, regional characteristics of the target area, monthly average environmental data of the desert, and the intensity of wind and sand movement and the thickness of the soil sand-fixing layer, multiple dynamic threshold-type target prevention and control indicators for the target area are determined.

[0082] In this embodiment, the normalized vegetation index The calculation formula is as follows: ,in, The NDVI value is obtained from traditional multispectral remote sensing inversion. It is dimensionless and ranges from -1 to 1. The porosity of the psammophytic plant canopy is obtained from texture analysis of remote sensing images. It is dimensionless and ranges from 0 to 0.6. The sand burial coefficient is a dimensionless value that is obtained by the ratio of the plant's sand burial height to its plant height, and ranges from 0 to 0.8.

[0083] The movement pattern of desert sand flow refers to the movement characteristics of sand flow in desert areas under the action of wind, including the direction of sand flow, sand transport volume, near-ground movement height, and flow characteristics affected by topography. For example, the sand flow movement pattern in the target desert area is dominated by northwest wind, and the sand transport volume in the 0-50cm height near the ground accounts for 90% of the total sand transport volume. When it encounters the dune ridge, it will flow around the dune and form vortices on the leeward slope of the dune.

[0084] The spatial constraint information of the key zone for windbreak and sand fixation refers to the spatial range, geographical coordinates, width, and orientation of the area that plays a core role in windbreak and sand fixation, determined by integrating the movement patterns of desert wind and sand flow. For example, in the target area, combined with the pattern that the northwest wind dominates the sand flow, the key zone for windbreak and sand fixation is determined to be distributed along the northwest-southeast direction, with a width of 500m, covering the sand dune front area of ​​the severely desertified area in the north, with geographical coordinates of 104.0°E-104.2°E and 38.0°N-38.5°N. This area is the core area for wind and sand flow input.

[0085] The porosity of the canopy of psammophytes refers to the ratio of the pore area of ​​the canopy to the total area of ​​the canopy.

[0086] The sand burial coefficient refers to the ratio of the sand burial height of psammophytic plants to the total height of the plant.

[0087] The gridded vegetation type constraint parameters refer to the parameters that limit the types of psammophytic plants suitable for planting in each grid after the target area is divided into grids, based on the spatial distribution of soil moisture content. This ensures that the vegetation type is compatible with the soil environment. For example, after the target area is divided into 500m×500m grids, the western grid has a soil moisture content of 1.5%, and the vegetation type constraint parameters are Haloxylon ammodendron and Caragana korshinskii; the eastern grid has a soil moisture content of 0.8%, and the vegetation type constraint parameters are drought-resistant psammophytic plants such as Hippophae rhamnoides and Calligonum mongolicum.

[0088] The gridded vegetation density constraint parameter refers to the parameter determined based on the spatial distribution of soil moisture content in each grid after the target area is divided into grids, limiting the planting density of psammophytic plants within that grid to ensure that the vegetation density is compatible with the carrying capacity of the soil environment. For example, if the soil moisture content in the western grid of the target area is 1.5%, the gridded vegetation density constraint parameter for Haloxylon ammodendron is 3 plants / square meter. The soil moisture content in the eastern grid is 0.8%, and the gridded vegetation density constraint parameter for Haloxylon ammodendron is 1 plant / .

[0089] The regional characteristics of the target area refer to the inherent regional characteristics of the target area, such as topography, desertification degree, climate type, and soil type. For example, the regional characteristics of the desert target area are temperate continental climate, mainly mobile and semi-mobile sand dunes, desertification degree is more severe in the north and less severe in the south, soil type is aeolian sandy soil, and organic matter content is less than 0.5%.

[0090] The wind speed at a height of 10m near the ground, the average monthly sand transport, and the number of days of continuous sandstorm activity were selected as evaluation indicators. The weights were determined by the analytic hierarchy process (AHP) and were 0.4, 0.3, and 0.3, respectively. The comprehensive index of sandstorm movement intensity was calculated. Specifically, after normalizing each indicator, the sandstorm movement intensity was calculated using the weighted sum method. A sandstorm movement intensity ≥ 0.7 was considered high intensity, 0.3 ≤ sandstorm movement intensity < 0.7 was considered medium intensity, and sandstorm movement intensity < 0.3 was considered low intensity.

[0091] The thickness of the soil sand-fixing layer refers to the vertical thickness of the sand-fixing layer (such as straw checkerboard or chemical sand-fixing layer) constructed in desertification control projects.

[0092] The beneficial effects of the above technical solutions are as follows: By clarifying that vegetation cover information includes four core components—spatial constraint information of key zones for windbreak and sand fixation, improved normalized vegetation index, gridded vegetation type, and density constraint parameters—the representation of vegetation cover information is made more aligned with the actual needs of desert windbreak and sand fixation, solving the problem of the single representation method of traditional vegetation cover information. By introducing canopy porosity and sand burial coefficient of psammophytic plants to correct the normalized vegetation index, the accuracy of monitoring the growth status of psammophytic plants is improved. By combining vegetation cover information, inherent regional characteristics, monthly average environmental data with wind and sand flow intensity and soil sand-fixing layer thickness to determine dynamic threshold indicators, the determination of indicators is made more scientific and comprehensive, further improving the matching degree between dynamic threshold-type target prevention and control indicators and the actual desert environment and prevention and control projects.

[0093] This invention proposes a remote sensing monitoring method for desertification control projects to determine real-time project progress information, including: According to the preset rules for differentiating desertification levels, the target area is divided into multiple grids, and based on the vegetation cover information and soil information, the vegetation richness, vegetation density and soil moisture content of each grid are calculated. Based on the prevention and control plan and the coordinate information of each grid, the topographic features and wind erosion level of each grid are determined, and based on the coordinate information of each grid, the target vegetation richness, target vegetation density and target soil moisture content of each grid are extracted from the dynamic threshold-type target prevention and control indicators. Based on the topographic influence coefficient and wind erosion level coefficient of each grid, an orthogonal two-dimensional weight matrix is ​​constructed. Calculate the first difference between the vegetation richness and the target vegetation richness, the second difference between the vegetation density and the target vegetation density, and the third difference between the soil moisture content and the target soil moisture content for each grid. Based on the orthogonal two-dimensional weight matrix of each grid, the first difference, second difference, and third difference of the grid are weighted and corrected to obtain the grid engineering progress information of each grid. Connecting the center points of every two adjacent grids yields the topology of the project progress network for the target region; The topology, edge features, and node features of the project progress network are input into a pre-trained neural network with a desert spatial heterogeneity attention layer to obtain the first project progress information of the target area. The edge features are the length and angle of the connecting lines in the topology, and the node features are the grid project progress information of each grid. From the optical remote sensing images of the target area, dust interference information is extracted by combining dust concentration inversion with image texture analysis and using a vector analysis algorithm based on dust concentration gradient and image texture direction, and the dust interference weight and interference direction of each project progress item in the first project progress information are determined. Based on the sandstorm interference weight and direction of each project progress, interference correction is performed on each project progress in the first project progress information, and a second correction is performed based on the desert control effectiveness attenuation coefficient to obtain the real-time project progress information of the target area.

[0094] The preset rules for differentiating desertification levels refer to the rules for different grid division scales and grid sizes based on the desertification level of the target area. The grid scale is smaller in severely desertified areas and larger in moderately and lightly desertified areas, so that the grid division is adapted to the degree of desertification.

[0095] vegetation richness ,in, This represents the actual number of types. Number of target categories; This represents the normalized value of plant growth status.

[0096] The topographic influence coefficient refers to a coefficient determined based on the topographic characteristics of each grid, such as dune slope, dune type, and whether it is a ridge area. It characterizes the degree of influence of topography on the progress of the prevention and control project; a larger coefficient indicates a more significant influence. For example, if a grid in the target area is a dune ridge area with a slope of 20°, the topographic influence coefficient is 0.9; if another grid is a gentle dune slope area with a slope of 5°, the topographic influence coefficient is 0.3. Implementation method: Based on the topographic characteristics of the grids, a basic coefficient is set for different terrain types, and then corrected for by dune slope. The topographic influence coefficient for each grid is calculated, with a range of 0-1.

[0097] The wind and sand erosion level coefficient refers to the coefficient determined according to the wind and sand erosion level (severe, moderate, and mild) of each grid, which characterizes the degree of impact of wind and sand erosion on the progress of prevention and control projects. The coefficient is positively correlated with the wind and sand erosion level. For example, the wind and sand erosion level coefficient of the severe wind and sand erosion grid is 1.0, that of the moderate grid is 0.6, and that of the mild grid is 0.2. The standard for the wind and sand erosion level is set in advance.

[0098] In this embodiment, the orthogonal two-dimensional weight matrix ,in, The topographic and geomorphological influence coefficient is obtained by normalizing the proportion of sand dunes and slope within the grid. It is dimensionless and ranges from 0 to 1. The wind and sand erosion level coefficient is determined by the grid wind and sand erosion level (1 for severe, 0.6 for moderate, and 0.3 for mild), and is dimensionless. The orthogonalization angle is set to 45°, which is derived from the statistical regularity of desert terrain characteristics.

[0099] In this embodiment, the sandstorm interference weight With the direction of interference : ; ; in, L represents the dust concentration gradient; L is the average length of the sand texture within the grid, in meters; L0 is the preset base length, with a value of 50 meters. denoted as the partial derivatives of dust concentration in the X and Y directions; Min is the minimum value of the global dust concentration gradient; Max is the maximum value of the global dust concentration gradient. In this embodiment, Where k is the calibration coefficient, which is fitted by the on-site dust concentration monitoring data of the target area, and k=0.85; Reflectivity in the blue and red light bands; The relative concentration of dust; ; It should be noted that, if If the value is greater than 0.8, then take 0.8 to avoid excessive interference from a single grid.

[0100] In this embodiment, the pre-trained neural network with the desert spatial heterogeneity attention layer is achieved by introducing a desert spatial heterogeneity factor to correct the node attention weights of the graph neural network.

[0101] ,in, Let be the attention weight of the i0th grid node, which is dimensionless; The feature vectors of the i0th and j2th nodes represent the grid engineering progress information; The attention layer weight matrix is ​​obtained from desert engineering samples; is the desert spatial heterogeneity factor of the i0th node, which is obtained by normalizing the degree of desertification of the grid and the terrain complexity. It is dimensionless and its value ranges from 0 to 1; N0 is the total number of nodes in the graph neural network.

[0102] The first difference, the second difference, and the third difference refer to the differences between the actual vegetation richness and the target value, the actual vegetation density and the target value, and the actual soil moisture content and the target value, respectively. A negative difference indicates that the actual value has not reached the target value, and a positive difference indicates that the actual value has exceeded the target value. All of them are dimensionless relative differences.

[0103] The topology of a project schedule network refers to a network structure that uses each grid as a node and connects the center points of adjacent grids to form edges, reflecting the spatial relationships of project progress within the target area. This structure visually represents the spatial connections between grids. For example, if the target area is divided into 800 grids, and the geometric center point of each grid is used as a node, connecting adjacent nodes to form straight edges creates an undirected project schedule network topology with 800 nodes and 1500 edges. Adjacent grids are determined using a dual rule of Euclidean distance and geographical connectivity. Euclidean distance: The Euclidean distance between grid center points is ≤ 1.5 times the preset scale of the corresponding desertification area. For example, when the preset scale of a severely desertified area is 500m, the Euclidean distance is ≤ 750m. Geographic connectivity: There are no obvious geographical barriers such as dune ridges or shifting sand zones between grids. The geographical connectivity threshold is defined as dune ridge height ≥ 5m.

[0104] A grid is considered adjacent if it meets both of the above conditions; if it only meets one of the conditions, it is not considered adjacent.

[0105] Edge features are the length and angle of connecting lines in the project progress network topology, reflecting the spatial distance and relative orientation of adjacent grids; node features are the grid project progress information corresponding to each grid, reflecting the core attributes of the node. For example, a connecting line with a length of 200m and an angle of 315° (northwest-southeast) is an edge feature; the grid project progress completion rates of the two nodes corresponding to this connecting line are 85% and 90% respectively, which are node features. It should be noted that the grid project progress completion rate = , in, The values ​​are the first, second, and third differences; w01, w02, and w03 are the correction weights, and the values ​​of w01, w02, and w03 determined based on the entropy weight method are 0.4, 0.3, and 0.3, respectively.

[0106] In this embodiment, the basic framework of the spatial heterogeneous attention layer neural network is: Graph Neural Network (GNN) + Desert Spatial Heterogeneous Attention Layer. The input layer consists of node features (grid engineering progress information) and edge features (connection line length and angle), for a total of 5 input dimensions. The network structure consists of: 1 spatial heterogeneous attention layer (4 attention heads), 2 GNN layers (64 and 32 hidden layer dimensions), 1 fully connected layer (32 neurons), and the output layer is the overall engineering progress completion rate of the target area, with 1 output dimension. Sample type: gridded engineering progress data + geographic feature data of the target area. The engineering progress data is the difference between the actual and target values ​​of the grid vegetation richness, vegetation density and soil moisture content. The geographic feature data is the grid topography and geomorphology influence coefficient and wind and sand erosion level coefficient. Sample size: ≥800 samples per target area, with each sample group corresponding to one grid, covering all gridded sub-regions of the target area; Sample preprocessing: Normalize all feature data with values ​​ranging from 0 to 1, and construct the topology sample of the project progress network; The neural network uses a loss function value ≤ 0.05 as the convergence condition, and the edge / node features are processed by linear normalization.

[0107] The first project progress information refers to the overall project progress information of the target area obtained after inputting the topology, edge features, and node features of the project progress network into a neural network with a heterogeneous attention layer in desert space, without considering the impact of sandstorm interference. For example, the first project progress information of the target area calculated by the neural network shows an overall progress completion rate of 88%, a progress completion rate of 82% in the severely desertified northern area, and a progress completion rate of 95% in the lightly desertified southern area.

[0108] Vector analysis algorithms combine spatial variations in dust concentration gradients with the texture direction of desert images to extract dust interference information using vector analysis methods. This approach can accurately characterize the range and extent of dust interference on project progress. For example, using this algorithm, the area with a large dust concentration gradient in the north of the target area's optical remote sensing image can be identified as the core area of ​​dust interference, with the interference direction being southeast.

[0109] Dust disturbance information refers to various types of information extracted using vector analysis algorithms that characterize the interference of dust on the progress of dust control projects, including the core area of ​​interference, interference intensity, and interference direction. For example, the dust disturbance information for the target area is as follows: the core area of ​​interference is the severely desertified northern area, the interference intensity is 0.8, and the interference direction is southeast.

[0110] In this embodiment, the first completion rate of each project progress is combined with the sandstorm interference weight and interference direction to calculate the corrected completion rate, thereby correcting the interference of the first project progress information. The project progress completion rate after interference correction is multiplied by the desert control effectiveness attenuation coefficient to complete the second correction and obtain the real-time project progress information of the target area.

[0111] The beneficial effects of the above technical solution are as follows: By dividing the grid according to the degree of desertification, the calculation of the project progress is more in line with the spatial heterogeneity of the desert; by constructing an orthogonal two-dimensional weight matrix to correct the grid index difference, the independent weight correction of topography and wind and sand factors is realized, improving the calculation accuracy of grid project progress information; by constructing a project progress network topology and introducing a neural network with a desert spatial heterogeneity attention layer, the spatial correlation characteristics of the project progress are fully explored, making the overall progress calculation more spatially reasonable; by extracting sand and dust interference information and correcting the interference through vector analysis algorithm, and combining it with the attenuation coefficient of prevention and control effectiveness for secondary correction, double error elimination is achieved, and the accuracy of the final real-time project progress information is greatly improved, which can more accurately reflect the actual progress of desertification prevention and control projects.

[0112] This invention proposes a remote sensing monitoring method for desertification control projects, step 4 of which includes: The multi-dimensional environmental data of the desert in the target area during the engineering operation period are divided according to the prevention and control stage. The multi-dimensional environmental data includes the frequency of wind and sand activity, dust concentration, soil moisture, precipitation distribution and phenological data of desert plants. The environmental data of each stage are preprocessed by spatiotemporal fusion. The preprocessed environmental information of each prevention and control stage is input into a pre-trained wind-sand-precipitation coupled spatiotemporal sequence prediction model, and the simulated prevention and control progress information with environmental impact confidence intervals for each prevention and control stage is output. The wind-sand-precipitation coupled spatiotemporal sequence prediction model is implemented by integrating long short-term memory network and geographical weighted regression algorithm.

[0113] In this embodiment, the phenological data of psammophytes refers to the data reflecting the phenological characteristics of psammophytes in different growth cycles, including the time nodes of budding period, leaf unfolding period, flowering period, fruiting period, and dormancy period, as well as plant height, canopy, root growth parameters, etc. in each phenological period. It is the core data reflecting the growth status of psammophytes.

[0114] Precipitation distribution refers to the spatial and temporal distribution characteristics of precipitation in the target area during the project operation period, including monthly average precipitation, regional precipitation differences, and precipitation period distribution, and is an important component of multidimensional environmental data in the desert.

[0115] Spatiotemporal fusion preprocessing refers to the preprocessing operation of multi-dimensional environmental data divided according to prevention and control stages, which integrates time series features and spatial distribution features to eliminate the interference of temporal fluctuations and spatial heterogeneity of the data, making the data more suitable for input into the prediction model.

[0116] Preprocessed environmental information refers to multi-dimensional desert environmental data that has undergone spatiotemporal fusion preprocessing. It integrates time series features and spatial distribution features, significantly improving the completeness and standardization of the data, and serves as the input data for the prediction model.

[0117] In this embodiment, the algorithm for the wind-sand-precipitation coupled spatiotemporal sequence prediction model is as follows: ,in, To simulate the predicted progress of prevention and control; is the dimensionless value extracted by the LSTM network from the temporal features of the environmental data sequence; GWR() is the fitted value of the geographic weighted regression on the spatial features. The phenological growth attenuation coefficient for psammophytes is determined by taking 0.2 for the dormant period, 0.05 for the growing period, and 0.3 for the withering period, based on psammophyte phenological observation data.

[0118] Environmental impact confidence interval ,in, ,in, The confidence interval half-width is set to 1.96 with a confidence level of 95%. The standard deviation of the model prediction residuals is calculated from historical prediction data; The desert spatial heterogeneity coefficient is determined by the variation coefficients of topography and desertification degree of the target area grid. It is dimensionless and ranges from 1.0 to 1.5.

[0119] In this embodiment, the basic framework of the wind-sand-precipitation coupled spatiotemporal sequence prediction model is: Long Short-Term Memory Network (LSTM) + Geographically Weighted Regression (GWR). LSTM extracts temporal features, GWR fits spatial features, and a fusing layer of psammophyte phenological features is added in the middle. LSTM hyperparameters: input layer dimension 5 (corresponding to 5 types of environmental data), hidden layers 2 (64 and 32 neurons respectively), time step 12 months, dropout coefficient 0.2 (to prevent overfitting), activation function tanh, loss function MSE, optimizer Adam, learning rate 0.001.

[0120] GWR hyperparameters: the spatial weighting function is a Gaussian kernel function, the bandwidth is determined by cross-validation, and the regression term is the time feature value output by LSTM + phenological data of psammophytes, for a total of 2 regression factors; Phenological Feature Fusion Layer: Incorporating the phenological growth attenuation coefficient of psammophytes As attention weights, they are multiplied with the LSTM output features and then input into the GWR model.

[0121] The sample types for this wind-sand-precipitation coupled spatiotemporal sequence prediction model are: historical multi-dimensional environmental data of the target area + historical engineering progress data. The environmental data include the frequency of wind and sand activities, dust concentration, soil moisture, precipitation distribution, and phenological data of psammophytes. The engineering progress data is the comprehensive completion rate of each prevention and control stage.

[0122] The beneficial effects of the above technical solution are as follows: by dividing multi-dimensional environmental data according to the prevention and control stages and performing spatiotemporal fusion preprocessing, the temporal fluctuations and spatial heterogeneity interference of the data are eliminated, providing high-quality input data for the prediction model; by integrating long short-term memory networks and geographic weighted regression algorithms, the prediction model simultaneously takes into account the temporal regularity of project progress, the spatial heterogeneity of desert areas, and the phenological regularity of desert plant growth, solving the problem of traditional prediction models ignoring plant phenological regularity; the simulated prevention and control progress information with environmental impact confidence intervals output by the model not only improves the accuracy of the prediction results but also provides a reliable reference for the progress results, significantly enhancing the engineering practice reference value of the simulated prevention and control progress information.

[0123] This invention proposes a remote sensing monitoring method for desertification control projects, step 5 of which includes: The real-time project progress information is compared with the corresponding simulated prevention and control progress information with environmental confidence intervals in multiple dimensions to obtain quantitative progress deviation information covering progress completion rate, ecological effectiveness achievement rate, and project stability. The quantitative progress deviation information is compared with the dynamic deviation range determined based on the characteristics of the prevention and control stage and real-time desert environmental data; When there are deviation data in the quantified schedule deviation information that exceed the corresponding dynamic deviation range, the core stage of the schedule deviation and the source of deviation transmission are located based on the spatial correlation characteristics of the project schedule. Based on the core deviation stage, the source of deviation transmission, and combined with the dynamic environmental sensitivity coefficient and real-time adjustment coefficient of engineering construction at that stage, the real-time engineering progress information is corrected to obtain the precise prevention and control progress of the target area.

[0124] In this embodiment, the ecological effectiveness compliance rate refers to the proportion of actual ecological effectiveness indicators in the target area, such as vegetation coverage, soil moisture content, and wind and sand activity reduction rate, that reach the dynamic threshold-type target prevention and control indicators. For example, there are a total of 5 ecological effectiveness indicators in the sand fixation stage of the target area, of which 4 reach the dynamic threshold, and the ecological effectiveness compliance rate is 80%.

[0125] In this embodiment, engineering stability = 0.5 × stability of sand fixation layer integrity rate + 0.5 × stability of plant survival rate.

[0126] The dynamic deviation range refers to the reasonable fluctuation range of quantitative progress deviation information, determined based on the characteristics of the prevention and control stage and real-time desert environmental data. It is dynamically adjusted according to the prevention and control stage and environmental data; exceeding this range indicates significant progress deviation. For example, during the sand fixation stage in the target area, due to strong wind and sand activity, the dynamic deviation range for the progress completion rate is set at -8% to +5%, and the dynamic deviation range for the ecological effectiveness achievement rate is set at -10% to +5%. During the vegetation restoration stage, when the environment is relatively stable, the dynamic deviation range for the progress completion rate is set at -5% to +5%. It should be noted that the deviation range for the ecological effectiveness achievement rate is 2% larger than the progress completion rate.

[0127] Spatial correlation characteristics of project progress refer to the spatial correlation between project progress in each grid and region of the target area, including spatial clustering characteristics, spatial transmission characteristics, and mutual influence characteristics between regions. It is the core basis for deviation tracing.

[0128] The core deviation stage refers to the core prevention and control stage that caused the progress deviation, as determined through source tracing, when the quantified progress deviation information exceeds the dynamic deviation range. It serves as the time dimension for correction. For example, if the quantified progress deviation information in the target area exceeds the dynamic deviation range, and the core deviation stage is determined to be the first year of the sand fixation stage, the progress is delayed because wind and sand activity in this stage exceeds expectations.

[0129] The source of deviation transmission refers to the core grid or region that, after being traced back to its origin, causes the progress deviation and transmits it to other areas. It serves as the spatial dimension basis for correction. For example, if the source of deviation transmission in the target area is a grid group in a severely desertified area in the north, the progress deviation in this area is caused by the construction quality of the sand fixation layer, and it is transmitted to the southeast.

[0130] The real-time adjustment coefficient for construction projects is determined based on the real-time conditions of the construction project, such as construction efficiency and material arrival rate, as shown in Table 1: Table 1 Real-time Adjustment Coefficients for Engineering Construction

[0131] It should be noted that: construction efficiency = actual completed construction area / planned construction area × 100%; material arrival rate = actual material arrival quantity / planned material quantity × 100%.

[0132] In this embodiment, the intensity of deviation transmission in the project progress between grids is quantified, and the source of deviation transmission is located. The intensity of deviation transmission... ,in, The engineering spatial correlation between the i3rd and j3rd grids; Let be the schedule deviation rate of the i3rd grid. k is the center distance between the i3rd and j3rd grids; k0 is the number of grids adjacent to the i3rd grid.

[0133] It should be noted that the sand-fixing zone is a connected grid: =1.0; Grids within the same vegetation restoration area: =0.8; Adjacent grids without engineering connectivity: =0.5; Non-adjacent grids: =0.

[0134] Deviation propagation source determination rule: Calculate the deviation propagation intensity of all grids within the target area. Select Grids with a value ≥0.8 are considered core deviation grids, and the set of core deviation grids represents the source of deviation propagation; if If the value is less than 0.8, it is determined that there is no obvious source of transmission, and the deviation is caused by local grid construction problems.

[0135] The beneficial effects of the above technical solution are as follows: By comparing real-time and simulated progress from three dimensions—progress completion rate, ecological effectiveness achievement rate, and project stability—more comprehensive quantitative progress deviation information is obtained; by determining the dynamic deviation range based on the characteristics of the prevention and control stage and real-time desert environmental data, deviation judgment is made more consistent with the actual desert environment and project stage characteristics; by locating the core deviation stage and the source of deviation transmission through the spatial correlation characteristics of project progress, precise location and tracing of progress deviations are achieved; by correcting based on the deviation tracing results, combined with the dynamic environmental sensitivity coefficient and the real-time adjustment coefficient of project construction, progress correction is made more targeted and accurate. The final accurate prevention and control progress can reflect the actual progress of desertification prevention and control projects to the greatest extent, providing a precise decision-making basis for adjusting the project plan.

[0136] This invention provides a remote sensing monitoring system for desertification control projects, such as... Figure 2 As shown, it includes: The image preprocessing module is used to acquire remote sensing images of the target area in real time using remote sensing monitoring equipment and based on the spatial heterogeneity characteristics of desertification areas, and to preprocess the remote sensing images of the target area, wherein the remote sensing images include optical remote sensing images and radar remote sensing images. The dynamic prevention and control module is used to determine the dynamic target prevention and control information of the target area based on the prevention and control plan of the desertification prevention and control project in the target area, combined with the coupling matching algorithm of desert ecological evolution law and psammophyte growth cycle. The real-time engineering determination module is used to analyze the preprocessed remote sensing images based on the desert multi-source data fusion inversion model to obtain the actual prevention and control information of the target area. Based on the actual prevention and control information and the dynamic target prevention and control information, the module introduces the desert prevention and control effectiveness attenuation coefficient dynamically calculated based on the wind and sand flow intensity and soil moisture to determine the real-time engineering progress information of the target area. The simulation prevention and control module is used to obtain the simulation prevention and control progress information of the target area with environmental confidence intervals based on the engineering operation time period of the target area and the desert multi-dimensional environmental data during the engineering operation time period, and combined with the wind-sand-precipitation coupled spatiotemporal sequence prediction model. The progress correction module is used to correct the real-time project progress information based on the simulated prevention and control progress information and combined with the dynamic environmental sensitivity coefficient of the desert prevention and control stage to obtain the accurate prevention and control progress of the target area.

[0137] In this embodiment, standardized data interfaces are used between modules, with data transmission latency ≤12 hours, meeting the real-time monitoring requirements of desertification control projects, and supporting switching between offline / online computing modes; the system hardware is supported by industrial-grade servers (CPU ≥32 cores, GPU ≥4 cards, memory ≥128G), supporting parallel processing of large amounts of data, data storage adopts a distributed database, historical data is stored for ≥10 years, and a built-in data backup strategy is provided to ensure data security.

[0138] The beneficial effects of the above technical solution are: it effectively realizes the dynamic, precise and desert environment-adapted remote sensing monitoring of the entire process of desertification control project, and effectively improves the matching degree between monitoring results and actual project progress and desert environment changes.

[0139] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A remote sensing monitoring method for desertification control projects, characterized in that, include: Step 1: Real-time acquisition of remote sensing images of the target area using remote sensing monitoring equipment based on the spatial heterogeneity characteristics of desertification areas, and preprocessing of the remote sensing images of the target area, wherein the remote sensing images include optical remote sensing images and radar remote sensing images. Step 2: Based on the desertification control project in the target area, and combined with the coupling matching algorithm of desert ecological evolution law and psammophyte growth cycle, determine the dynamic target control information of the target area; Step 3: Analyze the preprocessed remote sensing images based on the desert multi-source data fusion inversion model to obtain the actual prevention and control information of the target area. Based on the actual prevention and control information and the dynamic target prevention and control information, introduce the desert prevention and control effectiveness attenuation coefficient dynamically calculated based on the wind and sand flow intensity and soil moisture to determine the real-time engineering progress information of the target area. Step 4: Based on the engineering operation time period of the target area and the desert multi-dimensional environmental data during the engineering operation time period, and combined with the wind-sand-precipitation coupled spatiotemporal sequence prediction model, obtain the simulated prevention and control progress information of the target area with environmental confidence intervals; Step 5: Based on the simulated prevention and control progress information and combined with the dynamic environmental sensitivity coefficient of the desert prevention and control stage, the real-time engineering progress information is corrected to obtain the accurate prevention and control progress of the target area.

2. The remote sensing monitoring method for desertification control projects according to claim 1, characterized in that, Step 1 includes: Differentiated sub-regions are divided based on the location information and desertification level of the target area: severely desertified areas are divided with a higher density according to the first preset scale, and moderately and lightly desertified areas are divided in a regular manner according to the second preset scale. The center point of each sub-region is determined as the collection point by adaptive offsetting the vector offset algorithm based on the slope and orientation of the dune terrain combined with the desert terrain features. Using remote sensing monitoring equipment, with the collection point as the center point, sub-remote sensing images of each sub-region are collected at a resolution matched to the degree of desertification in the sub-region; Each sub-remote sensing image is seamlessly stitched together according to geographic coordinates to obtain the remote sensing image of the target area; The remote sensing images of the target area are sequentially subjected to atmospheric correction, geometric correction, dust removal, and noise removal based on anisotropic filtering of desert sand texture direction to obtain preprocessed remote sensing images.

3. The remote sensing monitoring method for desertification control projects according to claim 1, characterized in that, Step 2 includes: According to the coupling matching algorithm, multiple progressive prevention and control stages of the target area are extracted from the prevention and control plan, and the core governance objectives, wind and sand adaptable engineering layout and dynamic control requirements of each prevention and control stage are determined. The dynamic control requirements are adjusted in real time based on soil moisture and monthly average wind and sand activity data. Based on the project operation time, actual construction progress data, and remote sensing monitoring ecological response data of the desertification control project, the current control stage corresponding to the target area is determined, and the dynamic governance objectives, wind and sand adaptable engineering layout, and real-time adjusted control requirements of the current control stage are determined as the dynamic target control information of the target area.

4. The remote sensing monitoring method for desertification control projects according to claim 3, characterized in that, Step 3 includes: The preprocessed remote sensing images are input into the desert multi-source data fusion inversion model to obtain the actual prevention and control information of the target area. The actual prevention and control information includes vegetation cover information and soil information. The soil information is fused with multi-dimensional data of desert soil moisture, sand-fixing layer structure and water content and spatial correlation is established. The vegetation cover information of the target area is extracted from the dynamic target prevention and control information, and multiple dynamic threshold-type target prevention and control indicators of the target area are determined based on the vegetation cover information of the target area and the monthly average environmental data of the desert. The actual prevention and control information is compared with the dynamic threshold-type target prevention and control indicators, and combined with the desert prevention and control effectiveness decay coefficient to determine the real-time project progress information.

5. The remote sensing monitoring method for desertification control projects according to claim 4, characterized in that, Determine multiple dynamic threshold-based target prevention and control indicators for the target area, including: The vegetation cover information of the target area is extracted from the dynamic target prevention and control information. The vegetation cover information includes: spatial constraint information of the key zone for windbreak and sand fixation that integrates the movement law of desert wind and sand flow; improved normalized vegetation index for psammophytic plants that introduces canopy porosity and sand burial coefficient; gridded vegetation type constraint parameters and gridded vegetation density constraint parameters that combine the spatial distribution of soil moisture content. Based on the vegetation cover information, regional characteristics of the target area, monthly average environmental data of the desert, and the intensity of wind and sand movement and the thickness of the soil sand-fixing layer, multiple dynamic threshold-type target prevention and control indicators for the target area are determined.

6. The remote sensing monitoring method for desertification control projects according to claim 4, characterized in that, Determine real-time project progress information, including: According to the preset rules for differentiating desertification levels, the target area is divided into multiple grids, and based on the vegetation cover information and soil information, the vegetation richness, vegetation density and soil moisture content of each grid are calculated. Based on the prevention and control plan and the coordinate information of each grid, the topographic features and wind erosion level of each grid are determined, and based on the coordinate information of each grid, the target vegetation richness, target vegetation density and target soil moisture content of each grid are extracted from the dynamic threshold-type target prevention and control indicators. Based on the topographic influence coefficient and wind erosion level coefficient of each grid, an orthogonal two-dimensional weight matrix is ​​constructed. Calculate the first difference between the vegetation richness and the target vegetation richness, the second difference between the vegetation density and the target vegetation density, and the third difference between the soil moisture content and the target soil moisture content for each grid. Based on the orthogonal two-dimensional weight matrix of each grid, the first difference, second difference, and third difference of the grid are weighted and corrected to obtain the grid engineering progress information of each grid. Connecting the center points of every two adjacent grids yields the topology of the project progress network for the target region; The topology, edge features, and node features of the project progress network are input into a pre-trained neural network with a desert spatial heterogeneity attention layer to obtain the first project progress information of the target area. The edge features are the length and angle of the connecting lines in the topology, and the node features are the grid project progress information of each grid. From the optical remote sensing images of the target area, dust interference information is extracted by combining dust concentration inversion with image texture analysis and using a vector analysis algorithm based on dust concentration gradient and image texture direction, and the dust interference weight and interference direction of each project progress item in the first project progress information are determined. Based on the sandstorm interference weight and direction of each project progress, interference correction is performed on each project progress in the first project progress information, and a second correction is performed based on the desert control effectiveness attenuation coefficient to obtain the real-time project progress information of the target area.

7. The remote sensing monitoring method for desertification control projects according to claim 1, characterized in that, Step 4 includes: dividing the multi-dimensional environmental data of the desert in the target area during the engineering operation period into prevention and control stages. The multi-dimensional environmental data includes the frequency of wind and sand activity, dust concentration, soil moisture, precipitation distribution and phenological data of desert plants. The environmental data of each stage is then subjected to spatiotemporal fusion preprocessing. The preprocessed environmental information of each prevention and control stage is input into a pre-trained wind-sand-precipitation coupled spatiotemporal sequence prediction model, and the simulated prevention and control progress information with environmental impact confidence intervals for each prevention and control stage is output. The wind-sand-precipitation coupled spatiotemporal sequence prediction model is implemented by integrating long short-term memory network and geographical weighted regression algorithm.

8. The remote sensing monitoring method for desertification control projects according to claim 7, characterized in that, Step 5 includes: The real-time project progress information is compared with the corresponding simulated prevention and control progress information with environmental confidence intervals in multiple dimensions to obtain quantitative progress deviation information covering progress completion rate, ecological effectiveness achievement rate, and project stability. The quantitative progress deviation information is compared with the dynamic deviation range determined based on the characteristics of the prevention and control stage and real-time desert environmental data; When there are deviation data in the quantified schedule deviation information that exceed the corresponding dynamic deviation range, the core stage of the schedule deviation and the source of deviation transmission are located based on the spatial correlation characteristics of the project schedule. Based on the core deviation stage, the source of deviation transmission, and combined with the dynamic environmental sensitivity coefficient and real-time adjustment coefficient of engineering construction at that stage, the real-time engineering progress information is corrected to obtain the precise prevention and control progress of the target area.

9. A remote sensing monitoring system for desertification control projects, characterized in that, include: The image preprocessing module is used to acquire remote sensing images of the target area in real time using remote sensing monitoring equipment and based on the spatial heterogeneity characteristics of desertification areas, and to preprocess the remote sensing images of the target area, wherein the remote sensing images include optical remote sensing images and radar remote sensing images. The dynamic prevention and control module is used to determine the dynamic target prevention and control information of the target area based on the prevention and control plan of the desertification prevention and control project in the target area, combined with the coupling matching algorithm of desert ecological evolution law and psammophyte growth cycle. The real-time engineering determination module is used to analyze the preprocessed remote sensing images based on the desert multi-source data fusion inversion model to obtain the actual prevention and control information of the target area. Based on the actual prevention and control information and the dynamic target prevention and control information, the module introduces the desert prevention and control effectiveness attenuation coefficient dynamically calculated based on the wind and sand flow intensity and soil moisture to determine the real-time engineering progress information of the target area. The simulation prevention and control module is used to obtain the simulation prevention and control progress information of the target area with environmental confidence intervals based on the engineering operation time period of the target area and the desert multi-dimensional environmental data during the engineering operation time period, and combined with the wind-sand-precipitation coupled spatiotemporal sequence prediction model. The progress correction module is used to correct the real-time project progress information based on the simulated prevention and control progress information and combined with the dynamic environmental sensitivity coefficient of the desert prevention and control stage to obtain the accurate prevention and control progress of the target area.

Images