A multi-modal prairie ecological restoration image feature extraction and identification system

Abstract

The present application relates to the technical field of multi-modal image, and more particularly to a multi-modal grassland ecological restoration image feature extraction and identification system.The system comprises a data acquisition module, a feature extraction module, a grassland degradation identification module, and an effect evaluation module.The data acquisition module is used to acquire multi-source monitoring data of the grassland area, and performs geographic registration based on the multi-source monitoring data to construct an ecological monitoring dataset.The feature extraction module is used to extract vegetation coverage features using the ecological monitoring dataset, calculate community canopy density based on the vegetation coverage features, evaluate the precipitation interception rate based on the community canopy density, and determine the grassland vegetation water retention capacity based on the precipitation interception rate.The grassland degradation identification module is used to determine the soil water retention index based on the grassland vegetation water retention capacity.The present application realizes grassland degradation grade identification and ecological restoration effect evaluation based on multi-modal image technology, thereby improving the accuracy and scientificity of grassland ecological restoration monitoring.

Description

Technical Field

[0001] This invention relates to the field of multimodal image technology, and in particular to a multimodal grassland ecological restoration image feature extraction and recognition system. Background Technology

[0002] In the process of grassland ecological restoration monitoring and degradation assessment, vegetation structure, eco-hydrological processes, and soil wind erosion evolution have a complex coupling relationship, which is a key factor affecting the accuracy of degradation identification and restoration effect evaluation. However, existing technologies still have shortcomings: First, they are mostly based on single indicators such as vegetation cover or remote sensing indices, lacking the ability to comprehensively analyze structural characteristics such as vegetation patch connectivity, canopy light transmittance, and community canopy closure, making it difficult to accurately reflect the spatial continuity and restoration status of vegetation; second, they lack the ability to analyze the correlation between eco-hydrological processes such as chlorophyll enrichment, photosynthetic intensity, canopy closure level, and precipitation interception capacity, making it difficult to reveal the intrinsic connection between vegetation physiological activity and water retention function; third, they lack the ability to comprehensively analyze soil water content, pore water storage characteristics, capillary adhesion, and wind erosion resistance, making it difficult to accurately identify wind erosion-sensitive areas and degradation levels, resulting in a lack of systematicness and dynamism in the assessment of ecological restoration effects. Summary of the Invention

[0003] Therefore, it is necessary for the present invention to provide a multimodal grassland ecological restoration image feature extraction and recognition system to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a multimodal grassland ecological restoration image feature extraction and recognition system includes the following modules:

[0005] The data acquisition module is used to collect multi-source monitoring data from grassland areas; and to perform georeferencing based on the multi-source monitoring data to construct an ecological monitoring dataset.

[0006] The feature extraction module is used to extract vegetation cover features from the ecological monitoring dataset; calculate community canopy closure based on the vegetation cover features; assess the precipitation interception rate based on the community canopy closure; and determine the water retention capacity of grassland vegetation based on the precipitation interception rate.

[0007] The grassland degradation identification module is used to determine the soil moisture retention index based on the water retention capacity of grassland vegetation; identify wind erosion-sensitive areas based on the soil moisture retention index; and determine the grassland degradation level in wind erosion-sensitive areas.

[0008] The effect evaluation module is used to calculate the vegetation coverage growth rate based on the ecological monitoring dataset; determine the reduction ratio of degraded areas based on the vegetation coverage growth rate; calculate the ecological restoration index based on the reduction ratio of degraded areas; and evaluate the grassland ecological restoration effect based on the ecological restoration index.

[0009] The beneficial effects of this invention are as follows:

[0010] (1) The data acquisition module integrates and collects multi-source monitoring data of the grassland area, and performs georegistration to construct an ecological monitoring dataset, thereby realizing the spatial consistency of data from different sensor sources and effectively solving the problem that traditional single remote sensing data cannot reflect the multi-scale ecological information of the grassland.

[0011] (2) The feature extraction module constructs multi-layer structural indicators such as community canopy closure, canopy void index and canopy closure coefficient based on vegetation coverage characteristics, and combines precipitation interception rate and water retention capacity to conduct ecological function quantitative analysis, so as to realize the synergistic characterization of grassland vegetation structure and eco-hydrological function and improve the comprehensiveness and accuracy of ecological status representation.

[0012] (3) The grassland degradation identification module constructs a soil moisture retention index based on the water retention capacity of grassland vegetation, and combines the soil pore water storage state, interparticle capillary adhesion and soil wind erosion resistance to identify wind erosion sensitive areas and classify grassland degradation levels, thereby realizing the mechanistic characterization of grassland degradation process and improving the accuracy and reliability of degradation identification.

[0013] (4) The effect evaluation module calculates the vegetation coverage growth rate based on the ecological monitoring dataset, and combines the reduction ratio of degraded areas with the ecological restoration index to achieve a quantitative evaluation of the spatial changes and ecological improvement before and after grassland restoration, thereby improving the dynamic monitoring and comprehensive evaluation capabilities of grassland ecological restoration effects. Attached Figure Description

[0014] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0015] Figure 1 This is a schematic diagram of the execution flow of the multimodal grassland ecological restoration image feature extraction and recognition system of the present invention;

[0016] Figure 2 This is a schematic diagram of a grassland satellite remote sensing image as an example.

[0017] Figure 3 This is a schematic diagram of grassland restoration as an example;

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] The technical method of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this invention.

[0020] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0021] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0022] To achieve the above objectives, please refer to Figures 1 to 3 This invention provides a multimodal grassland ecological restoration image feature extraction and recognition system. The system includes a data acquisition module, a feature extraction module, a grassland degradation recognition module, and an effect evaluation module. The system performs the following steps:

[0023] S1: Collect multi-source monitoring data of grassland areas; perform georeferencing based on the multi-source monitoring data to construct an ecological monitoring dataset;

[0024] In one embodiment, multi-source monitoring data of a grassland area is collected using a UAV multispectral camera, satellite remote sensing imagery, soil moisture monitoring sensors, and an automatic weather station. The UAV is equipped with a five-band multispectral camera to acquire images in the red, green, blue, red-edge, and near-infrared bands; satellite remote sensing imagery is used to obtain information on large-scale vegetation distribution; the soil moisture monitoring sensor is used to acquire soil moisture content data at different depths; and the automatic weather station is used to acquire precipitation, temperature, wind speed, and humidity data. GNSS positioning information is used to spatially locate various monitoring data, and ground control points are selected for coordinate correction. An affine transformation model is used to spatially register the UAV imagery and satellite imagery, and the ground monitoring data and remote sensing data are time-synchronized based on the acquisition timestamps to construct an ecological monitoring dataset.

[0025] In another embodiment, assuming the monitoring area is 36 km², the UAV flies at an altitude of 150 m, acquiring 1500 multispectral images, each with a resolution of 6000×4000 pixels; 50 ground control points are deployed, and after registration, the average positioning error is reduced from 4.1 m to 0.22 m; 40 soil moisture monitoring points are set up, including 40 monitoring points at a depth of 10 cm, 40 at a depth of 30 cm, and 40 at a depth of 50 cm, collecting data every 30 minutes; an automatic weather station continuously monitors for 45 days, acquiring 2160 sets of precipitation data, 2160 sets of wind speed data, and 4320 sets of temperature and humidity data. The system comprises approximately... One spatial pixel and approximately An ecological monitoring dataset consisting of time-series monitoring records.

[0026] S2: Extract vegetation cover features using ecological monitoring datasets; calculate community canopy closure based on vegetation cover features; assess precipitation interception rate based on community canopy closure, and determine the water retention capacity of grassland vegetation based on precipitation interception rate.

[0027] In one embodiment, assuming a monitoring area has a total of 500,000 pixels, of which 365,000 are vegetation pixels, the vegetation coverage is 73%. The calculated vegetation patch connectivity is 0.82, and the vegetation spatial continuity index is 0.79. The canopy transmittance is 28%, the canopy porosity index is 0.31, the canopy closure coefficient is 0.76, and the community canopy closure is 0.74. Assuming a cumulative precipitation of 38 mm during the monitoring period, the canopy interception is calculated to be 15.2 mm based on the canopy interception model, resulting in a precipitation interception rate of 40%. The precipitation infiltration is calculated to be 18.5 mm, the soil moisture replenishment index is 0.71, and the grassland vegetation water retention capacity index is determined to be 0.75.

[0028] S3: Determine the soil moisture retention index based on the water retention capacity of grassland vegetation; identify wind erosion-sensitive areas based on the soil moisture retention index; determine the grassland degradation level in wind erosion-sensitive areas;

[0029] In one embodiment, assuming the grassland vegetation water retention capacity index is 0.75, corresponding to a soil effective water content of 24.8%; the calculated soil pore water storage capacity is... The pore water saturation level was 0.72; the liquid film extension range was 68% of the particle surface area, the capillary bridge formation probability was 0.81, the number of capillary bridges per unit square meter was 1520, and the capillary bridge formation density was 1520 / m²; based on this, the soil capillary adhesion force was calculated to be 48.3 N / m², the soil particle cohesion was 0.79, and the soil wind erosion resistance was 0.84. Under the condition of an average wind speed of 6.8 m / s, the surface sand lifting risk index was 0.36, with high-risk areas accounting for 11%, medium-risk areas accounting for 24%, and low-risk areas accounting for 65%, indicating that the area was classified as slightly degraded.

[0030] S4: Calculate the vegetation coverage growth rate based on the ecological monitoring dataset; determine the reduction ratio of degraded areas based on the vegetation coverage growth rate; calculate the ecological restoration index based on the reduction ratio of degraded areas; and evaluate the grassland ecological restoration effect based on the ecological restoration index.

[0031] In one embodiment, vegetation cover information is extracted using ecological monitoring datasets corresponding to different monitoring periods, and the change in vegetation cover between adjacent monitoring periods is calculated. The vegetation cover growth rate is calculated based on the change in vegetation cover. The boundaries of degraded areas within each monitoring period are extracted, and the changes in degraded area area are statistically analyzed. The reduction ratio of degraded areas is determined based on the changes in degraded area area. An ecological restoration index evaluation model is constructed by combining the vegetation cover growth rate, the reduction ratio of degraded areas, and the degree of improvement in soil moisture retention index. The grassland ecological restoration effect is then assessed based on the ecological restoration index.

[0032] In another embodiment, assuming the vegetation cover was 46% before restoration and increased to 72% after restoration, the vegetation cover growth rate is 56.5%; the degraded area decreased from 14.2 km² to 7.8 km², a reduction of 45.1%; and the soil moisture retention index increased from 0.43 to 0.76, an increase of 76.7%. Calculations using the ecological restoration index model show that vegetation restoration contributed 0.38, degraded area restoration contributed 0.27, and soil improvement contributed 0.21, resulting in an ecological restoration index of 0.86. Based on the preset evaluation criteria, an ecological restoration index greater than 0.80 is considered a significant restoration level. Therefore, the ecological restoration effect of this grassland area is determined to be good, with significant improvements in vegetation structure, moisture retention capacity, and wind erosion resistance.

[0033] Preferably, the feature extraction module calculates community canopy closure based on vegetation cover features, including:

[0034] Vegetation continuous distribution areas are marked based on vegetation cover characteristics; vegetation patch connectivity is calculated based on vegetation patch connectivity; and vegetation spatial continuity index is determined based on vegetation patch connectivity.

[0035] In one embodiment, vegetation cover features are extracted using an ecological monitoring dataset, and a vegetation cover raster map is generated. Raster cells with vegetation cover greater than 30% are marked as vegetation cells, and those with vegetation cover less than 30% are marked as non-vegetation cells. An eight-neighborhood connectivity analysis method is used to identify interconnected vegetation cells, and connected regions with an area greater than 100 m² are defined as continuously distributed vegetation areas. Then, the spatial distances between continuously distributed vegetation areas are calculated; when the distance between the boundaries of two vegetation patches is less than 15 m, a connection is considered to exist. A vegetation patch connectivity network is constructed based on the connectivity relationships between vegetation patches, and the vegetation patch connectivity degree is calculated based on the ratio of the actual number of connections to the theoretical maximum number of connections. The vegetation spatial continuity index is calculated by combining the vegetation patch connectivity degree and the area ratio of continuously distributed vegetation areas to the target area to characterize the degree of spatial continuity of grassland vegetation.

[0036] In another embodiment, assuming the target grassland area is 20 km², 560,000 vegetation cover raster cells were extracted from remote sensing imagery, of which 428,000 cells had a vegetation cover greater than 30%, accounting for 76.4% of the total area. Eight-neighbor connectivity analysis yielded 72 vegetation patches, of which 53 had effective continuous vegetation distribution areas larger than 100 m². 386 connectivity relationships with boundary distances less than 15 m between patches were identified, while the theoretical maximum number of connectivity relationships for the 53 patches is 1378. Therefore, the vegetation patch connectivity is 386 / 1378 = 0.28. Combined with the 76.4% vegetation area, the calculated vegetation spatial continuity index is 0.82, indicating a high degree of vegetation contiguousness and good spatial continuity in the target area.

[0037] The degree of canopy shading is assessed based on the vegetation spatial continuity index; the shading area is marked according to the degree of canopy shading; the canopy transmittance is calculated in the shading area; the canopy void index is determined based on the canopy transmittance; the canopy closure coefficient is calculated based on the canopy void index; and the community canopy closure is calculated based on the canopy closure coefficient.

[0038] In one embodiment, assuming a vegetation spatial continuity index of 0.82, which is higher than the continuous canopy determination threshold of 0.60, the target area is determined to have a continuous canopy structure. Analysis of UAV imagery yields a canopy projection area of ​​7.85 km², corresponding to a vegetation coverage area of ​​10 km². The canopy shading level is 78.5%, higher than the 70% shading area determination threshold. Therefore, the area of ​​7.85 km² is marked as a shading area. Canopy images with a resolution of 0.05 m are acquired, yielding a total of 3,200,000 pixels. Of these, 704,000 are translucent pixels with a brightness value greater than 180, and 2,496,000 are non-translucent pixels with a brightness value not greater than 180. The canopy transmittance is calculated based on the statistical results. The canopy void index was determined to be 0.22, and the canopy closure coefficient was 1 − 0.22 = 0.78. Combined with the vegetation spatial continuity index of 0.82, the community canopy closure was calculated as: Community Canopy Closure = 0.78 × 0.82 = 0.64. The calculation results show that the community canopy closure in the target area reaches 0.64, indicating that the grassland vegetation canopy has a high degree of closure, can form a relatively stable canopy shading structure, and provides basic data support for subsequent analysis of precipitation interception and water retention capacity.

[0039] Preferably, calculating the canopy transmittance in the shaded area includes:

[0040] Acquire vegetation canopy images corresponding to the shaded areas; identify light-transmitting and non-light-transmitting areas based on the vegetation canopy images; calculate the light-transmitting area based on the light-transmitting areas; calculate the total shaded area of ​​the non-light-transmitting areas;

[0041] In one embodiment, a drone equipped with an RGB camera acquires canopy images at a height of 20m above the vegetation canopy, with a spatial resolution of 0.05m. Radiometric correction and denoising are performed on the canopy images, and then translucent and opaque areas are identified based on pixel brightness differences. Areas with brightness values ​​higher than a preset threshold of 0.65 are marked as translucent areas, while areas with brightness values ​​lower than or equal to 0.65 are marked as opaque areas. Connected component analysis is used to extract translucent patches, and the actual area corresponding to each translucent area is calculated as the translucent area. Simultaneously, the area of ​​opaque areas is calculated to obtain the total shading area. To avoid interference from weed shadows and noise, isolated areas with an area smaller than 0.02m² are removed, thus obtaining more accurate canopy structure information.

[0042] In another embodiment, assuming a grassland monitoring area of ​​10,000 m², 400 canopy images were collected. After image segmentation, 980 translucent areas were identified, with a total translucent area of ​​2,300 m²; the total non-translucent area was 7,700 m². Among these, 85 translucent areas were larger than 5 m², accounting for 62% of the total translucent area; 18 translucent areas were larger than 20 m², with the largest translucent area reaching 32 m². This indicates that the canopy cover in this area is relatively complete, but there are still some areas where light penetrates significantly.

[0043] The canopy shading rate is calculated based on the total shading area and the light-transmitting area; the canopy transmittance is determined based on the ratio of the canopy shading rate to the light-transmitting area.

[0044] In one embodiment, the proportion of the light-transmitting area to the total shading area is statistically analyzed based on the light-transmitting area and the total shading area. This proportion is then combined with the shading area proportion to assess the degree of canopy shading. Shading level thresholds are set: a shading area proportion below 40% is considered a low-shading canopy; a shading area proportion between 40% and 70% is considered a medium-shading canopy; and a shading area proportion above 70% is considered a high-shading canopy. Canopy transmittance is determined based on the light-transmitting area proportion, and transmittance level thresholds are set: transmittance above 50% is considered a high-transmittance canopy; transmittance between 25% and 50% is considered a medium-transmittance canopy; and transmittance below 25% is considered a low-transmittance canopy. Through comprehensive analysis of canopy shading degree and transmittance, a quantitative assessment of the density characteristics of the vegetation canopy structure is achieved.

[0045] In another embodiment, assuming a monitoring area has a light-transmitting area of ​​2300 m² and a total shading area of ​​7700 m², the shading area accounts for 77% of the entire canopy area, corresponding to a high shading level; the light-transmitting area accounts for 23%, corresponding to a low light-transmitting level, indicating that the vegetation in this area is relatively dense and the canopy closure is relatively high. Assuming another monitoring area has a light-transmitting area of ​​4200 m² and a total shading area of ​​5800 m², the shading area accounts for 58%, corresponding to a medium shading level; the light-transmitting area accounts for 42%, corresponding to a medium light-transmitting level. Assuming a third monitoring area has a light-transmitting area of ​​6200 m² and a total shading area of ​​3800 m², the shading area accounts for only 38%, while the light-transmitting area accounts for 62%, corresponding to a low shading and high light-transmitting level, indicating that the vegetation distribution in this area is relatively sparse and the canopy structure is relatively loose.

[0046] Preferably, identifying translucent and non-translucent areas based on vegetation canopy images includes:

[0047] The vegetation canopy image is converted to grayscale to obtain pixel grayscale values; the pixel brightness value is calculated based on the pixel grayscale value; if the pixel brightness value is greater than the preset light transmittance threshold, it is marked as a light-transmitting area; if the pixel brightness value is less than the preset light transmittance threshold, it is marked as a non-light-transmitting area.

[0048] In one embodiment, a UAV equipped with a visible light camera acquires images of the grassland canopy. The spatial resolution of the images is 0.05m, and the image size is 4000×3000 pixels. Radiometric correction and median filtering are applied to the canopy images to eliminate light fluctuations and noise interference. A grayscale algorithm is used to convert the RGB images to grayscale images, obtaining the grayscale values ​​corresponding to each pixel and normalizing them to the range of 0-1 to characterize the pixel brightness level. Then, the brightness distribution characteristics of the entire canopy image are statistically analyzed, and a light transmittance threshold is determined based on the brightness histogram. When a pixel brightness value is greater than 0.65, it is determined that there is significant light penetration at that location, and the corresponding pixel is marked as a transparent area; when a pixel brightness value is less than or equal to 0.65, it is determined that the location is blocked by vegetation leaves, and the corresponding pixel is marked as a non-transparent area. A 5×5 pixel window is used for neighborhood consistency testing. When the proportion of transparent pixels within the window exceeds 70%, the entire area is marked as a transparent area; otherwise, it is marked as a non-transparent area, thereby improving the stability and accuracy of transparent area identification.

[0049] Preferably, the feature extraction module assesses the precipitation interception rate based on the community canopy closure and determines the water retention capacity of grassland vegetation based on the precipitation interception rate, including:

[0050] Determine the canopy closure level based on community canopy closure; calculate the raindrop blocking ratio based on the canopy closure level; determine the canopy interception capacity based on the raindrop blocking ratio; obtain precipitation data; calculate the precipitation interception rate based on the canopy interception capacity and precipitation.

[0051] In one embodiment, assuming a grassland community canopy closure of 0.78, it is classified as a high closure level based on a preset threshold. Using UAV canopy imagery, the canopy coverage area is determined to be 8500 m², with a total monitoring area of ​​10000 m², corresponding to a raindrop obstruction rate of 85%. Combined with a leaf area index of 4.2, an average canopy thickness of 0.65 m, and a plant density of 180 plants / 100 m², the canopy interception capacity level is determined to be Level IV. Simultaneously, data from a rainfall event are acquired, with a cumulative rainfall of 42 mm, of which the vegetation canopy actually intercepted 15.5 mm of rainfall, resulting in a calculated rainfall interception rate of 36.9%. In another monitoring area, the community canopy closure is 0.52, the raindrop obstruction rate is 57%, the cumulative rainfall is 38 mm, and the intercepted rainfall is 9.8 mm, corresponding to a rainfall interception rate of 25.8%. Multi-regional comparisons revealed that as community canopy closure increased, both the raindrop blocking ratio and precipitation interception rate showed a significant upward trend.

[0052] The surface runoff reduction coefficient is calculated based on the precipitation interception rate; the precipitation infiltration is determined based on the surface runoff reduction coefficient; the soil moisture replenishment index is calculated based on the precipitation infiltration; and the water retention capacity of grassland vegetation is assessed based on the soil moisture replenishment index.

[0053] In one embodiment, surface runoff reduction is assessed based on the obtained precipitation interception rate. When the precipitation interception rate is greater than 35%, the surface runoff reduction coefficient is set to 0.70–0.90; when the precipitation interception rate is between 20% and 35%, the surface runoff reduction coefficient is set to 0.50–0.70; and when the precipitation interception rate is less than 20%, the surface runoff reduction coefficient is set to 0.20–0.50. The actual precipitation infiltration amount into the soil is determined by combining data on surface slope, soil texture, and precipitation intensity. The soil moisture replenishment index is calculated based on the precipitation infiltration amount, soil field capacity, and root depth. When the soil moisture replenishment index is greater than 0.75, the grassland vegetation's water retention capacity is rated as excellent; when the soil moisture replenishment index is between 0.50 and 0.75, it is rated as good; and when the soil moisture replenishment index is less than 0.50, it is rated as average or poor. This method enables the quantitative analysis of the transfer of precipitation interception effect to soil water retention capacity.

[0054] In another embodiment, assuming a precipitation interception rate of 36.9%, the corresponding surface runoff reduction coefficient is 0.82. Combining the regional average slope of 6°, sandy loam soil content of 62%, and precipitation of 42 mm, the precipitation infiltration rate is determined to be 26.4 mm. Based on the soil field capacity of 32%, average root depth of 35 cm, and initial soil moisture content of 18%, the soil moisture replenishment index is calculated to be 0.81, thus the grassland vegetation's water retention capacity is rated as excellent. In another sample area, the precipitation interception rate is 25.8%, corresponding to a surface runoff reduction coefficient of 0.61, precipitation infiltration of 17.2 mm, and a soil moisture replenishment index of 0.63, resulting in a water retention capacity rating of good. Further assuming a degraded grassland area has a precipitation interception rate of only 12.4%, corresponding to a surface runoff reduction coefficient of 0.35, precipitation infiltration of 8.6 mm, and a soil moisture replenishment index of 0.38, then the water retention capacity is rated as poor. Comparative analysis of different regions can effectively reflect the effect of improved vegetation structure on soil water storage and moisture retention capacity.

[0055] Preferably, determining the canopy closure level based on community canopy closure includes:

[0056] Dominant vegetation community areas are identified based on canopy closure; spectral absorption characteristics are extracted from the dominant vegetation community areas; and chlorophyll enrichment is calculated based on the spectral absorption characteristics.

[0057] In one embodiment, vegetation classification is performed on the target grassland area based on remote sensing imagery or UAV multispectral imagery. Dominant vegetation communities are identified by setting a canopy closure threshold. Areas with a canopy closure greater than 0.75 are classified as high-coverage dominant communities, areas with a canopy closure between 0.50 and 0.75 are classified as moderately dominant communities, and areas with a canopy closure less than 0.50 are excluded from the dominant community analysis. Within the dominant vegetation community areas, the spectral absorption characteristics of the vegetation canopy are extracted, including red light absorption intensity, near-infrared reflectance response, and red edge position shift characteristics. The spectral curves are locally smoothed to reduce noise. Chlorophyll enrichment is calculated based on absorption differences across different bands. When the red edge band shift exceeds 720 nm and near-infrared reflectance is significantly enhanced, chlorophyll enrichment is considered high; when the red edge is between 705 nm and 720 nm, it is considered moderately enriched; and when it is below 705 nm, it is considered low enriched. This achieves a graded characterization of vegetation physiological activity.

[0058] Of particular importance is the calculation of chlorophyll enrichment based on spectral absorption characteristics, which includes:

[0059] Extract the green band reflectance, red band absorbance, and near-infrared band reflectance values ​​of the spectral absorption characteristics; calculate the green intensity value of the leaf based on the green band reflectance value; calculate the chlorophyll absorption intensity value based on the red band absorbance value; and calculate the leaf cell activity value based on the near-infrared band reflectance value.

[0060] In one embodiment, reflectance data of the vegetation canopy in the green (G), red (R), and near-infrared (NIR) bands are acquired using a multispectral sensor (e.g., Mica Sense Red Edge or a similar five-band imaging device), and atmospheric correction and radiometric calibration are performed on the raw image. Based on the green band reflectance value RG, a leaf green intensity index is constructed. This is used to characterize the structural integrity of leaf chloroplasts and the explicit features of photosynthetic activity; based on the absorption characteristics of the red band, chlorophyll absorption intensity values ​​are constructed. This reflects chlorophyll's ability to absorb red light; based on near-infrared reflectance. Construct leaf cell activity value These three indicators are used to characterize the integrity of the internal cell structure of leaves and the reflectivity of the spongy tissue. They are fused at the pixel level to provide the basic input for subsequent calculations of chlorophyll accumulation.

[0061] In another embodiment, assume that normalized reflectance data were collected in a grassland ecological monitoring area as follows: green band Red band Near-infrared band The leaf green intensity value is then calculated. Approximately [0.48, 0.62] respectively; chlorophyll absorption intensity value Approximately [0.73, 0.82] respectively; leaf cell activity value The value was approximately [0.88, 1.15] in different samples. Specifically, in areas with high vegetation cover ( The leaf exhibited a high cell activity value (approximately 1.15), indicating a dense leaf structure and active photosynthetic tissue, while in the mildly degenerated areas ( The cell viability value decreased to about 0.88, indicating a trend of loosening of the leaf structure.

[0062] Of particular importance is the calculation of leaf cell viability based on near-infrared reflectance values, which includes:

[0063] The internal reflected energy of the leaf is calculated based on the near-infrared band reflectance value; the uniformity of mesophyll cell arrangement is determined based on the internal reflected energy of the leaf; and the degree of intercellular connectivity is calculated based on the uniformity of mesophyll cell arrangement.

[0064] In one embodiment, a near-infrared imaging device is used to acquire the reflectance values ​​of vegetation leaves in the near-infrared band (NIR, typically in the 800–900 nm range), and the raw reflectance data is subjected to radiometric normalization. Based on the near-infrared reflectance values... Constructing the internal reflection energy index of the blade Its expression is ,in The structural magnification factor reflects the enhancement effect of the leaf spongy tissue on near-infrared scattering. A structural magnification factor is constructed by statistically analyzing the spatial distribution of scattered energy within the leaf. This index, calculated jointly based on local variance and spatial spectral entropy, characterizes the regularity and density of mesophyll cell arrangement. Based on the spatial consistency of cell arrangement and the topological relationship of pore structure, the degree of intercellular connectivity is calculated. By using a graph structure connectivity analysis model, the intercellular spaces are equivalent to a connected network of nodes and edges, thereby characterizing the continuity of diffusion channels for gas and water inside the leaf.

[0065] In another embodiment, it is assumed that the near-infrared reflectance value of a certain herbaceous plant leaf is collected. The range is [0.48, 0.72]. The internal reflected energy of the blade is then calculated. Approximately [0.52, 0.79] (taking...) Based on the high-resolution leaf microstructure reconstruction results, the uniformity of mesophyll cell arrangement... The value is approximately [0.61, 0.88] in different samples, where the blades have a compact structure ( )correspond In leaves under mild stress, the cells are arranged regularly with uniform intercellular spacing; while in leaves under mild stress ( )middle, A decrease to approximately 0.61 indicates disordered cell arrangement. The intercellular connectivity is then calculated. Approximately [0.55, 0.93], where highly connected samples ( This indicates the presence of continuous gas exchange channels within the mesophyll tissue, while low-connectivity samples ( This manifests as localized blockage of stomata and intercellular spaces.

[0066] The leaf tissue development status is assessed based on the degree of intercellular connectivity; the leaf cell activity value is determined based on the leaf tissue development status.

[0067] In one embodiment, the degree of intercellular connectivity is... Using these as core structural input parameters, and combining them with morphological characteristics of leaf tissue development (including changes in cell wall thickness, degree of mesophyll differentiation, and air cell ratio), a leaf tissue development status evaluation model is constructed. This model, through... A hierarchical mapping was performed to divide leaf tissues into three stages: nascent development stage, stable maturity stage, and decline / degeneration stage. Simultaneously, a leaf cell metabolic activity function was introduced to classify the tissue development stages. By coupling and mapping with cellular energy metabolism levels, leaf cell activity values ​​can be obtained. It is used to characterize the photosynthetic capacity, gas exchange efficiency and cellular physiological activity of leaves.

[0068] In another embodiment, it is assumed that The value range is [0.55, 0.93], which can be divided as follows: When At that time, the leaf tissue development state was determined to be "initial developmental unstable state"; when When, it is determined to be a "stable developmental state"; when At that time, it was determined to be in a "highly mature developmental state". The leaf cell activity value was calculated. For example, in different samples: when hour, This manifests as lower cellular metabolic activity and limited photosynthetic efficiency; when hour, It exhibits a normal active state; when hour, The leaves exhibit a highly active state, demonstrating strong photosynthetic electron transport and material exchange capabilities. This indicator system allows for continuous quantitative assessment of the physiological state of the leaves.

[0069] The vegetation chlorophyll accumulation is calculated based on leaf green intensity value, chlorophyll absorption intensity value, and leaf cell activity value; the degree of chlorophyll enrichment is assessed based on the chlorophyll accumulation.

[0070] In one embodiment, the leaf green intensity value Chlorophyll absorption intensity value and leaf cell activity value We performed weighted fusion to construct a chlorophyll accumulation model: ,in , , These represent spectral sensitivity weighting coefficients, used to balance the contributions of leaf appearance greenness, absorption characteristics, and cell structure activity to chlorophyll accumulation. Based on the amount of chlorophyll accumulated... By setting a grading threshold range, the degree of chlorophyll enrichment in vegetation is divided into three levels: low, medium, and high, thereby achieving a comprehensive characterization of vegetation nutritional status and photosynthetic capacity.

[0071] In another embodiment, it is assumed that the weighting coefficients are taken. , , Calculate the following for the above samples: When , , At that time, chlorophyll accumulation ;when , , hour, ;when , , hour, According to the set hierarchical rules (e.g.: For low enrichment, For enrichment, (High enrichment) corresponds to low enrichment, medium enrichment and high enrichment states respectively, thus realizing the quantitative differentiation of different vegetation growth vitality and nutrient accumulation levels.

[0072] The photosynthetic intensity is determined based on the degree of chlorophyll enrichment; the leaf expansion capacity is calculated based on the photosynthetic intensity; the canopy cover increment is determined based on the leaf expansion capacity; and the canopy closure level is determined based on the canopy cover increment.

[0073] In one embodiment, photosynthetic intensity is graded and assessed based on chlorophyll enrichment levels. High chlorophyll enrichment corresponds to a photosynthetic intensity range of 0.75–0.90, moderate enrichment to 0.55–0.75, and low enrichment to 0.30–0.55. Leaf expansion capacity is assessed based on photosynthetic intensity combined with temperature, light duration, and soil moisture conditions. When photosynthetic intensity is high and water conditions are sufficient, leaf expansion capacity is enhanced, manifested as increased leaf elongation rate and branching density. Canopy cover increment is calculated based on leaf expansion capacity. Strong expansion capacity results in significant horizontal canopy expansion and a marked increase in cover increment; moderate expansion capacity results in slow canopy expansion; and weak expansion capacity results in a relatively stable canopy. Canopy closure level is determined based on canopy cover increment. An increment exceeding 15% is considered a high closure level, between 5% and 15% is considered a medium closure level, and less than 5% is considered a low closure level, thus achieving a continuous mapping from vegetation growth state to structural changes.

[0074] Preferably, the calculation of raindrop blocking ratio based on canopy closure level includes:

[0075] Determine the degree of leaf overlap based on the canopy closure level; calculate the projected area of ​​leaves per unit area based on the degree of leaf overlap; determine the precipitation interception area based on the projected area of ​​leaves per unit area.

[0076] In one embodiment, assuming a high canopy closure level of 0.85 in a certain grassland area, corresponding to a high degree of leaf overlap and coverage, the leaves exhibit a distinct multi-layered structure in space. Image inversion yields a leaf projection area of ​​0.82 per unit area (approximately 82% coverage), with a primary interception area of ​​approximately 5200 m² and a secondary interception area of ​​approximately 1800 m². In a medium-closure area (canopy closure level 0.68), the leaf overlap is moderate, with a leaf projection area of ​​approximately 0.64 per unit area, corresponding to a primary interception area of ​​approximately 3600 m² and a secondary interception area of ​​approximately 2900 m². In a low-closure area (canopy closure level 0.52), the leaf overlap is low, with a leaf projection area of ​​only 0.41 per unit area, a primary interception area of ​​less than 2000 m², and most areas exhibiting a clearly defined gap structure.

[0077] Acquire leaf distribution data in the precipitation interception area; calculate leaf coverage per unit area based on leaf distribution data; determine the number of raindrop collisions with the leaves based on leaf coverage per unit area; calculate the raindrop blocking ratio based on the number of raindrop collisions with the leaves.

[0078] In one embodiment, assuming that the leaf distribution data of a high-density interception area yields a leaf coverage rate of 0.86 per unit area, an average raindrop collision frequency of 7.5 times per rainfall event, and a corresponding raindrop obstruction rate of approximately 88%, the leaf coverage rate in the medium-density area is 0.67 per unit area, the average raindrop collision frequency is 4.2 times per event, and the raindrop obstruction rate is approximately 63%. In the low-density area, the leaf coverage rate is 0.43 per unit area, the average raindrop collision frequency is only 1.8 times per event, and the raindrop obstruction rate is approximately 34%. Statistical analysis of three consecutive rainfall events reveals that the raindrop obstruction rate in the high-coverage area remains stable between 85% and 90%, in the medium-coverage area it fluctuates between 60% and 68%, and in the low-coverage area it decreases to 30% to 40%, indicating that leaf distribution density directly affects the raindrop collision path and canopy interception capacity.

[0079] Preferably, the grassland degradation identification module identifies wind erosion-sensitive areas based on the soil moisture retention index, including:

[0080] Calculate the available soil moisture content based on the soil moisture retention index; calculate the soil pore water storage based on the available soil moisture content; calculate the soil capillary adhesion based on the soil pore water storage.

[0081] In one embodiment, soil sampling points in the target area were stratified to obtain soil moisture data from three layers: 0–10 cm, 10–20 cm, and 20–30 cm. The raw moisture content data were normalized to eliminate the influence of sampling depth differences. Then, the effective moisture state of each layer was determined based on the soil moisture retention index. A sliding window (window length of 5 days) was used to smooth moisture content fluctuations and eliminate abnormal abrupt changes (the threshold was set to mean ± 2 standard deviations). Based on this, the effective moisture content was mapped to the pore scale, and the pore water storage was estimated using soil pore structure parameters (porosity approximately 0.35–0.48). The capillary adhesion distribution was calculated in conjunction with the soil particle size distribution (35% sand, 45% silt, and 20% clay). The results showed that the capillary adhesion varied in gradient across different depths, with the maximum at the surface layer being approximately 1.3 times that of the middle layer.

[0082] In another embodiment, assuming the soil moisture retention index of a certain monitoring area is 0.62, the effective water content after treatment is approximately 18%–24% (22% in the surface layer, 19% in the middle layer, and 17% in the deep layer); the corresponding pore water storage is approximately per cubic meter. , , The calculated distribution of capillary adhesion in different regions is as follows: approximately 0.8 kPa in loose sandy areas, approximately 1.6 kPa in medium loam areas, and approximately 2.4 kPa in clayey areas. Overall, the higher the clay content, the stronger the capillary adhesion, and in the toe area, due to water accumulation, the local capillary adhesion increases by approximately 15%–20%.

[0083] Soil particle cohesion is assessed based on soil capillary adhesion; soil wind erosion resistance is determined based on soil particle cohesion; surface sand-raising risk index is calculated based on soil wind erosion resistance; and wind erosion-sensitive areas are identified based on the surface sand-raising risk index.

[0084] In one embodiment, capillary adhesion is spatially interpolated based on the capillary bridging mechanism between soil particles of different sizes to form a continuous distribution field, and the degree of cohesion between particles is assessed accordingly. A grading method is used to classify the cohesion into three levels: low, medium, and high, and the wind erosion resistance parameter is corrected by combining surface roughness (calculated from micro-topographic elevation differences of 5–30 cm). Under wind speed boundary conditions (range varying from 2–12 m / s), the critical state of particle initiation is simulated, and the wind erosion resistance is converted into a surface sand-raising risk index. Based on the risk index, wind erosion sensitive areas are spatially divided to predict potential desertification areas.

[0085] In another embodiment, it is assumed that the soil capillary adhesion distribution range in a certain area is 0.7 kPa–2.6 kPa, of which the low-cohesion area (<1.0 kPa) accounts for 28%, the medium-cohesion area (1.0–2.0 kPa) accounts for 49%, and the high-cohesion area (>2.0 kPa) accounts for 23%. The corresponding wind erosion resistance intensities are approximately: 3.2 m / s for low-intensity areas, 5.8 m / s for medium-intensity areas, and 8.4 m / s for high-intensity areas. When calculating the surface sand-lifting risk index, the risk value is approximately 0.82 for low-cohesion areas, approximately 0.46 for medium-intensity areas, and approximately 0.18 for high-intensity areas (the higher the value, the greater the risk). Three types of wind erosion sensitive areas were identified: highly sensitive areas are mainly concentrated in bare land and sparse vegetation areas, accounting for approximately 31% of the total area; medium sensitive areas account for 44%; and low-sensitive stable areas account for 25%, mainly distributed in fixed sand dune areas with good vegetation cover.

[0086] Preferably, the calculation of soil capillary adhesion based on soil pore water storage includes:

[0087] Assess pore water saturation based on soil pore water storage capacity; determine wetting state based on pore water saturation; delineate particle liquid film extension range based on wetting state; identify overlapping areas of particle contact interfaces based on liquid film extension range.

[0088] In one embodiment, a layered pore structure analysis was performed on the target soil sample, and the three-dimensional pore structure was reconstructed using CT scans (resolution approximately...). The pore connectivity volume distribution was obtained, and the pore volume fraction was coupled with the water content distribution field to obtain the storage state of pore water at different scales. Based on the continuity of pore water distribution, the wetting state was divided into zones, classifying the soil into non-wetting, partially wetting, and fully wetting zones. The extent of liquid film expansion on the particle surface was dynamically tracked based on the wetting continuity index. The liquid film bridging region at the particle contact interface was identified by interfacial energy gradient analysis, where the liquid film thickness in the range of 0.5–3 μm was considered an effective overlapping contact area, which could be used for subsequent capillary action modeling.

[0089] In another embodiment, it is assumed that the pore water storage capacity of a sandy loam soil sample is approximately The calculated pore water saturation level was approximately 65%–88% (with a local center value of approximately 76%). The low-saturation zone (<60%) accounted for 22%, the medium-saturation zone (60%–85%) for 51%, and the high-saturation zone (>85%) for 27%. Wetting state classification showed that the fully wetted zone accounted for approximately 38%, the partially wetted zone for approximately 44%, and the non-wetting zone for approximately 18%. Analysis of the particle liquid film extension range showed an average liquid film influence radius of approximately 12 μm, with the high-wetting zone extending to 18 μm, while the low-wetting zone only reached approximately 6 μm. Overlapping regions at the particle contact interface accounted for approximately 41% of the total particle contacts, with the fine particle agglomeration region being the most significant.

[0090] The probability of capillary bridge formation is calculated based on the overlapping area of ​​the contact interface; the number of capillary bridges per unit area is counted based on the probability of capillary bridge formation; the capillary bridge formation density is determined based on the number of capillary bridges per unit area; and the soil capillary adhesion force is determined based on the capillary bridge formation density.

[0091] In one embodiment, assuming a particle contact interface overlap rate of approximately 41% in a certain region, the corresponding capillary bridge formation probability is approximately 0.72–0.89 (average approximately 0.81). Within a unit area of ​​1 cm², the distribution of capillary bridge numbers is as follows: approximately 35–60 bridges / cm² in sandy areas, approximately 70–110 bridges / cm² in loamy areas, and approximately 120–180 bridges / cm² in clayy areas. The corresponding capillary bridge formation densities are approximately 45 bridges / cm² (low-density area), 95 bridges / cm² (medium-density area), and 150 bridges / cm² (high-density area), respectively. The calculated soil capillary adhesion is approximately 0.6–1.1 kPa for sandy structures, approximately 1.3–2.0 kPa for loamy structures, and approximately 2.2–3.1 kPa for clayy structures. Overall, for every approximately 10% increase in capillary bridge density, capillary adhesion increases by approximately 8%–12%, exhibiting a significant enhancing effect in high-clay areas.

[0092] Preferably, assessing soil particle cohesion based on soil capillary adhesion includes:

[0093] Calculate soil particle binding force based on soil capillary adhesion force; determine soil disturbance resistance value based on soil particle binding force; evaluate wind stripping resistance based on soil disturbance resistance value;

[0094] In one embodiment, soil capillary adhesion is modeled as an equivalent force field resulting from the combined effects of pore water meniscus tension and particle contact area. The capillary adhesion force per unit particle is calculated based on soil moisture content and pore structure parameters. And combined with particle contact coordination number With contact area coefficient To obtain the soil particle binding force Introducing the soil internal structure disturbance response coefficient (Reflecting sensitivity to vibration, shear, and external disturbances), defining soil disturbance resistance value. Considering the shear stress under the action of wind field With equivalent windward area Wind stripping resistance is established through mechanical equilibrium relationships. , is used to characterize the detachment threshold of soil particles under wind action.

[0095] In another embodiment, assuming the surface soil moisture content of a grassland is 18%, the capillary adhesion force of a single particle is calculated accordingly. Mean coordination number Contact area coefficient Then soil particle binding force Take the structural disturbance coefficient. Then the disturbance resistance value Set the windward equivalent area. They are respectively The corresponding calculation of wind stripping resistance Approximately Among them, under strong gusts (wind speed of about 12 m / s), when When the value exceeds 0.8N, it can be determined that there is a risk of particle loosening in the area.

[0096] The soil structure stability index is calculated based on wind stripping resistance; the soil particle cohesion is assessed based on the soil structure stability index.

[0097] In one embodiment, wind power is stripped of drag. With soil cohesion and gravity stability terms Normalization was performed to construct a soil structure stability index. The higher the index, the stronger the soil's resistance to wind erosion. Simultaneously, it maps interparticle cementation (including organic matter cementation and clay mineral cementation) into an equivalent cohesion index. And establish mapping relationships ,in It is a structural amplification factor used to reflect the enhancing effect of micro-cementation on macroscopic stability, thereby realizing a quantitative conversion from mechanical stability to particle cohesion.

[0098] In another embodiment, assuming that soil cohesion is measured in a wind-erosion sensitive area... Gravity stability term Combined with the wind stripping resistance in the previous embodiment. value range The corresponding soil structure stability index is... The values ​​are [4.07, 3.24, 2.50]. The structural amplification factor is taken. Soil particle cohesion The values ​​are approximately [3.46, 2.75, 2.13] respectively. When the viscosity drops below 2.5, the soil in the area is considered to be in a medium-low cohesion state, with significantly weakened resistance to wind erosion, making it prone to surface particle stripping and structural damage under continuous wind action.

[0099] It should be noted that you should refer to [link / reference]. Figure 2 This image has a spatial resolution of 10m, uses a B4 / B3 / B2 true-color band combination, and covers a grassland ecological area of ​​approximately 80km. The vegetation in the image is generally dark green with a high NDVI value (up to 0.8), indicating healthy grassland growth. A winding river flows through the central part of the image, surrounded by several dark-colored lakes. Some areas show brownish-yellow bare land or degraded patches. Image metadata is located in the upper left corner, the NDVI legend is in the lower left corner, the scale bar is in the lower right corner, and a north arrow is marked in the upper right corner.

[0100] It should be noted that you should refer to [link / reference]. Figure 3 On the left is the degraded grassland before restoration, with large areas of bare ground, sparse and withered vegetation, and obvious dirt roads. On the right is the ecologically restored grassland after restoration, with dense and lush grass, significantly increased vegetation coverage, and rolling hills. The two sides are separated by white dotted lines and arrows, which visually demonstrate the significant changes before and after ecological restoration.

[0101] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0102] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A multimodal grassland ecological restoration image feature extraction and recognition system, characterized in that, Includes the following modules: The data acquisition module is used to collect multi-source monitoring data from grassland areas; and to perform georeferencing based on the multi-source monitoring data to construct an ecological monitoring dataset. The feature extraction module is used to extract vegetation cover features from the ecological monitoring dataset; Calculate community canopy closure based on vegetation cover characteristics; assess precipitation interception rate based on community canopy closure; and determine the water retention capacity of grassland vegetation based on precipitation interception rate. The grassland degradation identification module is used to determine the soil moisture retention index based on the water retention capacity of grassland vegetation; identify wind erosion-sensitive areas based on the soil moisture retention index; and determine the grassland degradation level in wind erosion-sensitive areas. The effect evaluation module is used to calculate the vegetation cover growth rate based on the ecological monitoring dataset; and to determine the reduction ratio of degraded areas based on the vegetation cover growth rate. The ecological restoration index is calculated based on the reduction rate of degraded areas. The effectiveness of grassland ecological restoration is assessed based on the ecological restoration index.

2. The multimodal grassland ecological restoration image feature extraction and recognition system according to claim 1, characterized in that, The feature extraction module calculates community canopy closure based on vegetation cover features, including: Vegetation continuous distribution areas are marked based on vegetation cover characteristics; vegetation patch connectivity is calculated based on vegetation patch connectivity; and vegetation spatial continuity index is determined based on vegetation patch connectivity. The degree of canopy shading is assessed based on the vegetation spatial continuity index; the shading area is marked according to the degree of canopy shading; the canopy transmittance is calculated in the shading area; the canopy void index is determined based on the canopy transmittance; the canopy closure coefficient is calculated based on the canopy void index; and the community canopy closure is calculated based on the canopy closure coefficient.

3. The multimodal grassland ecological restoration image feature extraction and recognition system according to claim 2, characterized in that, Calculating canopy transmittance in shaded areas includes: Acquire vegetation canopy images corresponding to the shaded areas; identify light-transmitting and non-light-transmitting areas based on the vegetation canopy images; calculate the light-transmitting area based on the light-transmitting areas; calculate the total shaded area of ​​the non-light-transmitting areas; The canopy shading rate is calculated based on the total shading area and the light-transmitting area; the canopy transmittance is determined based on the ratio of the canopy shading rate to the light-transmitting area.

4. The multimodal grassland ecological restoration image feature extraction and recognition system according to claim 3, characterized in that, Based on vegetation canopy images, the areas that are translucent and those that are not translucent include: The vegetation canopy image is converted to grayscale to obtain pixel grayscale values; the pixel brightness value is calculated based on the pixel grayscale value; if the pixel brightness value is greater than the preset light transmittance threshold, it is marked as a light-transmitting area; if the pixel brightness value is less than the preset light transmittance threshold, it is marked as a non-light-transmitting area.

5. The multimodal grassland ecological restoration image feature extraction and recognition system according to claim 1, characterized in that, The feature extraction module assesses precipitation interception rate based on community canopy closure and determines grassland vegetation water retention capacity based on precipitation interception rate, including: Determine the canopy closure level based on community canopy closure; calculate the raindrop blocking ratio based on the canopy closure level; determine the canopy interception capacity based on the raindrop blocking ratio; obtain precipitation data; calculate the precipitation interception rate based on the canopy interception capacity and precipitation. The surface runoff reduction coefficient is calculated based on the precipitation interception rate; the precipitation infiltration is determined based on the surface runoff reduction coefficient; the soil moisture replenishment index is calculated based on the precipitation infiltration; and the water retention capacity of grassland vegetation is assessed based on the soil moisture replenishment index.

6. The multimodal grassland ecological restoration image feature extraction and recognition system according to claim 5, characterized in that, Determining canopy closure level based on community canopy closure includes: Dominant vegetation community areas are identified based on canopy closure; spectral absorption characteristics are extracted from the dominant vegetation community areas; and chlorophyll enrichment is calculated based on the spectral absorption characteristics. The photosynthetic intensity is determined based on the degree of chlorophyll enrichment; the leaf expansion capacity is calculated based on the photosynthetic intensity; the canopy cover increment is determined based on the leaf expansion capacity; and the canopy closure level is determined based on the canopy cover increment.

7. The multimodal grassland ecological restoration image feature extraction and recognition system according to claim 5, characterized in that, The calculation of raindrop blocking ratio based on canopy closure level includes: Determine the degree of leaf overlap based on the canopy closure level; calculate the projected area of ​​leaves per unit area based on the degree of leaf overlap; determine the precipitation interception area based on the projected area of ​​leaves per unit area. Acquire leaf distribution data in the precipitation interception area; calculate leaf coverage per unit area based on leaf distribution data; determine the number of raindrop collisions with the leaves based on leaf coverage per unit area; calculate the raindrop blocking ratio based on the number of raindrop collisions with the leaves.

8. The multimodal grassland ecological restoration image feature extraction and recognition system according to claim 1, characterized in that, The grassland degradation identification module identifies wind erosion-sensitive areas based on soil moisture retention index, including: Calculate the available soil moisture content based on the soil moisture retention index; calculate the soil pore water storage based on the available soil moisture content; calculate the soil capillary adhesion based on the soil pore water storage. Soil particle cohesion is assessed based on soil capillary adhesion; soil wind erosion resistance is determined based on soil particle cohesion; surface sand-raising risk index is calculated based on soil wind erosion resistance; and wind erosion-sensitive areas are identified based on the surface sand-raising risk index.

9. The multimodal grassland ecological restoration image feature extraction and recognition system according to claim 8, characterized in that, The calculation of soil capillary adhesion force based on soil pore water storage capacity includes: Assess pore water saturation based on soil pore water storage capacity; determine wetting state based on pore water saturation; delineate particle liquid film extension range based on wetting state; identify overlapping areas of particle contact interfaces based on liquid film extension range. The probability of capillary bridge formation is calculated based on the overlapping area of ​​the contact interface; the number of capillary bridges per unit area is counted based on the probability of capillary bridge formation; the capillary bridge formation density is determined based on the number of capillary bridges per unit area; and the soil capillary adhesion force is determined based on the capillary bridge formation density.

10. The multimodal grassland ecological restoration image feature extraction and recognition system according to claim 8, characterized in that, Soil particle cohesion is assessed based on soil capillary adhesion, including: Calculate soil particle binding force based on soil capillary adhesion force; determine soil disturbance resistance value based on soil particle binding force; evaluate wind stripping resistance based on soil disturbance resistance value; The soil structure stability index is calculated based on wind stripping resistance; the soil particle cohesion is assessed based on the soil structure stability index.

Images