Method for quantifying regional grazing intensity based on unmanned aerial vehicle observation of livestock manure
By using UAV image data processing and image segmentation algorithms, combined with linear regression and clustering algorithms, the system identifies grazing intensity and activity preference areas, and plans rotational grazing routes. This solves the problem of refining and dynamizing grassland grazing intensity assessment, and achieves a balance between efficient utilization of grassland resources and ecological protection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- LANZHOU UNIV
- Filing Date
- 2026-05-08
- Publication Date
- 2026-07-31
AI Technical Summary
Existing technologies are insufficient for precise and dynamic assessment of grazing intensity, cannot accurately identify the distribution of livestock manure and local differences, and lack a closed-loop management mechanism from intensity assessment to overgrazing warning, rotational grazing route planning and ecological restoration feedback, resulting in local grassland degradation and uneven resource utilization.
High-resolution imagery was acquired using drones, and image segmentation algorithms were used to extract manure distribution features to generate density distribution maps. A linear regression model was used to calculate grazing intensity, and a clustering algorithm was combined to generate heat maps to identify areas of activity preference. Threshold monitoring was employed to trigger early warnings, and rotational grazing routes were planned using terrain data and path optimization algorithms. Grazing-prohibited and restricted grazing areas were demarcated, and virtual fencing technology was used to limit livestock activity. Multispectral sensor data was integrated to update manure freshness indicators to assess the effectiveness of ecological restoration.
It enables refined and dynamic assessment and control of grassland grazing intensity, improves the balance of grassland resource utilization and the controllability of ecological restoration, and alleviates grassland degradation caused by local overgrazing.
Smart Images
Figure CN122491664A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of grazing intensity quantification, and in particular relates to a method for quantifying regional grazing intensity based on UAV observation of livestock manure. Background Technology
[0002] In the intersection of modern animal husbandry and grassland ecological protection, the scientific management of grazing activities is crucial for maintaining the sustainable use of grasslands. Traditional methods for assessing grazing intensity mainly rely on manual ground patrols, livestock headcount, or experience-based judgment. These methods are simple to operate and easy to implement, and are widely used in small-scale ranches or traditional farming models. In recent years, with the development of remote sensing technology and geographic information systems, some studies have begun to explore using satellite imagery to assess vegetation cover changes or to obtain activity trajectories through livestock tracking collars, aiming to obtain more macroscopic information on grazing distribution. These methods provide some data support for grassland management and help identify areas of long-term overgrazing.
[0003] However, existing technologies still have significant shortcomings in practical applications. First, manual patrols and census counting are insufficient to comprehensively capture the refined spatial impact of livestock activities on grasslands, especially in large-scale grasslands with complex terrain. They often overlook differences in grazing intensity in local areas and cannot reflect the dynamic changes between livestock activity and grassland conditions in a timely manner. Second, satellite imagery has limited resolution, making it difficult to identify subtle surface features such as livestock manure. While livestock manure, as the most direct trace of livestock activity, contains crucial information such as activity range, density, and time, its spatial distribution is highly uneven and easily affected by environmental factors such as terrain and vegetation. This makes it impossible for existing methods to accurately extract grazing intensity indicators from manure distribution. Furthermore, even if high-density manure areas are identified, existing technologies lack a complete closed-loop management mechanism that extends from intensity assessment to overgrazing warnings, rotational grazing route planning, and ecological restoration feedback. This makes it difficult for managers to take timely and targeted measures after overgrazing is detected, leading to increased local grassland degradation and uneven resource utilization. Summary of the Invention
[0004] To address the aforementioned technical problems, this invention provides a method for quantifying regional grazing intensity based on unmanned aerial vehicle (UAV) observation of livestock manure, comprising: Acquire drone image data, extract fecal distribution features based on the drone image data, and obtain a fecal density distribution map; Calculate the grazing intensity value based on the fecal density distribution map to determine the current grazing intensity level; Identify activity preference areas based on the current grazing intensity level; The overload status is determined based on the activity preference area, and an overload area list is obtained; The warning activation zone is determined based on the overload zone list; Based on the planned rotational grazing routes in the early warning activation area, an ordered rotation sequence is obtained; The protection boundary is determined based on the ordered rotation sequence; An ecological restoration assessment map is generated based on the protection boundary; Adjust rotational grazing routes based on the aforementioned ecological restoration assessment map to determine the optimized grazing utilization rate.
[0005] Preferably, the process of obtaining the fecal density distribution map includes: Image data is acquired using drone equipment to obtain initial image files; Based on the initial image file, the fecal distribution area is identified, and a segmented fecal area image is obtained; Based on the segmented fecal region image, the density of fecal distribution is calculated to determine fecal density data; A density distribution matrix is constructed based on the fecal density data, and a density distribution map is generated. The density distribution map is filtered to obtain an optimized density distribution map; Based on the optimized density distribution map, spatial mapping processing is performed to output the final fecal density distribution map.
[0006] Preferably, the process of determining the current grazing intensity level includes: Density information is extracted from the fecal density distribution map to obtain a density distribution dataset; Based on the density distribution dataset, the density value per unit area is fitted to determine the grazing intensity value; Based on the grazing intensity values, high-intensity areas are marked, and the distribution information of high-intensity areas is obtained; Based on the distribution information of the high-intensity areas, local clustering phenomena are identified, and identification data of the local clustering areas are obtained; Based on the identification data of the local clustered areas, determine the spatial distribution characteristics; Based on the spatial distribution characteristics, determine the equilibrium state of grazing intensity levels; Based on the equilibrium state of the grazing intensity level, a regionalized intensity distribution map is generated to determine the final grazing intensity assessment result.
[0007] Preferably, the process of identifying the active preference region includes: Spatial distribution information was obtained from grazing intensity records to obtain a processed spatial distribution dataset; Based on the organized spatial distribution dataset, the data points are grouped to determine the preliminary intensity distribution categories; A spatial distribution matrix is generated based on the preliminary intensity distribution categories, and the spatial distribution matrix is converted into a heat map; Based on the heat map, the location information of high-intensity clustering areas is extracted to determine the range of activity preference areas; The region boundaries are divided based on the range of the activity preference region to determine the preference region boundaries; The average intensity value is calculated based on the boundary of the preferred region to obtain the intensity distribution result of the active preferred region.
[0008] Preferably, the process of obtaining the list of overloaded areas includes: Acquire grazing intensity data for the activity preference areas and determine the intensity monitoring results for each area; Based on the intensity monitoring results, the overload condition is determined, and the overload condition determination result is obtained; Based on the overload status determination results, the overload region is extracted to form an overload region set; Spatial clustering is performed on the set of overloaded regions to obtain overloaded region grouping information; Based on the overloaded area grouping information, priority sorting is performed to obtain a sorted list of overloaded areas; Based on the sorted list of overloaded areas, a preliminary plan for intervention strategies is generated, and the basis for allocating intervention strategies is determined. Based on the allocation criteria of the intervention strategies, the intervention strategies are matched to obtain a list of regional intervention plans.
[0009] Preferably, the process of determining the early warning activation zone includes: Real-time monitoring data is obtained from the overloaded area to obtain a set of density anomaly points; Based on the set of density anomaly points, filter out anomaly points that exceed the threshold to determine a list of high-risk density points; Based on the list of high-risk density locations, determine the degree of clustering between locations and mark potential warning areas; The density change trend is extracted from the potential warning area to obtain the density anomaly persistence assessment result; Based on the results of the density anomaly persistence assessment, an early warning signal is triggered to determine the activation range that requires close attention. Based on the activation range, generate the warning activation boundary and output the range division result.
[0010] Preferably, the process of obtaining the ordered rotation sequence includes: Based on the early warning activation signal, obtain the boundary data of the area division and determine the distribution of monitoring points; Based on the distribution of the monitoring points, terrain data is extracted to obtain information on terrain elevation and obstacle distribution; Based on the terrain elevation and obstacle distribution information, determine the suitable area for passage; Based on the suitable traversable area, a path optimization algorithm is used to calculate and obtain the rotational grazing route; An ordered sequence is generated based on the direction of the rotational grazing path, and the order of each path node is determined. Based on the ordered sequence, the logical arrangement of the rotation order is processed to generate rotation sequence data; The path nodes of the circumduction sequence data are verified to obtain a complete grazing path plan.
[0011] Preferably, the process of determining the protection boundary includes: Based on ordered and cyclic sequence data, initial regional division information is obtained, and preliminary distribution of grazing-prohibited and grazing-restricted areas is obtained. Based on the preliminary distribution of grazing-prohibited and grazing-restricted areas, the protection boundaries are adjusted to determine the final protection boundary range. Based on the final protection boundary range, the constraint range of the virtual fence is optimized to obtain the restriction data of the livestock activity range; Based on the restricted data of the livestock's activity range, adjust the deployment points of the virtual fence and determine the new activity range constraints; Based on the new activity range constraints, analyze the livestock activity trajectories to determine whether they comply with the activity range restrictions. Based on the analysis of livestock activity trajectories, new constraint strategies are determined through a dynamic adjustment mechanism. Based on the new constraint strategy, the grazing ban area and grazing restriction area are dynamically adjusted to obtain the latest regional distribution data.
[0012] Preferably, the process of generating an ecological restoration assessment map includes: Data was collected from the protected area using a multispectral sensor to obtain an initial dataset; Preprocessing is performed on the initial dataset to obtain a cleaned image set; Based on the image set after cleaning, spectral features related to fecal freshness are extracted to determine the distribution of freshness indicators. Based on the distribution of the freshness index, the effectiveness of the restrictive measures was analyzed, and preliminary results on the impact of the measures were obtained. Based on the preliminary results of the impact of the measures, the progress of ecological restoration is determined, and a restoration trend layer is generated; A comprehensive ecological restoration distribution map is generated by overlaying historical data onto the restoration trend layer.
[0013] Preferably, the process of determining the optimized grazing utilization rate includes: Data on the distribution characteristics of unused areas are obtained from ecological restoration assessment maps to determine their location and extent. Based on the location and extent information of the unused area, the unused area is divided into blocks, and the boundaries and area data of each block are determined. Based on the area data of each block, the grazing path is adjusted using a distribution balancing algorithm to obtain the adjusted path distribution scheme. Based on the adjusted path distribution scheme, the grazing utilization intensity of each block is calculated, and the path is locally optimized to obtain the optimized path plan. Based on the optimized path planning, the utilization rate of each block is calculated to determine the overall grazing utilization rate distribution of the region. Based on the distribution of grazing utilization rates in the overall region, the final utilization rate value and path adjustment results are generated.
[0014] Compared with the prior art, the present invention has the following advantages and technical effects: This invention acquires high-resolution image data from drones and combines it with image segmentation algorithms to accurately extract fecal distribution features and generate density distribution maps, effectively solving the identification problems caused by the uneven spatial distribution of feces and environmental interference. On this basis, it uses a linear regression model to calculate the grazing intensity value per unit area and combines it with a clustering algorithm to generate a heat map, which can accurately identify activity preference areas and overgrazing states, overcoming the shortcomings of traditional methods that ignore local differences and lack dynamic feedback.
[0015] This invention plans rotational grazing routes based on overgrazing warning areas and terrain data, divides prohibited and restricted grazing areas using virtual fencing technology, and integrates multispectral sensor data to update manure freshness indicators to assess the ecological restoration effect. Finally, it uses a distributed equilibrium algorithm to optimize grazing utilization, forming a complete closed-loop management system from "identification-assessment-early warning-intervention-feedback".
[0016] This invention enables refined and dynamic assessment and control of grassland grazing intensity, significantly improving the balance of grassland resource utilization and the controllability of ecological restoration, and effectively alleviating grassland degradation caused by local overgrazing. Attached Figure Description
[0017] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention. Detailed Implementation
[0018] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.
[0019] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.
[0020] like Figure 1 As shown, this embodiment provides a method for quantifying regional grazing intensity based on unmanned aerial vehicle (UAV) observation of livestock manure, including: Acquire drone image data, extract fecal distribution features based on the drone image data, and obtain a fecal density distribution map; The grazing intensity value is calculated based on the manure density distribution map to determine the current grazing intensity level; Identify activity preference areas based on current grazing intensity levels; The overload status is determined based on the activity preference area, and a list of overloaded areas is obtained; Determine the warning activation zone based on the list of overloaded areas; Based on the rotational grazing routes planned in the early warning activation zone, an orderly rotation sequence was obtained; Determine the protection boundary based on the ordered rotation sequence; An ecological restoration assessment map is generated based on the protection boundaries; Adjust rotational grazing routes based on ecological restoration assessment maps to determine optimized grazing utilization rates.
[0021] This embodiment acquires high-resolution imagery using drones, extracts manure distribution features using image segmentation algorithms, generates a density distribution map, calculates grazing intensity using a linear regression model, and then generates a heat map using a clustering algorithm to identify areas of favored activity. For overgrazing areas, this embodiment uses threshold monitoring to trigger early warnings, combines terrain data and path optimization algorithms to plan rotational grazing routes, delineates prohibited and restricted grazing areas, and restricts livestock activity range using virtual fencing technology. Simultaneously, it integrates multispectral sensor data to update manure freshness indicators, assesses the effectiveness of ecological restoration, optimizes rotational grazing routes for unused areas, and ultimately improves grazing utilization. This embodiment achieves a balance between efficient use of grassland resources and ecological protection, demonstrating significant practical value.
[0022] Furthermore, the process of obtaining the fecal density distribution map includes: Image data is acquired using drone equipment to obtain initial image files; Based on the initial image file, the fecal distribution area is identified, and the segmented fecal area image is obtained. Based on the segmented fecal region image, the density of fecal distribution is calculated to determine fecal density data; A density distribution matrix is constructed based on fecal density data, and a density distribution map is generated. The density distribution map is filtered to obtain an optimized density distribution map. Spatial mapping is performed based on the optimized density distribution map to output the final fecal density distribution map.
[0023] Furthermore, this embodiment uses a drone to take aerial photographs of the target area, acquiring high-resolution image data and storing it as an initial image file. Based on the initial image file, an image segmentation algorithm is used to process the image, identifying the fecal distribution areas and obtaining segmented fecal region images. Pixel-level statistical analysis is performed on the segmented fecal region images to calculate the density of fecal distribution in each region and determine the fecal density data. A density distribution matrix is constructed using the fecal density data to generate a preliminary density distribution map. If noise points or abnormal areas exist in the density distribution map, a preset threshold is used for filtering to remove areas that do not meet the criteria, resulting in an optimized density distribution map. The optimized density distribution map is then combined with geographic information system data for spatial mapping processing to determine the consistency between the map and the actual terrain, outputting the final fecal density distribution map. Finally, a visualization tool is used to render the final fecal density distribution map, generating an intuitive color-coded distribution map and determining the final presentation result.
[0024] In one embodiment, an unmanned aerial vehicle (UAV) cruises at a constant speed at an altitude of 50 meters to acquire image data with a ground resolution of 0.02 meters, and stores it in a tagged image format.
[0025] Specifically, this embodiment uses a semantic segmentation algorithm based on deep convolutional neural networks to perform pixel-level classification of background and target in the image.
[0026] For example, when the algorithm identifies a cluster of pixels with a specific texture and tone, it marks it as a fecal region and outputs a binarized image.
[0027] In one embodiment, this embodiment performs sliding window statistics on the binarized image, with the window size set to 50 by 50 pixels.
[0028] Specifically, the percentage of target pixels within the window is calculated, and if the percentage exceeds 0.1, it is determined to be a high-density area.
[0029] For example, if three fecal points are detected within a 1-square-meter actual ground area, the density value of that area is recorded as 3.
[0030] In one embodiment, a 100x100 density distribution matrix is constructed, and the density values of each window are filled into the corresponding coordinates.
[0031] Specifically, if the density value of an isolated pixel in the matrix is much higher than the surrounding average, for example, more than 5 times the average, it is determined to be sensor noise.
[0032] For example, by setting an area threshold of 10 pixels, excessively small fragmented areas can be eliminated to ensure the accuracy of the map.
[0033] In one embodiment, the optimized map is projected onto a global geodetic coordinate system.
[0034] Specifically, by combining digital elevation model data, we can check whether the distribution of feces conforms to the terrain logic, such as excluding steep areas with a slope greater than 45 degrees.
[0035] In one embodiment, the spatial mapping process aligns the image coordinate system with the vector layer in the geographic information system using an affine transformation model.
[0036] Specifically, resampling technology is used to unify the spatial resolution to 0.5 meters, ensuring that the density data and terrain features completely overlap in spatial location.
[0037] For example, by comparing the attributes of ground features, if a high-density point is found to be located on a hardened road surface, its confidence weight is reduced to 0.3 to eliminate non-target interference, and finally a validated fecal density distribution map is output.
[0038] In one embodiment, the visualization tool uses a heatmap rendering algorithm to process the distribution map.
[0039] Specifically, energy diffusion is calculated by applying a Gaussian kernel function with a radius of 5 pixels to each fecal pixel, thereby generating a smooth density field.
[0040] For example, the calculated energy values can be mapped to a gradient from cool to warm colors, where blue represents areas with a density value of 0 to 1 and red represents areas with a density value greater than 10, thus visually demonstrating the spatial aggregation trend of pollutants.
[0041] In one embodiment, the final results are output in the form of a digital report, which includes area statistics for each density level.
[0042] Specifically, by counting pixels in the color-coded map, the distribution ratio of different pollution levels can be obtained.
[0043] For example, if statistics show that the area of high-density areas reaches 500 square meters, a cleanup task list containing specific latitude and longitude coordinates can be generated to determine the final presentation result.
[0044] Furthermore, the process of determining the current grazing intensity level includes: Density information is extracted from the fecal density distribution map to obtain a density distribution dataset; Based on the density distribution dataset, the density value per unit area is fitted to determine the grazing intensity value; Based on grazing intensity values, high-intensity areas are marked to obtain distribution information of these high-intensity areas. Based on the distribution information of high-intensity areas, local clustering phenomena are identified, and the identification data of local clustering areas are obtained; Based on the identification data of local clustered areas, determine the spatial distribution characteristics; Based on spatial distribution characteristics, determine the equilibrium state of grazing intensity levels; Based on the equilibrium state of grazing intensity levels, a regionalized intensity distribution map is generated to determine the final grazing intensity assessment result.
[0045] Furthermore, this embodiment obtains raw data through a fecal density distribution map, and uses automated tools to extract density information from the distribution map to obtain a preliminary density distribution dataset. Based on the density distribution dataset, a linear regression model is constructed to fit the density values per unit area, determining the grazing intensity value corresponding to that unit area. Using the output of the linear regression model, combined with a preset intensity level threshold, if the calculated grazing intensity value exceeds the threshold range, it is marked as a high-intensity area, and the distribution information of high-intensity areas is obtained. Through the distribution information of high-intensity areas, the correlation between fecal density and grazing intensity is analyzed to determine whether local clustering exists, obtaining identification data for locally clustered areas. Based on the identification data of locally clustered areas, relevant location information is extracted from the distribution map to determine the spatial relationship between clustered areas and the overall distribution map, obtaining spatial distribution characteristics. After obtaining the spatial distribution characteristics, combined with the grazing intensity value per unit area, the differences in intensity levels across different regions are analyzed to determine the equilibrium state of the overall grazing intensity level. Through the equilibrium state of the overall grazing intensity level, a regionalized intensity distribution mapping is generated to determine the final grazing intensity assessment result.
[0046] For example, this embodiment uses an automated analysis tool to perform pixel-level scanning of the fecal density distribution map, extracts the density feature values within each grid cell, and forms a preliminary dataset containing coordinate information and density values.
[0047] In one possible implementation, this embodiment uses the extracted density dataset as an independent variable to input a linear regression model, and fits the model using a pre-established ratio between manure retention and livestock feed intake.
[0048] For example, when 0.8 fecal targets are detected per unit area, the model outputs a grazing intensity of 1.5 standard livestock units per hectare. This quantification transforms visual distribution characteristics into specific production and management indicators.
[0049] For example, a grazing intensity warning threshold can be set to 2.0. When the value output by the linear regression model reaches 2.5, the area is automatically marked as a high-intensity grazing area.
[0050] In one possible implementation, this embodiment uses a spatial autocorrelation algorithm to analyze the clustering of feces distribution in these high-intensity areas. By calculating the local Moran's index, if the index is greater than 0.6 and the significance level is less than 0.05, it is determined that there is local clustering at that location, and corresponding clustering area identification data is generated. This helps to identify the preferred resting points of livestock in pastures, such as water sources or shaded areas.
[0051] For example, the geometric center and coverage radius of the clustered area are extracted based on the acquired identification data to determine its relative position in the overall grassland space.
[0052] For example, the analysis found that 70% of the clusters were located in gentle slopes of less than 5 degrees.
[0053] In one possible implementation, the intensity values per unit area are combined to compare the intensity differences between different plots. If the standard deviation of intensity between regions is greater than 1.2, the overall grazing level is determined to be unbalanced. Finally, spatial interpolation is used to transform the discrete intensity values into a continuous regionalized mapping map, outputting a grazing intensity assessment result that reflects grassland utilization efficiency.
[0054] Furthermore, the process of identifying regions of activity preference includes: Spatial distribution information was obtained from grazing intensity records to obtain a processed spatial distribution dataset; Based on the organized spatial distribution dataset, the data points are grouped to determine the preliminary intensity distribution categories; A spatial distribution matrix is generated based on the preliminary intensity distribution categories, and the spatial distribution matrix is converted into a heat map; The location information of high-intensity clustering areas is extracted from the heat map to determine the range of activity preference areas; The boundaries of the preferred activity areas are defined by dividing the area into zones. The average intensity value is calculated based on the boundary of the preferred region to obtain the intensity distribution result of the activity preferred region.
[0055] Furthermore, this embodiment obtains spatial distribution information from grazing intensity records, cleans and formats the original dataset, removes invalid values and outliers, and obtains a cleaned spatial distribution dataset. For the cleaned spatial distribution dataset, a clustering algorithm is applied to group the data points according to geographical location and intensity value to determine preliminary intensity distribution categories. Based on the preliminary intensity distribution categories, a corresponding spatial distribution matrix is generated, and a visualization tool is used to convert the matrix data into a heatmap, obtaining an intuitive visualization of the intensity distribution. By analyzing the intensity distribution visualization results of the heatmap, the location information of high-intensity clusters is extracted to determine the potential activity preference area range. For the potential activity preference area range, combined with the category information in the spatial distribution matrix, regional boundary delineation is performed to determine the specific preference area boundaries. Based on the determined preference area boundaries, the intensity values from the original spatial distribution dataset are superimposed, and the average intensity value of each area is calculated to obtain the final activity preference area intensity distribution result.
[0056] For example, when obtaining spatial distribution information from grazing intensity records, the original dataset is first cleaned and formatted to remove invalid values and outliers.
[0057] Specifically, this cleaning process involves checking data integrity, such as removing outliers with fecal density values below 0 or above a reasonable upper limit, such as 50 feces / m², to obtain a cleaned spatial distribution dataset. This ensures the accuracy of subsequent analyses, avoids biases caused by noise interference, and improves the reliability of grazing intensity assessments.
[0058] For example, for the organized spatial distribution dataset, clustering algorithms are applied to group the data points, classify them according to geographical location and intensity value, and determine the preliminary intensity distribution categories.
[0059] In one embodiment, the K-means algorithm is used to divide data points into 3-5 clusters, for example, using latitude and longitude coordinates and a density value such as an average of 20 clusters per square meter as cluster centers. This grouping helps to identify the distribution of grazing activity patterns, thereby optimizing pasture management strategies.
[0060] For example, based on the initial intensity distribution categories, a corresponding spatial distribution matrix is generated, and visualization tools are used to convert the matrix data into a heat map format, resulting in an intuitive visualization of the intensity distribution.
[0061] Specifically, the matrix can be represented as a grid, with each cell corresponding to the intensity value of a 1 square kilometer area. For example, red indicates a high-density area with more than 30 cells per square meter. This heat map can intuitively reveal the location of high-intensity areas, facilitating rapid decision-making for grazing adjustments and reducing the risk of overgrazing.
[0062] For example, by analyzing the intensity distribution visualization results of heat maps, the location information of high-intensity clustering areas can be extracted to determine the potential range of activity preference areas.
[0063] In one embodiment, the analysis process includes setting a threshold, such as marking areas with a density exceeding 25 animals per square meter as high intensity, thereby identifying grassland areas where livestock tend to congregate. This helps in understanding animal behavioral preferences and facilitates sustainable grazing planning.
[0064] For example, based on the potential activity preference area range, the category information in the spatial distribution matrix is combined to perform area boundary division processing to determine the specific preference area boundary.
[0065] Specifically, the convex hull algorithm is used to define boundaries, for example, enclosing a point set with a density of 15-40 points per square meter to form a closed region. This division can accurately quantify the size of the preferred area, such as an area of 10 hectares, improving the efficiency of area management.
[0066] For example, based on the determined preference region boundaries, the intensity values in the original spatial distribution dataset are superimposed to calculate the average intensity value in each region, thus obtaining the final result of the activity preference region intensity distribution.
[0067] In one embodiment, the average density of all points within a region is calculated. For example, an average of 28 points per square meter indicates a high-intensity preference. This not only quantifies differences in grazing intensity but also supports balanced resource allocation and prevents localized degradation.
[0068] Furthermore, the process of obtaining the list of overloaded areas includes: Acquire grazing intensity data for the activity preference areas and determine the intensity monitoring results for each area; Based on the strength monitoring results, the overload condition is determined, and the overload condition judgment result is obtained; Based on the overload status determination results, the overloaded areas are extracted to form an overloaded area set; Spatial clustering is performed on the overloaded region set to obtain overloaded region grouping information; Based on the overloaded area grouping information, prioritize and sort the overloaded areas to obtain a sorted list of overloaded areas. Based on the sorted list of overloaded areas, a preliminary plan for intervention strategies is generated, and the basis for allocating intervention strategies is determined. Based on the allocation criteria of intervention strategies, match intervention strategies to obtain a list of regional intervention plans.
[0069] Furthermore, this embodiment acquires grazing intensity data for activity preference areas, and uses a monitoring system to collect grazing intensity data in real time for each area, determining the intensity monitoring results for each area. Based on the intensity monitoring results, if the grazing intensity value of a certain area exceeds a preset carrying capacity threshold, the area is determined to be in an overloaded state using a threshold comparison method, resulting in an overloaded state determination result. Based on the overloaded state determination result, areas meeting the overloaded conditions are extracted to form a preliminary set of overloaded areas, determining the distribution of overloaded areas. For the distribution of overloaded areas, regional analysis technology is used to perform spatial clustering processing on the overloaded areas, obtaining clustered overloaded area grouping information. Using the clustered overloaded area grouping information, combined with regional division data, the overloaded areas within each group are prioritized, resulting in a ranked list of overloaded areas. Based on the ranked list of overloaded areas, the activity preference data and intensity monitoring data are correlated to generate a preliminary plan for targeted intervention strategies, determining the basis for allocating intervention strategies. Based on the basis for allocating intervention strategies, an information processing system automatically matches intervention strategies for each overloaded area, obtaining a final list of regional intervention plans.
[0070] Real-time collection of grazing intensity data is fundamental to assessing regional carrying capacity. In this embodiment, sensors deployed at various monitoring points in the grassland acquire data on the number of livestock per unit area and grazing time, which is then input into a data processing terminal. When the monitored intensity value exceeds a preset carrying capacity threshold of 15 livestock per hectare, an overgrazing status determination is triggered. This determination process employs a threshold comparison algorithm, comparing real-time data with the preset carrying capacity limit one by one, and outputting the geographical coordinates of the overgrazing area and the overgrazing degree value.
[0071] In one possible implementation, this embodiment employs a density-based spatial clustering algorithm to group geographically adjacent areas with similar levels of overgrazing into the same group, based on the distribution of overgrazing areas. The algorithm takes the coordinates of the region's center point as input, calculates the Euclidean distance between points, merges overgrazing points less than 500 meters apart, and outputs the clustered overgrazing area grouping information. This approach effectively identifies contiguous overgrazing areas, providing spatial reference for subsequent resource allocation.
[0072] Specifically, based on the clustered grouping information, and combining data on vegetation restoration capacity and historical grazing intensity within the region, each group was prioritized. A weighted scoring method was used for ranking, assigning a weight of 0.6 to the degree of vegetation degradation and a weight of 0.4 to the duration of overgrazing, to calculate the intervention priority score for each region. A higher score indicates greater ecological pressure in the region, requiring priority for intervention.
[0073] For example, when generating intervention strategies, a sorted list of regions is used as input, linked to activity preference data, and processed automatically by a rule matching engine. If a region belongs to a high-frequency activity preference area and is severely overgrazing, a grazing ban and rotation strategy is matched; if it belongs to a low-frequency activity preference area, a supplementary feeding guidance strategy is matched. This process outputs a final list of regional intervention plans, clearly defining the type of intervention measures and implementation time for each region. Through this logically rigorous correlation analysis, ecological vulnerabilities can be accurately identified, a dynamic balance of grazing intensity can be achieved, and the sustainable use of grassland resources can be ensured.
[0074] Furthermore, the process of determining the early warning activation zone includes: Real-time monitoring data is obtained from the overloaded area to obtain a set of density anomaly points; Based on the set of density anomaly points, filter out anomaly points that exceed the threshold to determine a list of high-risk density points; Based on the list of high-risk density locations, determine the degree of clustering between locations and mark potential warning areas; Density change trends are extracted from potential early warning areas to obtain the results of density anomaly persistence assessment; Based on the results of the density anomaly persistence assessment, an early warning signal is triggered to determine the activation range that requires key attention; Based on the activation range, generate the warning activation boundary and output the range division result.
[0075] Furthermore, this embodiment obtains real-time monitoring data from overloaded areas, analyzes density distribution, and obtains a preliminary set of density anomaly points. Based on this preliminary set, a preset threshold comparison method is used to filter out anomaly points exceeding the threshold, determining a list of high-risk density points. For the high-risk density point list, the spatial coordinate information corresponding to the points is obtained, and the degree of clustering between points is determined. If the degree of clustering is higher than a preset standard, it is marked as a potential warning area. The density change trend of core points is extracted from the potential warning area, and the density anomaly persistence assessment result is obtained through time series analysis. Based on the density anomaly persistence assessment result, if the persistence exceeds a preset duration, a warning signal is triggered, and the activation range requiring key attention is determined. Detailed regional data within the activation range is obtained, and combined with spatial segmentation technology, a precise warning activation boundary is generated, outputting the final range segmentation result. Based on the range segmentation result, the focus of the density monitoring system is automatically updated, and the data acquisition frequency is adjusted to obtain more accurate subsequent monitoring data.
[0076] In one embodiment, this embodiment obtains real-time coordinate data transmitted back by positioning terminals worn by livestock in the grazing area, uses a kernel density estimation algorithm to perform spatial smoothing on the number of individuals per unit area, identifies points with density values significantly higher than the surrounding average, and forms a preliminary set of density anomaly points.
[0077] For example, when the instantaneous grazing density in a certain pasture grid reaches 60 standard sheep units per hectare, that location is marked as the object to be analyzed.
[0078] Specifically, for the initially screened abnormal locations, their density values are compared with preset ecological carrying capacity thresholds. If the measured density of a location exceeds the threshold by more than 20%, it is included in the list of high-risk density locations.
[0079] For example, this embodiment uses a spatial proximity analysis algorithm to calculate the Euclidean distance between high-risk locations. When the number of locations clustered within 50 meters exceeds 8, the area is determined to have a risk of excessive trampling and is marked as a potential warning area.
[0080] In one embodiment, this embodiment extracts density change data of core locations within the potential warning area over the past 72 hours and uses time series analysis to assess the persistence of the abnormal state. If the cumulative duration of the density exceeding the standard exceeds 60% within the observation period, a warning signal is triggered.
[0081] For example, in this embodiment, the outermost boundary of these continuously exceeding points is extracted using the convex hull algorithm, and combined with the spatial partitioning technology of the geographic information system, an accurate early warning activation range is generated.
[0082] Specifically, based on the generated warning boundary, an automatic instruction is sent to the monitoring terminal to increase the frequency of location data collection in that area from once per hour to once every 10 minutes. Through this dynamic adjustment mechanism, higher resolution trajectory data can be obtained, thereby more accurately depicting the movement path of livestock within the warning range.
[0083] For example, when the system identifies an area as being under warning, it can calculate the average dwell time and movement speed of livestock in that area using adjusted high-frequency data, thereby determining whether there is a risk of vegetation degradation due to fixed-point grazing. This monitoring strategy based on warning boundaries can effectively avoid wasting resources in non-abnormal areas while ensuring real-time control of high-risk plots, achieving refined management of grazing intensity.
[0084] Furthermore, the process of obtaining the ordered rotation sequence includes: Based on the early warning activation signal, obtain the boundary data of the area division and determine the distribution of monitoring points; Topographic data is extracted based on the distribution of monitoring points to obtain information on terrain elevation and obstacle distribution; Based on the terrain elevation and obstacle distribution information, determine the suitable area for passage; Based on the suitable area for passage, a path optimization algorithm is used to calculate and obtain the rotational grazing route; Generate an ordered sequence based on the direction of the rotational grazing path, and determine the order of each path node; Based on the ordered sequence, process the logical arrangement of the rotation order to generate rotation sequence data; By validating the path nodes of the rotation sequence data, a complete rotational grazing path plan is obtained.
[0085] Furthermore, in one possible implementation of this embodiment, the vector boundary data of the target grassland is retrieved based on the early warning activation signal. Latitude and longitude coordinates, such as 112.55 and 42.30, are used as vertices to enclose a closed monitoring area. Within this area, gridded monitoring points are deployed at 50-meter intervals to ensure coverage density meets the needs of refined management.
[0086] Specifically, by extracting elevation data and land feature classification information from monitoring points, steep slope areas with a slope greater than 15 degrees and pitted areas with a depth of more than 0.5 meters are identified and marked as impassable obstacles.
[0087] For example, the constraints of terrain on path planning are analyzed, and the terrain undulation and vegetation coverage are used as weighting factors to construct a passage cost matrix. If there are multiple paths leading to the target grassland in the area, a heuristic search algorithm is used to perform multi-objective optimization calculations, avoiding high-risk obstacles while selecting the path with the shortest total length and gentle slope change.
[0088] For example, when comparing route A and route B, if route A is 1.2 kilometers long but contains many steep slopes, while route B is 1.5 kilometers long but has flat terrain, then route B is preferred as the rotational grazing route.
[0089] In one possible implementation, the coordinates of key nodes are extracted based on the determined rotational grazing route to generate an ordered sequence containing the starting point, intermediate points, and ending point. For the logical arrangement of the rotation order, a direction vector analysis method is introduced to ensure that the livestock herd does not collide during movement, generating rotation sequence data that conforms to a clockwise or counterclockwise pattern.
[0090] Specifically, the generated sequence undergoes integrity verification, checking for duplicate coordinates or path breaks through topological relationships. If abnormal node spacing exceeding 200 meters or missing regions are found, a rearrangement mechanism is triggered, using interpolation algorithms to fill in missing nodes, ultimately outputting a closed and continuous grazing path planning scheme.
[0091] Furthermore, the process of determining the protection boundary includes: Based on ordered and cyclic sequence data, initial regional division information is obtained, and preliminary distribution of grazing-prohibited and grazing-restricted areas is obtained. Based on the preliminary distribution of grazing-prohibited and grazing-restricted areas, the protection boundaries were adjusted to determine the final protection boundary range. Based on the final protection boundary range, the constraint range of the virtual fence is optimized to obtain the restriction data of the livestock activity range; Based on the data on the restrictions on the activity range of livestock, adjust the deployment points of the virtual fence and determine the new activity range constraints; Based on the new activity range constraints, analyze the livestock activity trajectories to determine whether they comply with the activity range restrictions. Based on the analysis of livestock activity trajectories, new constraint strategies are determined through a dynamic adjustment mechanism. Based on the new constraint strategy, the grazing ban and grazing restriction areas are dynamically adjusted to obtain the latest regional distribution data.
[0092] Furthermore, this embodiment obtains initial regional division information through ordered sequences and rotational data, processes the data using a pre-established classification model, and obtains a preliminary distribution of grazing-prohibited and grazing-restricted areas. For the preliminary distribution of grazing-prohibited and restricted areas, corresponding protection boundary information is obtained. If the protection boundary overlaps with or deviates from the regional division, the boundary line is adjusted through a boundary determination module to determine the final protection boundary range. Based on the final protection boundary range, deployment parameters for the virtual fence are obtained, and the constraint range of the virtual fence is optimized using a support vector machine algorithm to obtain the restriction data for livestock activity range. For the restriction data of livestock activity range, specific technical constraints are obtained. If the restriction data exceeds a preset threshold, new activity range constraints are determined by adjusting the deployment points of the virtual fence. Based on the new activity range constraints, real-time monitoring data from the livestock management module is obtained, and the livestock activity trajectory is analyzed using a data comparison method to determine whether it conforms to the preset activity range restrictions. Through the analysis results of the livestock activity trajectory, abnormal activity information is obtained. If the abnormal activity information indicates that livestock exceed the restriction range, a new constraint strategy is determined through the dynamic adjustment mechanism of the virtual fence. Based on the new constraint strategy, the update requirements for regional division are obtained, and an information feedback mechanism is used to dynamically adjust the grazing-prohibited and grazing-restricted areas to obtain the latest regional distribution data.
[0093] For example, the coordinate points in the rotation sequence are combined with historical vegetation cover data and input into a pre-established classification model.
[0094] For example, when the vegetation index of a specific area is below 0.35, the model identifies it as an ecologically fragile area, thus obtaining a preliminary distribution of grazing-prohibited and grazing-restricted areas.
[0095] In one possible implementation, the initially delineated area is compared with the ecological red line in the geographic information system. If a deviation of more than 3 meters is found between the protection boundary and the area delineation, the coordinates are corrected through the boundary determination module to ensure that the final protection boundary avoids sensitive areas such as water sources.
[0096] For example, the latitude and longitude set of the final protected boundary is obtained as input parameters, and the constraint range of the virtual fence is optimized using a support vector machine algorithm. A nonlinear kernel function is used to map the geographical boundary to a high-dimensional feature space, constructing an optimal classification hyperplane to determine the limiting data for livestock activities.
[0097] For example, the effective sensing distance of a virtual fence can be set to 5 meters to ensure that livestock receive timely feedback when they approach the boundary.
[0098] In one possible implementation, real-time location data collected by the livestock management module is acquired, with a sampling frequency set to once every 10 minutes. A data comparison method is used to match the real-time trajectory coordinates of the livestock with the constraint range generated by a support vector machine to determine whether their activities are compliant.
[0099] For example, if the analysis results show that livestock exceed the restricted area five times consecutively within one hour, it is determined to be abnormal activity. At this time, through the dynamic adjustment mechanism of the virtual fence, the pulse intensity or sound warning frequency of the fence is increased by 20%, and the boundary range on that side is automatically reduced by about 2 meters according to the escape direction of the livestock, thus determining a new constraint strategy.
[0100] In one possible implementation, specific technical constraints are obtained, such as stipulating that the maximum distance between fence deployment points must not exceed 50 meters. If the constraint data exceeds a preset threshold, the deployment points of the virtual fence are adjusted, and three control points are added at corners with complex terrain to determine new activity range constraints.
[0101] For example, the update requirements for regional delineation are obtained according to the new constraint strategy. If the vegetation recovery rate of a certain region is lower than expected within three consecutive observation periods, an information feedback mechanism is used to adjust the region from a restricted grazing area to a prohibited grazing area.
[0102] For example, the area of the grazing-prohibited zone was increased from the original 15% to 20%, and the latest regional distribution data was obtained through dynamic adjustment.
[0103] Furthermore, the process of generating an ecological restoration assessment map includes: Data was collected from the protected area using a multispectral sensor to obtain an initial dataset; Preprocessing is performed on the initial dataset to obtain a cleaned image set; Based on the cleaned image set, spectral features related to fecal freshness were extracted to determine the distribution of freshness indicators. Based on the distribution of freshness indicators, analyze the effectiveness of the restrictive measures and obtain preliminary results on the impact of the measures; Based on the preliminary results of the measures' impact, assess the progress of ecological restoration and generate a restoration trend layer; A comprehensive ecological restoration distribution map is generated by overlaying historical data onto the restoration trend layer.
[0104] Furthermore, this embodiment uses a multispectral sensor to collect data from the protected area, acquiring image information covering different spectral bands and storing it as an initial dataset. Based on the initial dataset, the image information of different spectral bands is preprocessed to remove noise and invalid values, resulting in a cleaned image set. Using the cleaned image set, spectral features related to fecal freshness are extracted, and combined with preset classification criteria, the distribution of freshness indicators is determined. If the distribution of freshness indicators is lower than a preset threshold, the effectiveness of restrictive measures within the protected area is analyzed to obtain preliminary results on the impact of the measures. Based on the preliminary results on the impact of the measures, combined with environmental change data within the protected area, the progress of ecological restoration is judged, and a restoration trend layer is generated. The restoration trend layer is obtained, and historical monitoring data of the area is overlaid to generate a comprehensive ecological restoration distribution map, completing the assessment of the protected area.
[0105] Multispectral sensors acquire the spectral response of surface vegetation and organic matter by collecting reflectance data in the visible to near-infrared bands.
[0106] In one embodiment, the normalized vegetation index (NDI) and red-edge position offset are used as inputs. Principal component analysis (PCA) is used to reduce the dimensionality of the original image, removing noise caused by atmospheric scattering and outputting a reflectance matrix with a high signal-to-noise ratio. For fecal freshness analysis, the absorption peak depth of specific bands is extracted, such as the moisture absorption characteristics around 1450 nm. Combined with a random forest classifier, freshness is divided into a continuous interval from 0 to 1, with values closer to 1 indicating higher freshness. When the freshness index falls below a threshold of 0.3, an effectiveness assessment of the conservation area restriction measures is triggered.
[0107] Specifically, the biomass growth rate per unit area is calculated by comparing the changes in vegetation cover between grazing-prohibited and grazing-restricted areas.
[0108] In one possible implementation, historical monitoring data is used as a benchmark. A time-series analysis algorithm is employed, inputting image data from different time points to output an ecological restoration trend layer. This layer, by overlaying vector information of the grazing ban boundary, analyzes the difference in vegetation restoration rates inside and outside the boundary, thereby quantifying the contribution of the virtual fence to ecological restoration.
[0109] For example, in grazing-prohibited areas, if the average annual growth rate of vegetation coverage exceeds 5% and the spectral characteristics of soil organic matter are significantly enhanced, then the ecological restoration is considered to be progressing well.
[0110] It should be noted that the process of generating the comprehensive ecological restoration distribution map involves spatial registration and weighted fusion of multi-source data.
[0111] Preferably, the weights of each indicator are determined using the analytic hierarchy process (AHP), with vegetation cover weighted at 0.4, soil moisture content at 0.3, and manure degradation rate at 0.3. By mapping these indicators to a spatial grid, an intuitive ecological restoration level distribution map is output. This method can accurately identify ecologically sensitive areas under virtual fencing constraints, providing data support for subsequent adjustments to grazing ban boundaries. By analyzing the restoration trends in different areas, the pressure distribution of livestock activities on surface vegetation can be clearly identified, thereby achieving refined management of protected areas.
[0112] Furthermore, the process of determining the optimized grazing utilization rate includes: Data on the distribution characteristics of unused areas are obtained from ecological restoration assessment maps to determine their location and extent. Based on the location and extent information of the unused area, the unused area is divided into blocks, and the boundaries and area data of each block are determined. Based on the area data of each block, the grazing path is adjusted using a distribution balancing algorithm to obtain the adjusted path distribution scheme. Based on the adjusted path distribution scheme, the grazing utilization intensity of each block is calculated, and the path is locally optimized to obtain the optimized path plan. Based on the optimized route plan, the utilization rate of each block is calculated to determine the overall grazing utilization rate distribution of the region. Based on the overall regional grazing utilization rate distribution, the final utilization rate value and path adjustment results are generated.
[0113] Furthermore, in one embodiment, for the extraction of unused areas in the ecological restoration assessment graphic, an image segmentation algorithm is used to process the raster data. By setting a pixel grayscale threshold, the pixels representing unused areas in the graphic are clustered, and the vector boundary coordinates of the unused areas are output.
[0114] For example, areas in the graphic with gray values between 0.2 and 0.3 are identified as unused land and divided into several grid units with an area of 500 square meters, which serve as the basic spatial units for subsequent path planning.
[0115] Specifically, for adjusting rotational grazing routes, a distributional balancing algorithm is applied, using the centroid of each block as a node to construct a minimum spanning tree model. Given the area data and geographic coordinates of each block, the algorithm generates an initial route plan by calculating the Euclidean distance between nodes.
[0116] For example, if the area of a certain block exceeds the average by 15%, the algorithm automatically increases the path branch density in that area to ensure that the grazing path coverage is proportional to the block area.
[0117] In one possible implementation, grazing utilization intensity is calculated using an evaluation model based on livestock carrying capacity per unit area. The path length, grazing duration, and livestock number for each block are input, and the utilization intensity value is calculated through a weighted summation. If the utilization intensity of a block exceeds a preset threshold of 0.8, a local optimization mechanism is triggered. This mechanism adjusts the weights of path nodes, shifting paths in overloaded areas to adjacent low-utilization blocks until the utilization intensity of each block tends to be balanced.
[0118] For example, during path optimization, the target utilization rate range for each block is set to 0.4 to 0.6. If the calculated utilization rate for a certain block is 0.75, then by moving path nodes, some grazing activities are guided to neighboring blocks with a utilization rate of 0.25. Through iterative calculations, the path adjustment is completed until the overall area's utilization rate variance is less than 0.05. This process, by quantifying spatial distribution and utilization intensity, ensures a dynamic balance of grazing activities within the protected area, avoiding localized overgrazing that could disrupt ecological restoration.
[0119] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A method for quantifying regional grazing intensity based on unmanned aerial vehicle (UAV) observation of livestock manure, characterized in that, include: Acquire drone image data, extract fecal distribution features based on the drone image data, and obtain a fecal density distribution map; Calculate the grazing intensity value based on the fecal density distribution map to determine the current grazing intensity level; Identify activity preference areas based on the current grazing intensity level; The overload status is determined based on the activity preference area, and an overload area list is obtained; The warning activation zone is determined based on the overload zone list; Based on the planned rotational grazing routes in the early warning activation area, an ordered rotation sequence is obtained; The protection boundary is determined based on the ordered rotation sequence; An ecological restoration assessment map is generated based on the protection boundary; Adjust rotational grazing routes based on the aforementioned ecological restoration assessment map to determine the optimized grazing utilization rate.
2. The method according to claim 1, characterized in that, The process of obtaining a fecal density distribution map includes: Image data is acquired using drone equipment to obtain initial image files; Based on the initial image file, the fecal distribution area is identified, and a segmented fecal area image is obtained; Based on the segmented fecal region image, the density of fecal distribution is calculated to determine fecal density data; A density distribution matrix is constructed based on the fecal density data, and a density distribution map is generated. The density distribution map is filtered to obtain an optimized density distribution map; Based on the optimized density distribution map, spatial mapping processing is performed to output the final fecal density distribution map.
3. The method according to claim 1, characterized in that, The process of determining the current grazing intensity level includes: Density information is extracted from the fecal density distribution map to obtain a density distribution dataset; Based on the density distribution dataset, the density value per unit area is fitted to determine the grazing intensity value; Based on the grazing intensity values, high-intensity areas are marked, and the distribution information of high-intensity areas is obtained; Based on the distribution information of the high-intensity areas, local clustering phenomena are identified, and identification data of the local clustering areas are obtained; Based on the identification data of the local clustered areas, determine the spatial distribution characteristics; Based on the spatial distribution characteristics, determine the equilibrium state of grazing intensity levels; Based on the equilibrium state of the grazing intensity level, a regionalized intensity distribution map is generated to determine the final grazing intensity assessment result.
4. The method according to claim 1, characterized in that, The process of identifying active preference regions includes: Spatial distribution information was obtained from grazing intensity records to obtain a processed spatial distribution dataset; Based on the organized spatial distribution dataset, the data points are grouped to determine the preliminary intensity distribution categories; A spatial distribution matrix is generated based on the preliminary intensity distribution categories, and the spatial distribution matrix is converted into a heat map; Based on the heat map, the location information of high-intensity clustering areas is extracted to determine the range of activity preference areas; The region boundaries are divided based on the range of the activity preference region to determine the preference region boundaries; The average intensity value is calculated based on the boundary of the preferred region to obtain the intensity distribution result of the active preferred region.
5. The method according to claim 1, characterized in that, The process of obtaining the list of overloaded areas includes: Acquire grazing intensity data for the activity preference areas and determine the intensity monitoring results for each area; Based on the intensity monitoring results, the overload condition is determined, and the overload condition determination result is obtained; Based on the overload status determination results, the overload region is extracted to form an overload region set; Spatial clustering is performed on the set of overloaded regions to obtain overloaded region grouping information; Based on the overloaded area grouping information, priority sorting is performed to obtain a sorted list of overloaded areas; Based on the sorted list of overloaded areas, a preliminary plan for intervention strategies is generated, and the basis for allocating intervention strategies is determined. Based on the allocation criteria of the intervention strategies, the intervention strategies are matched to obtain a list of regional intervention plans.
6. The method according to claim 1, characterized in that, The process of determining the early warning activation zone includes: Real-time monitoring data is obtained from the overloaded area to obtain a set of density anomaly points; Based on the set of density anomaly points, filter out anomaly points that exceed the threshold to determine a list of high-risk density points; Based on the list of high-risk density locations, determine the degree of clustering between locations and mark potential warning areas; The density change trend is extracted from the potential warning area to obtain the density anomaly persistence assessment result; Based on the results of the density anomaly persistence assessment, an early warning signal is triggered to determine the activation range that requires close attention. Based on the activation range, generate the warning activation boundary and output the range division result.
7. The method according to claim 1, characterized in that, The process of obtaining an ordered rotation sequence includes: Based on the early warning activation signal, obtain the boundary data of the area division and determine the distribution of monitoring points; Based on the distribution of the monitoring points, terrain data is extracted to obtain information on terrain elevation and obstacle distribution; Based on the terrain elevation and obstacle distribution information, determine the suitable area for passage; Based on the suitable traversable area, a path optimization algorithm is used to calculate and obtain the rotational grazing route; An ordered sequence is generated based on the direction of the rotational grazing path, and the order of each path node is determined. Based on the ordered sequence, the logical arrangement of the rotation order is processed to generate rotation sequence data; The path nodes of the circumduction sequence data are verified to obtain a complete grazing path plan.
8. The method according to claim 1, characterized in that, The process of determining the protection boundary includes: Based on ordered and cyclic sequence data, initial regional division information is obtained, and preliminary distribution of grazing-prohibited and grazing-restricted areas is obtained. Based on the preliminary distribution of grazing-prohibited and grazing-restricted areas, the protection boundaries are adjusted to determine the final protection boundary range. Based on the final protection boundary range, the constraint range of the virtual fence is optimized to obtain the restriction data of the livestock activity range; Based on the restricted data of the livestock's activity range, adjust the deployment points of the virtual fence and determine the new activity range constraints; Based on the new activity range constraints, analyze the livestock activity trajectories to determine whether they comply with the activity range restrictions. Based on the analysis of livestock activity trajectories, new constraint strategies are determined through a dynamic adjustment mechanism. Based on the new constraint strategy, the grazing ban area and grazing restriction area are dynamically adjusted to obtain the latest regional distribution data.
9. The method according to claim 1, characterized in that, The process of generating an ecological restoration assessment map includes: Data was collected from the protected area using a multispectral sensor to obtain an initial dataset; Preprocessing is performed on the initial dataset to obtain a cleaned image set; Based on the image set after cleaning, spectral features related to fecal freshness are extracted to determine the distribution of freshness indicators. Based on the distribution of the freshness index, the effectiveness of the restrictive measures was analyzed, and preliminary results on the impact of the measures were obtained. Based on the preliminary results of the impact of the measures, the progress of ecological restoration is determined, and a restoration trend layer is generated; A comprehensive ecological restoration distribution map is generated by overlaying historical data onto the restoration trend layer.
10. The method according to claim 1, characterized in that, The process of determining the optimized grazing utilization rate includes: Data on the distribution characteristics of unused areas are obtained from ecological restoration assessment maps to determine their location and extent. Based on the location and extent information of the unused area, the unused area is divided into blocks, and the boundaries and area data of each block are determined. Based on the area data of each block, the grazing path is adjusted using a distribution balancing algorithm to obtain the adjusted path distribution scheme. Based on the adjusted path distribution scheme, the grazing utilization intensity of each block is calculated, and the path is locally optimized to obtain the optimized path plan. Based on the optimized path planning, the utilization rate of each block is calculated to determine the overall grazing utilization rate distribution of the region. Based on the distribution of grazing utilization rates in the overall region, the final utilization rate value and path adjustment results are generated.