Grassland human settlement environment ecological space identification method and system

Abstract

The present application provides a kind of prairie human settlement environment ecological space identification method and system, it is related to computer vision and ecological space analysis technical field, obtains the continuous video stream of prairie containing different function partition and time span, dynamic scene analysis is carried out to the video stream unit therein, the interactive action and state change of scene element are identified, and video stream scene interaction sequence is generated;Based on the interaction information in the video stream scene interaction sequence, associate ecological core element and derivative element, and construct ecological element dynamic correlation model;Video stream scene interaction sequence is input into the ecological element dynamic correlation model operation, and the trajectory of the change of ecological space form is output, generates preliminary identification result, and the model parameters are updated from this, obtain calibrated model and again operation generates final identification result.The present application solves the problem that traditional static method cannot capture the dynamic interaction of ecological process, realizes the dynamic, accurate and adaptive identification of prairie ecological space.

Description

Technical Field

[0001] This invention relates to the field of computer vision and ecological space analysis technology, and more specifically, to a method and system for identifying the ecological space of grassland human settlements. Background Technology

[0002] In the fields of grassland ecology research and human settlement planning, accurately identifying grassland ecological spaces is crucial for ecological protection, rational resource utilization, and sustainable development of human settlements. Traditional methods for identifying grassland ecological spaces mainly rely on static field survey data, such as regular sampling and manual observation to record grassland vegetation cover and animal activity ranges. However, these methods have significant limitations.

[0003] On the one hand, static data can only reflect the state of grassland at a specific moment and cannot capture the dynamic changes of grassland ecological elements over different time spans. Grassland ecosystems are complex dynamic systems, and the interactions and state changes of their ecological elements evolve continuously over time. Static data cannot fully and accurately present these dynamic characteristics.

[0004] On the other hand, traditional methods are difficult to effectively link various ecological elements in grassland human settlements. Grassland human settlements include core ecological elements (such as soil, water sources, and vegetation) and derivative elements (such as human activity facilities and livestock activity areas). There are complex dynamic relationships among these elements, and static data cannot deeply explore and analyze these relationships, thus limiting the accurate identification and understanding of the ecological space of grassland human settlements. Summary of the Invention

[0005] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a method for identifying the ecological space of grassland human settlements, the method comprising:

[0006] Acquire a continuous video stream of the grassland, which includes uninterrupted video stream units collected from different functional zones and different time spans of the grassland. Each video stream unit fully records the dynamic changes of the scene and the interaction process of elements in the corresponding area during the collection period.

[0007] Dynamic scene analysis is performed on each video stream unit in the continuous grassland video stream to identify the interactive actions and state changes of scene elements within each video stream unit, and to generate the video stream scene interaction sequence corresponding to each video stream unit.

[0008] Based on the scene element interaction information in the video stream scene interaction sequence, the core ecological elements and derived elements in the grassland human settlement environment are associated to construct an ecological element dynamic association model that describes the dynamic interaction relationship between elements.

[0009] The video stream scene interaction sequence is input into the ecological element dynamic association model. Through the calculation of the ecological element dynamic association model, the morphological change trajectory of the grassland human settlement ecological space is output, and the preliminary ecological space identification result is generated.

[0010] Based on the difference information between the preliminary ecological space identification results and the video stream scene interaction sequence, the function parameters of the element association in the dynamic association model of ecological elements are updated to obtain the calibrated dynamic association model of ecological elements. The video stream scene interaction sequence is then input into the calibrated dynamic association model of ecological elements again to generate the final identification result of the grassland human settlement environment ecological space.

[0011] Furthermore, embodiments of the present invention also provide a grassland human settlement environment ecological space identification system, characterized in that it includes:

[0012] A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to perform the above-described grassland human settlement ecological space identification method by executing the machine-executable instructions.

[0013] In another aspect, embodiments of the present invention also provide a computer program product, the computer program product including machine-executable instructions, the machine-executable instructions being stored in a computer-readable storage medium, a processor of a computer device reading the machine-executable instructions from the computer-readable storage medium, the processor executing the machine-executable instructions, causing the computer device to execute the above-described grassland human settlement ecological space identification method.

[0014] Based on the above, by acquiring continuous grassland video streams containing uninterrupted video stream units collected across different functional zones and time spans, the dynamic changes and element interactions of the corresponding area during the acquisition period can be fully recorded. Dynamic scene analysis is performed on each video stream unit to generate a video stream scene interaction sequence, enabling the identification of interactive actions and state changes of scene elements. Based on this, a dynamic ecological element association model is constructed, effectively linking the core and derived ecological elements in the grassland human settlement environment. This model accurately describes the dynamic relationships between elements. By inputting the video stream scene interaction sequence into the dynamic ecological element association model, the model outputs the morphological change trajectory of the grassland human settlement environment's ecological space and generates preliminary identification results, achieving dynamic identification of the grassland human settlement environment's ecological space. Furthermore, the model parameters are updated based on the differences between the preliminary identification results and the video stream scene interaction sequence, resulting in a calibrated model and generating the final identification result. This invention deeply integrates the dynamic analysis capabilities of computer vision, the process simulation capabilities of ecology, and the deductive capabilities of spatial analysis, thereby making the finally identified ecological space not only more accurate but also capable of reflecting real-world dynamic evolution, improving the accuracy and reliability of the identification. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the execution flow of the grassland human settlement environment ecological space identification method provided in the embodiments of the present invention.

[0016] Figure 2 This is a schematic diagram of exemplary hardware and software components of the grassland human settlement environment ecological space identification system provided in an embodiment of the present invention. Detailed Implementation

[0017] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating a method for identifying the ecological space of a grassland human settlement environment according to an embodiment of the present invention. The following is a detailed description of this method for identifying the ecological space of a grassland human settlement environment.

[0018] Step S110: Acquire a continuous video stream of the grassland. The continuous video stream of the grassland includes uninterrupted video stream units collected from different functional zones of the grassland and different time spans. Each video stream unit fully records the dynamic changes of the scene and the interaction process of elements in the corresponding area during the collection period.

[0019] In this embodiment, the continuous video stream of the grassland is collected by multiple sets of high-definition monitoring devices deployed throughout the grassland. These devices are evenly distributed according to a preset grid spacing, covering four functional zones: native vegetation areas, human settlement areas, livestock activity areas, and water resource conservation areas. Each video stream unit corresponds to a single monitoring device continuously collecting uninterrupted video data for a duration of T, encompassing the dynamic changes of natural and man-made elements within the area covered by that device throughout the entire time period. Specifically, it covers the complete interaction process of elements such as the growth and decay of native vegetation, the movement and foraging of livestock, the usage status of human settlements, and the flow and evaporation of surface water. The frame rate of each video stream unit is uniformly fixed to ensure consistent time intervals between individual frames. All video stream units use the same encoding format and resolution to ensure format uniformity in subsequent data processing. During the data acquisition phase, real-time differential privacy technology is used to blur sensitive information such as faces and license plates in the video stream, potentially involving privacy information related to human activities. Simultaneously, a collaborative edge-to-edge encryption transmission mechanism ensures secure data transmission and storage, preventing privacy data leakage.

[0020] Step S120: Perform dynamic scene analysis on each video stream unit in the continuous grassland video stream, identify the interactive actions and state changes of scene elements within each video stream unit, and generate the video stream scene interaction sequence corresponding to each video stream unit.

[0021] Step S121: Separate the independent video frames contained in each video stream unit of the grassland continuous video stream frame by frame according to the time sequence, and completely preserve the pixel information and color channel information of each independent video frame. Arrange the separated independent video frames in the original time sequence to form an ordered video frame sequence corresponding to each video stream unit.

[0022] In this embodiment, a video frame separation tool is used to process each video stream unit frame by frame. During the processing, the pixel value information of the three RGB color channels of each video frame is fully preserved, as well as the additional attribute information such as brightness, saturation, and contrast of each pixel within the frame. The separated independent video frames are arranged sequentially according to the time of acquisition, and each video frame is assigned a unique timestamp. This timestamp contains the unique code of the video stream unit and the frame acquisition sequence number, ensuring that the time order of the ordered video frame sequence is not disordered. The arranged ordered video frame sequence is stored in the form of structured folders, with each folder corresponding to a video stream unit. Each video frame file in the folder is named with a timestamp for easy subsequent retrieval and retrieval.

[0023] Step S122: Extract scene composition information of each video frame in the ordered video frame sequence, traverse each pixel of the video frame, identify different scene components within the video frame based on a preset image feature difference threshold, record the pixel distribution range, color feature value, and contour edge information of each scene component, and generate a structured scene composition information dataset. The scene components include three categories: artificial facility form, natural vegetation appearance, and land cover.

[0024] In this embodiment, a U-Net-based image semantic segmentation model is used to extract scene composition information from each video frame. It comprises four parts: an input layer, an encoder module, a decoder module, and an output layer. The encoder module extracts multi-scale features of the image through convolution and pooling operations, while the decoder module restores the spatial resolution of the image through upsampling and stitching operations, ultimately outputting a segmentation result containing the categories of scene components. The model's input is the RGB tensor data of a single video frame, and the output is a category mask tensor of the same size as the input video frame. The value of each pixel corresponds to its respective scene component category. When applying this model, the pixel values ​​of the video frame are first normalized and fed into the model's input layer. After processing by the encoder and decoder modules, the output layer uses the Softmax function to obtain the category probability distribution of each pixel, selecting the category with the highest probability value as the pixel's category to generate a category mask. Subsequently, based on the category mask, each pixel of each video frame is traversed to calculate the pixel distribution range of each scene component. Specifically, this involves determining the minimum and maximum x and y coordinates of each scene component within the video frame, thereby defining the rectangular area it covers. The color feature values ​​of each scene component are calculated, specifically the mean and standard deviation of the RGB color channels of all pixels within that component. The contour edge information of each scene component is extracted, specifically by using the Canny edge detection algorithm to identify the set of edge pixel coordinates for that component. These three types of information are then organized according to the category of scene components to generate a structured scene composition information dataset containing category identifiers, pixel distribution ranges, color feature values, and contour edge information.

[0025] Step S123: Compare the scene composition information datasets of two adjacent video frames, compare the changes in pixel distribution range, color feature value fluctuations, and contour edge offsets of the scene components dimension by dimension, capture dynamic change signs of the scene components, mark the starting and ending pixel coordinates of each dynamic change sign, and define the spatial range of the change. The dynamic change signs include three categories: position movement, shape change, and appearance / disappearance.

[0026] In this embodiment, a dimensional comparison method is used to compare the scene composition information datasets of two adjacent video frames. First, the pixel distribution range changes of the scene components are compared. Specifically, the difference between the minimum and maximum horizontal and vertical coordinates of the same scene component in the two consecutive frames is calculated. If the difference exceeds a preset pixel range change threshold, it is determined that the scene component has dynamic changes such as position movement or shape change. Next, the color feature value fluctuations of the scene components are compared. Specifically, the difference between the mean and standard deviation of the three RGB color channels of the same scene component in the two consecutive frames is calculated. If the difference exceeds a preset color feature fluctuation threshold, it is determined that the scene component has dynamic changes such as shape change or disappearance. Finally, the contour edge offset of the scene components is compared. Specifically, the overlap of the edge pixel coordinate set of the same scene component in the two consecutive frames is calculated. If the overlap is lower than a preset contour edge overlap threshold, it is determined that the scene component has dynamic changes such as position movement or shape change. For captured dynamic change indicators, their start and end pixel coordinates are marked. The start pixel coordinates are the coordinates of the starting pixel position of the change indicator in the current frame, and the end pixel coordinates are the coordinates of the ending pixel position of the change indicator in the current frame. The rectangular spatial range covered by the dynamic change indicator is determined by the start and end pixel coordinates. Furthermore, based on the manifestation of the dynamic change indicator, it is divided into three categories: position movement, shape change, and appearance / disappearance. Position movement refers to a change in the overall position of a scene component without a change in shape; shape change refers to a change in the shape of a scene component without a change in its overall position; and appearance / disappearance refers to a scene component appearing for the first time or disappearing completely in the current frame.

[0027] Step S124: Mark the scene components corresponding to each dynamic change sign, associate the same scene components in the previous and next frames, record three types of quantitative information of the scene component in the adjacent video frames: position offset, morphological deformation degree, and color change amplitude, and organize them into a single frame change record in a unified format.

[0028] In this embodiment, firstly, a corresponding scene component category identifier is added to each dynamic change sign to clarify the category of artificial facility form, natural vegetation appearance, or land cover to which the dynamic change sign belongs. Subsequently, a feature matching algorithm is used to associate the same scene components in consecutive frames. Specifically, the color feature values ​​and contour edge information of scene components of the same category in consecutive frames are extracted, the feature similarity between the two is calculated, and the scene component with the highest similarity is selected as the association result of the same scene component. For each associated scene component, its positional offset in adjacent video frames is calculated. Specifically, this is the difference between the center pixel coordinates of the scene component in the current frame and the center pixel coordinates of the corresponding scene component in the previous frame. This difference is in vector form, containing offsets in both horizontal and vertical dimensions. The degree of morphological deformation is also calculated, specifically the Hausdorff distance between the set of outline edge pixels of the scene component in the current frame and the set of outline edge pixels of the corresponding scene component in the previous frame. This distance reflects the degree of difference between the two outlines, i.e., the degree of morphological deformation. Finally, the magnitude of color change is calculated, specifically the average of the absolute differences between the mean values ​​of the three RGB color channels of the scene component in the current frame and the mean values ​​of the three RGB color channels of the corresponding scene component in the previous frame. This average value reflects the overall magnitude of the color change. The scene component category identifier, positional offset, degree of morphological deformation, and magnitude of color change are organized in a unified JSON format to form a single-frame change record. Each single-frame change record corresponds to the dynamic change information of a set of the same scene component in two adjacent video frames.

[0029] Step S125: Integrate the single-frame change records of all adjacent video frames in the ordered video frame sequence, arrange them sequentially according to the time order of the video frames, and construct a continuous change trajectory data chain.

[0030] In this embodiment, the single-frame change records of all adjacent video frames in the ordered video frame sequence are arranged sequentially according to the time order of the video frames. During the arrangement process, the timestamp identifier of each single-frame change record is retained. This timestamp identifier contains the timestamp information of the previous frame and the current frame, ensuring the temporal continuity of the change trajectory data chain. The arranged change trajectory data chain is stored in the form of a linked list. Each node in the linked list corresponds to a single-frame change record. Nodes are associated with each other through timestamp identifiers, which facilitates tracing the dynamic change trajectory of scene components from any time point. At the same time, an index structure is established for the change trajectory data chain. The index structure is classified according to the category identifier of scene components. Each category contains pointers to the change trajectory nodes of all scene components in that category, which facilitates quick retrieval of dynamic change information of scene components of a specific category.

[0031] Step S126: Identify the interaction actions between the scene components in the change trajectory data chain. Based on the positional relationship, change sequence, and morphological correlation of the scene components, distinguish the interaction forms between three types of scene elements: occlusion, contact, and accompanying movement. Collect the judgment criteria parameters and feature presentation data for each interaction form.

[0032] Step S1261: Extract the position coordinate information of all scene components in the change trajectory data chain, establish a unified coordinate reference system based on the pixel coordinate system of video frames, obtain the boundary coordinates and center coordinates of each scene component, and classify and store them according to the scene component identifier and time node.

[0033] In this embodiment, the position coordinate information of the corresponding scene component is extracted from each node of the change trajectory data chain. Specifically, this includes the minimum and maximum x and y coordinates of the scene component in the corresponding video frame, as well as the center pixel coordinates. A unified coordinate reference system is established based on the pixel coordinate system of the video frame. This coordinate reference system has the top left corner of the video frame as the origin, the x-axis as the horizontal axis, and the y-axis as the vertical axis. The coordinates of a pixel are the number of pixels on the x and y axes. Based on the position coordinate information, the boundary coordinates of each scene component are calculated, specifically the pixel coordinates of the four vertices of the area covered by the scene component; the center coordinates are also calculated, specifically the coordinates of the center pixel of the area covered by the scene component. The boundary coordinates and center coordinates of each scene component are classified and stored according to the unique identifier of the scene component and the corresponding time node. The storage structure adopts a multi-dimensional array form, where the first dimension of the array is the unique identifier of the scene component, the second dimension is the time node identifier, and the third dimension is the specific value of the boundary coordinates and center coordinates.

[0034] Step S1262: Track the position coordinate changes of each scene component in consecutive video frames, and generate an independent movement trajectory for each scene component based on the time series of coordinate information. The independent movement trajectory contains three types of information: position coordinates, movement direction, and movement distance at each time node.

[0035] In this embodiment, for each scene component, its center coordinates at consecutive time nodes are extracted from the categorized and stored location coordinate information to form a time series of center coordinates. Based on this time series, the movement direction of each time node relative to the previous time node is calculated, specifically by determining the angle value of the movement direction using the ratio of the horizontal difference to the vertical difference of the center coordinates; the movement distance is calculated, specifically the Euclidean distance between the center coordinates of the current time node and the center coordinates of the previous time node. The center coordinates, movement direction, and movement distance corresponding to each time node are organized into independent movement trajectories in chronological order. Each independent movement trajectory contains a unique identifier for the scene component, a list of time nodes, and the location coordinates, movement direction, and movement distance information corresponding to each time node.

[0036] Step S1263: Compare the independent movement trajectories of any two scene components, calculate the boundary coordinate distance and center coordinate distance between the two in the same video frame, record the sequence of distance data changes over time, and generate a distance change dataset.

[0037] In this embodiment, the independent movement trajectories of all scene components are traversed, and pairwise comparisons are performed on any two scene components. During the comparison, the position coordinate information at the same time node is selected for calculation. For the same time node, the boundary coordinate distance between the two scene components is calculated, specifically the minimum Euclidean distance between the boundary vertex coordinates of the two scene components; the center coordinate distance is also calculated, specifically the Euclidean distance between the center coordinates of the two scene components. The boundary coordinate distance and center coordinate distance at each time node are recorded sequentially to form a sequence of distance data changing over time. This sequence contains the unique identifiers of the two scene components, a list of time nodes, and the boundary coordinate distance and center coordinate distance information corresponding to each time node. All the distance change sequences obtained from the pairwise comparisons are organized to generate a distance change dataset. This distance change dataset is stored in the form of a structured database table, with each record corresponding to a set of distance change sequences between two scene components.

[0038] Step S1264: Observe the changing trend of the position coordinate distance, process the distance change dataset, and when the distance shrinks to the preset first distance threshold and the number of consecutive frames reaches the preset first frame number threshold, combine the overlapping signs of the morphological features of the scene components to determine that the two scene components have made contact, and collect the contact action determination threshold parameters and the number of consecutive frames parameters.

[0039] In this embodiment, a sliding window analysis method is used to process each distance change sequence in the distance change dataset. The size of the sliding window is a preset first frame number threshold. By traversing the distance change sequence through the sliding window, the changing trends of the boundary coordinate distance and the center coordinate distance within the window are observed. When the boundary coordinate distance of all time nodes within the window shrinks to below the preset first distance threshold, and the number of frames covered by the window reaches the preset first frame number threshold, the morphological feature information of the two scene components in the video frame corresponding to the window is further extracted. Specifically, this is the set of contour edge pixels of the two scene components, and the intersection ratio of the two contour edge pixel sets is calculated. This intersection ratio reflects the overlap of morphological features. When the intersection ratio exceeds a preset overlap ratio threshold, it is determined that the two scene components have made contact. The contact action determination threshold parameters are collected, specifically the first distance threshold and the overlap ratio threshold; the duration frame number parameter is collected, specifically the first frame number threshold. The above parameters are associated with and stored with the contact action determination results for subsequent model calibration and optimization.

[0040] Step S1265: When the boundary coordinates of one scene component completely cover the boundary coordinates of another scene component, and the covered part cannot be observed by pixel feature detection in subsequent consecutive video frames, the occlusion action is determined by combining the positional relationship and movement trajectory of the two scene components, and the coverage range determination parameters and observation disappearance duration parameters of the occlusion action are collected.

[0041] In this embodiment, each distance change sequence in the distance change dataset is traversed, and the boundary coordinate relationship between the two scene components at each time point is examined. When the rectangular area corresponding to the boundary coordinate of one scene component completely contains the rectangular area corresponding to the boundary coordinate of another scene component, pixel feature detection is further performed on multiple subsequent consecutive video frames. Specifically, the color feature values ​​and contour edge information of the covered scene component in the subsequent video frames are extracted. If pixel feature information matching the original features of the covered scene component cannot be detected in multiple consecutive video frames, the positional relationship and movement trajectory of the two scene components are combined for judgment. If the movement trajectory of the two scene components shows that the covered scene component has not moved or the movement direction is towards the covering scene component, then an occlusion action is determined to have occurred. Coverage range judgment parameters for occlusion actions are collected, specifically the coverage ratio threshold of the boundary coordinates, that is, when the proportion of the boundary area of ​​the covering scene component covering the boundary area of ​​the covered scene component reaches the threshold, it is determined to be complete coverage. Observation disappearance duration parameters are collected, specifically the threshold of the number of subsequent consecutive video frames in which the features of the covered scene component cannot be detected. The above parameters are associated with the occlusion action judgment results and stored for subsequent model calibration and optimization.

[0042] Step S1266: When the angle between the movement directions of the independent movement trajectories of the two scene components is less than a preset angle threshold, and the fluctuation amplitude of the position coordinate distance is less than a preset distance fluctuation threshold, verify that the duration of the corresponding state reaches a preset second frame number threshold, determine that an accompanying movement action has occurred, and collect the allowable deviation parameters of direction consistency and the fluctuation threshold parameters of distance stability.

[0043] In this embodiment, the independent movement trajectories of all scene components are traversed. For any two scene components, their independent movement trajectories are compared pairwise to extract the movement direction information of the two scene components at the same time node. The angle between the movement directions is calculated, reflecting the degree of difference between the two movement directions. Simultaneously, the center coordinate distance information of the two scene components at consecutive time nodes is extracted, and the fluctuation amplitude of the center coordinate distance is calculated, specifically the standard deviation of the center coordinate distance at consecutive time nodes. This standard deviation reflects the stability of the distance. When the angle between the movement directions is less than a preset angle threshold, and the standard deviation of the center coordinate distance is less than a preset distance fluctuation threshold, a sliding window analysis method is used to verify the duration of this state. The size of the sliding window is a preset second frame number threshold. When all time nodes within the window meet the above two conditions, it is determined that the two scene components have undergone accompanying movement. Allowable deviation parameters for direction consistency are collected, specifically the angle threshold; fluctuation threshold parameters for distance stability are collected, specifically the distance fluctuation threshold and the second frame number threshold. These parameters are associated and stored with the determination results of accompanying movement for subsequent model calibration and optimization.

[0044] Step S1267: Record the starting video frame number and ending video frame number of each interaction action. Based on the time nodes of the action occurrence and termination, calculate the duration of the action, accurate to the smallest time unit of the video frame.

[0045] In this embodiment, for the three types of interaction actions—contact, occlusion, and accompanying movement—the starting and ending video frame numbers of the action are extracted from the trajectory change data chain. The starting video frame number is the sequential number of the video frame that first meets the action determination condition, and the ending video frame number is the sequential number of the video frame that last meets the action determination condition. Based on the starting and ending video frame numbers, the duration of the action is calculated, specifically the time difference between the acquisition time of the starting and ending video frames. This time difference is accurate to the smallest time unit of the video frame, i.e., the acquisition interval between two adjacent video frames. The starting video frame number, ending video frame number, duration interval, and action determination result for each interaction action are associated and stored to form a complete interaction action record.

[0046] Step S1268: Extract three types of parameters—contour shape, color distribution, and texture details—of the two scene components when the interaction occurs. Compare the values ​​of these three types of parameters before and after the interaction to generate a dataset of morphological feature changes.

[0047] In this embodiment, for the two scene components that interact, three types of parameters are extracted: contour shape, color distribution, and texture details at the time point before the action, at each time point during the action, and at the time point after the action. Contour shape parameters specifically include the set of pixels at the contour edges of the scene component, the perimeter of the contour, and the area. Color distribution parameters specifically include the mean, standard deviation, and histogram distribution of the three RGB color channels of the scene component. Texture details parameters specifically include the gray-level co-occurrence matrix features of the scene component, including contrast, correlation, energy, and homogeneity. The same parameter before and after the action is compared, and the change amount and rate of change are calculated. The change amount is the difference between the parameter value after the action and the parameter value before the action, and the rate of change is the ratio of the change amount to the parameter value before the action. All parameter changes and rates of change are organized by time point to generate a morphological feature change dataset. This dataset contains unique identifiers for the two scene components, a list of time points, and the values, changes, and rates of change of the three types of parameters corresponding to each time point.

[0048] Step S1269: Analyze the changes in morphological features during the interaction process, calculate the magnitude and rate of change of morphological feature parameters, determine the degree of influence of the action on the morphology of the scene components, classify the level of influence, and collect the level standard parameters.

[0049] In this embodiment, for each parameter in the morphological feature change dataset, the magnitude of its change during the interaction process is calculated. Specifically, this is the difference between the maximum and minimum values ​​of the parameter during the action, reflecting the overall range of parameter change. The rate of change is also calculated, specifically the ratio of the magnitude of change to the duration of the action, reflecting how quickly the parameter changes. Based on the combined value of the magnitude and rate of change, the degree of influence of the interaction action on the morphology of the scene's constituent parts is determined. Specifically, the magnitude and rate of change are weighted and summed to obtain a comprehensive influence value. The weights of this weighted sum are set according to the importance of the parameter to the morphology of the scene's constituent parts. Based on the magnitude of the comprehensive influence value, the degree of influence is divided into four levels: no influence, slight influence, moderate influence, and severe influence. Each level corresponds to a range of comprehensive influence values, which are the level standard parameters. The influence level and level standard parameters are associated and stored with the determination results of the interaction action for subsequent ecological element association analysis.

[0050] Step S12610: Integrate the interaction action type, time interval, identification of the constituent parts of the scene, degree of influence of form and judgment basis parameters, and construct the interaction form record between scene elements using a standardized format.

[0051] In this embodiment, the types of interaction actions (contact, occlusion, accompanying movement), the duration of the action, the unique identifiers of the two scene components involved, the degree of morphological influence, and the judgment criteria parameters (including various thresholds, frame rate parameters, etc.) are organized using a unified XML standardized format to construct a record of interaction forms between scene elements. Each interaction form record contains a root node and multiple child nodes. The root node is the unique identifier of the interaction form, and the child nodes correspond to information such as action type, time interval, identifiers of the participating scene components, degree of morphological influence, and judgment criteria parameters. Each child node contains corresponding attribute values ​​and descriptive text. The constructed record of interaction forms between scene elements is stored in chronological order to form an interaction form record sequence, which serves as the core component of the video stream scene interaction sequence.

[0052] Step S127: Record the identifiers of the participating scene components, the starting video frame number, the ending video frame number, the duration of the action, and the state change data of each participating component for each interaction action, and organize them in chronological order to form a scene interaction record.

[0053] In this embodiment, the unique identifiers of the participating scene components, the starting video frame number, the ending video frame number, the duration of the action, and the state change data such as the positional offset, morphological deformation degree, and color change amplitude of each participating component during the action are organized. During the organization process, the data is arranged in chronological order of the actions. A unique action identifier number is assigned to each scene interaction record, which includes the action type code and the action sequence number. The organized scene interaction records are stored in the form of a structured database table, with each record corresponding to complete information about one interaction action.

[0054] Step S128: Align the change trajectory data chain with the scene interaction record on the time axis, integrate the aligned change trajectory data chain and scene interaction record in chronological order, organize the dynamic change information and interaction information into a scene dynamic dataset in a unified format, perform serialization processing on the scene dynamic dataset, and generate the video stream scene interaction sequence corresponding to each video stream unit.

[0055] In this embodiment, a timeline alignment algorithm is used to align the change trajectory data chain with the scene interaction record. During the alignment process, the timestamp of the video frame is used as a reference to match the action information of each node in the change trajectory data chain with the corresponding time node in the scene interaction record. After alignment, the dynamic change information of the change trajectory data chain and the interaction information of the scene interaction record are integrated in chronological order. During the integration process, the original identifier and attribute information of each data are preserved, and the above information is organized into a scene dynamic dataset in a unified format. This scene dynamic dataset is stored in the form of a multidimensional tensor. The first dimension of the tensor is the time node, the second dimension is the unique identifier of the scene component, and the third dimension is the specific parameter value of the dynamic change information and interaction information. Subsequently, the scene dynamic dataset is serialized, specifically, the multidimensional tensor is converted into serializable byte stream data, encoded using a preset encoding format, to generate a video stream scene interaction sequence corresponding to each video stream unit.

[0056] Step S130: Based on the scene element interaction information in the video stream scene interaction sequence, associate the core ecological elements and derived elements in the grassland human settlement environment, and construct an ecological element dynamic association model that describes the dynamic interaction relationship between elements.

[0057] Step S131: Collect relevant data on the core ecological elements and related elements derived from the core elements that maintain grassland ecological stability. The core ecological elements include grassland vegetation type, soil water retention capacity, and distribution of protozoa. The derived ecological elements include vegetation growth rate, soil moisture change, and animal activity range. Data is collected through three methods: grassland ecological survey literature retrieval, on-site ecological condition observation and recording, and acquisition through ecological monitoring data sharing platform.

[0058] In this embodiment, a grassland ecological survey literature search is used to extract basic attribute data, characteristic parameter data, and interaction relationship data of core ecological elements and derived ecological elements from published academic papers, research reports, ecological survey manuals, and other literature. Through on-site ecological observation and recording, a professional ecological observation team sets up multiple observation points throughout the grassland area to regularly collect measured data on core ecological elements such as grassland vegetation type, soil water retention capacity, and protozoan distribution, as well as measured data on derived ecological elements such as vegetation growth rate, soil moisture changes, and animal activity range. The observation period for each observation point is fixed, and the observation data includes information such as observation time, observation point location, and specific values ​​of the elements. Long-term ecological monitoring data for the entire grassland area is obtained from a public ecological monitoring data sharing platform. This data includes real-time and historical data collected by automatic monitoring equipment, covering multi-dimensional information on core ecological elements and derived ecological elements. All collected data is classified and organized according to element type, and duplicate and invalid data are removed to form a standardized ecological element dataset.

[0059] Step S132: Traverse each data node of the video stream scene interaction sequence, extract the interaction form, interaction frequency, and interaction impact range between the scene components, and generate a set of scene element interaction information.

[0060] In this embodiment, all data nodes of each video stream scene interaction sequence are traversed, and information such as the interaction form (contact, occlusion, accompanying movement), the time node of the interaction, and the identifiers of the participating scene components are extracted from each data node. Based on this information, the interaction frequency of each interaction form is calculated, specifically the ratio of the total number of times the interaction form occurs in the entire video stream scene interaction sequence to the total duration of the video stream; the interaction impact range is calculated, specifically the ratio of the total area covered by the participating scene components when the interaction occurs to the total area covered by the video stream unit. The interaction form, interaction frequency, and interaction impact range are organized according to the category of interaction form to generate a scene element interaction information set. This scene element interaction information set includes information such as interaction form identifier, interaction frequency, interaction impact range, and corresponding interaction record list. Each interaction record list contains specific information about all occurrence instances of that interaction form.

[0061] Step S133: Through feature matching, determine the correspondence between the core ecological elements and the interaction information set of scene elements. Compare the feature parameters of each core ecological element with the attribute parameters of the interaction information of scene elements one by one. Determine the scene element interaction actions with a matching degree higher than the preset threshold as actions that directly affect the state of the corresponding core ecological elements, and generate a preliminary list of associated correspondences.

[0062] In this embodiment, a cosine similarity matching algorithm is used to compare the feature parameters of the core ecological elements with the attribute parameters of the scene element interaction information one by one. The feature parameters of the core ecological elements include the species composition, coverage, and growth status of grassland vegetation types; soil texture, bulk density, and porosity of soil water retention capacity; and species types, population density, and activity areas of protozoa. The attribute parameters of the scene element interaction information include the interaction form, interaction frequency, interaction impact range, and the category of the scene components involved. The cosine similarity between the feature parameters of each core ecological element and the attribute parameters of each scene element interaction information is calculated. This similarity value reflects the degree of matching between the two. When the similarity value is higher than a preset matching threshold, it is determined that the scene element interaction action directly affects the state of the corresponding core ecological element. The unique identifier of the core ecological element, the unique identifier of the scene element interaction action, and the matching value are recorded to generate a preliminary association list. This association list is stored in tabular form, with each record corresponding to a set of association relationships between core ecological elements and scene element interaction actions.

[0063] Step S134: Based on the preliminary associated list, label the change type of the core ecological element for each scene element interaction action, collect the characteristic index data and judgment standard parameters of the change type, map the interaction action to the change type one by one, and generate a structured mapping table.

[0064] In this embodiment, for each set of relationships in the preliminary association list, combined with ecological mechanism knowledge and measured data, the type of change in the core ecological element caused by the interaction of scene elements is determined. For example, the contact action between livestock herds and native vegetation may cause a decrease in the coverage of grassland vegetation, and the contact action between livestock herds and soil may cause a decrease in the porosity of soil water retention capacity. Characteristic index data for each type of change is collected, specifically the range and trend of changes in the characteristic parameters of the corresponding core ecological element. Judgment standard parameters are also collected, specifically the threshold values ​​for the characteristic parameters that determine if the core ecological element has undergone this type of change. For example, when the decrease in grassland vegetation coverage exceeds a preset threshold, it is determined to be a decrease in coverage. The unique identifier of the scene element interaction action, the type of change in the core ecological element, the characteristic index data, and the judgment standard parameters are mapped one-to-one to generate a structured mapping table.

[0065] Step S135: Analyze the relationship between ecological derivative elements and ecological core elements. By combining ecological mechanism analysis and data statistical analysis, analyze the process by which changes in ecological core elements lead to changes in ecological derivative elements. Collect data on the transmission of changes between the two, and establish a correlation between the change characteristics of ecological core elements and the change characteristics of ecological derivative elements based on the statistical analysis results, and generate a correlation analysis report.

[0066] In this embodiment, an ecological mechanism analysis method is first employed. Based on existing grassland ecosystem theory, the intrinsic mechanism by which changes in core ecological elements trigger alterations in derived ecological elements is analyzed. For example, a decrease in grassland vegetation cover leads to increased soil moisture evaporation, resulting in an increase or decrease in soil humidity; changes in protozoan population density affect soil nutrient cycling, thus causing changes in vegetation growth rate, etc. Subsequently, a statistical data analysis method is used to extract historical observation data of core and derived ecological elements from the ecological element dataset, constructing time series of their characteristic parameters and calculating the Pearson correlation coefficient between the time series. This correlation coefficient reflects the degree of linear association between the two. Combining the results of the ecological mechanism analysis and the magnitude of the Pearson correlation coefficient, a correlation relationship is established between the characteristics of changes in core ecological elements and the characteristics of changes in derived ecological elements. For example, when the correlation coefficient between the type of grassland vegetation cover decrease and the type of vegetation growth rate decrease is higher than a preset threshold, a correlation relationship is established. Data on the transmission of changes between the two is collected, specifically the time difference between the occurrence of changes in core ecological elements and changes in derived ecological elements, and the ratio of the magnitude of changes in core ecological elements to the magnitude of changes in derived ecological elements, etc. The correlation and correspondence, change transmission data, ecological mechanism analysis results, and data statistical analysis results are compiled to generate a correlation analysis report. This correlation analysis report is stored in PDF format and includes text descriptions, charts, data tables, and other content.

[0067] Step S136: Based on the correlation analysis report, track the specific links and processes by which changes in each core ecological element trigger changes in corresponding derivative ecological elements, mark key node data and influencing factor data in the path, and generate element transmission links.

[0068] In this embodiment, based on the correlation and change transmission data in the correlation analysis report, a path tracing method is used to track the specific links and processes by which changes in each core ecological element trigger changes in corresponding derivative ecological elements. Specifically, starting from the initial time node of the change in the core ecological element, the nodes where changes in derivative ecological elements occur are sequentially searched in subsequent time nodes to determine the temporal order and causal relationship between the two. During the tracing process, key node data in the path is marked, specifically the time nodes when significant changes occur in the core ecological element and the derivative ecological element, as well as the corresponding element characteristic parameter values. Influencing factor data is also marked, specifically the characteristic parameter values ​​of other elements that may affect the transmission effect during the change process, such as the values ​​of meteorological elements (temperature, precipitation, wind speed, etc.). The key node data and influencing factor data are organized in chronological order to generate an element transmission link. This element transmission link is represented in the form of a directed graph, where the nodes are key nodes, the directed edges between nodes represent the transmission direction of element changes, and the attributes on the edges are the influencing factor data and the transmission time difference information.

[0069] Step S137: Integrate the mapping table and the element transmission link to form an element association structure, and organize the connection relationship between the core ecological elements, the derived ecological elements, and the interactive information of the scene elements in a structured manner.

[0070] In this embodiment, the structured mapping table and element transmission links are integrated to organize the interaction actions of scene elements, the change types of core ecological elements, the change types of core ecological elements, the change types of derived ecological elements, and the connection relationships between them in a structured manner, forming an element association structure. This element association structure is represented by a hierarchical tree structure. The root node is the scene element interaction information, the first-level child nodes are the corresponding change types of core ecological elements, the second-level child nodes are the corresponding core ecological elements, the third-level child nodes are the corresponding change types of derived ecological elements, and the fourth-level child nodes are the corresponding derived ecological elements. The connection edges between each node contain attribute information such as association strength, transmission time, and triggering conditions.

[0071] Step S138: Based on the basic attributes of ecological elements and actual data of scene interactions, collect scene condition data and element status data of each connection relationship of the triggering element association structure, and record specific trigger condition description data and judgment standard parameters.

[0072] In this embodiment, for each connection in the element association structure, combined with the basic attributes of ecological elements and actual data of scene interactions, scene condition data that triggers the connection is collected. Specifically, this includes environmental conditions when scene element interaction occurs, such as meteorological conditions and terrain conditions. Element state data is also collected, specifically the characteristic parameter values ​​of core ecological elements or derived ecological elements when the connection is triggered. Specific trigger condition description data is recorded. For example, when the contact action between livestock herds and native vegetation occurs under high-temperature, low-rainfall meteorological conditions, and the grassland vegetation coverage is within a preset low range, the connection between this contact action and the type of grassland vegetation coverage decrease is triggered. Judgment standard parameters are also recorded, specifically thresholds for meteorological conditions and thresholds for ecological element characteristic parameters. The trigger condition description data and judgment standard parameters are then associated and stored with the corresponding connection relationships in the element association structure.

[0073] Step S139: Based on the changing trends and quantitative indicators of ecological elements and the dynamic adjustment needs of their relationships, determine the timing data, scope data, and method data for adaptive adjustment of the element relationship structure after the element status changes, and generate a set of update rules.

[0074] Step S1391: Perform statistical analysis on the historical change data of the core ecological elements, and classify the state changes into gradual changes and abrupt changes based on a preset threshold for the change per unit time; wherein, state changes with a change per unit time lower than the threshold for the change per unit time are determined to be gradual changes, and state changes with a change per unit time higher than or equal to the threshold for the change per unit time are determined to be abrupt changes.

[0075] In this embodiment, a sliding window statistical method is used to statistically analyze the historical change data of core ecological elements. The size of the sliding window is a preset unit time length. By traversing the time series of characteristic parameters of the core ecological elements through the sliding window, the unit time change within each window is calculated, specifically the ratio of the change in the characteristic parameter within the window to the unit time length of the window. The calculated unit time change is compared with a preset unit time change threshold. When the unit time change is lower than the threshold, the change in the state of the core ecological element within that window is determined to be a gradual change; when the unit time change is higher than or equal to the threshold, it is determined to be a sudden change. The identifiers of gradual and sudden changes are associated and stored with the corresponding time windows.

[0076] Step S1392: For gradual changes, based on the range of changes per unit time corresponding to the determined gradual changes, set the corresponding update cycle, determine the update timing of the core ecological elements, and collect the specific duration data and adjustment condition data of the update cycle.

[0077] In this embodiment, for gradual changes, the change per unit time is divided into multiple range intervals, each corresponding to a preset update cycle. The smaller the change per unit time, the longer the corresponding update cycle. For example, when the change per unit time is in the first range interval (the smallest change range), the corresponding update cycle is longer; when the change per unit time is in the second range interval, the corresponding update cycle is medium; and when the change per unit time is in the third range interval, the corresponding update cycle is shorter. Based on the range of change per unit time corresponding to the gradual change, the corresponding update cycle is determined. This update cycle is the update timing for the core ecological element; that is, every update cycle, the connection relationships related to the core ecological element in the element association structure are updated once. Specific duration data of the update cycle is collected, i.e., the duration value of the update cycle corresponding to each range interval; adjustment condition data is also collected, i.e., the conditions for adjusting the update cycle when the change per unit time shifts from one range interval to another. For example, when the change per unit time rises from the first range interval to the second range interval, the update cycle is adjusted from longer to medium.

[0078] Step S1393: For mutation-type changes, establish an instant update triggering process to monitor the status changes of core ecological elements in real time. When a mutation-type change is detected in a core ecological element, initiate a structural update process and collect the detection frequency data, mutation judgment threshold parameters, and update initiation process data of the triggering process.

[0079] In this embodiment, an immediate update triggering process is established for abrupt changes. This update rule process uses real-time monitoring to detect changes in the state of core ecological elements. The detection frequency is a preset fixed interval, meaning that the characteristic parameters of the core ecological elements are collected and analyzed once every fixed interval. When the unit time change of the core ecological element is detected to be higher than or equal to the preset mutation judgment threshold parameter, an abrupt change is determined, and the structural update process of the element's associated structure is immediately initiated. This structural update process includes steps such as extracting key data of the abrupt change, determining the update scope, executing the update operation, and verifying the update results. Data on the detection frequency of the triggering process is collected, i.e., the fixed interval value of real-time monitoring; the mutation judgment threshold parameter is collected, i.e., the threshold value of the unit time change; and data on the update initiation process is collected, i.e., the specific steps and operating procedures of the structural update process.

[0080] Step S1394: Analyze the response time of ecological derivative elements after changes in ecological core elements based on historical data, obtain response delay duration data of ecological derivative elements, determine the update timing of ecological derivative elements based on response delay duration, and collect update time point data corresponding to different response delay durations.

[0081] In this embodiment, the time nodes when core ecological elements change and the corresponding time nodes when derived ecological elements change are extracted from the ecological element dataset. The time difference between the two is calculated, and this time difference is the response delay duration of the derived ecological elements. Statistical analysis is performed on all response delay duration data to obtain the average response delay duration and the distribution range of the response delay duration for derived ecological elements corresponding to different types of core ecological element changes. The update timing of derived ecological elements is determined based on the response delay duration. Specifically, after a core ecological element changes, the connection relationships related to that derived ecological element in the element association structure are updated after the corresponding response delay duration. Update time point data corresponding to different response delay durations are collected, i.e., the time node value obtained by adding the response delay duration to the time node when the core ecological element changes.

[0082] Step S1395: Based on the change transmission efficiency data of ecological core elements and ecological derivative elements, when the change in the change transmission efficiency exceeds a preset threshold, adjust the correlation strength between the two according to preset rules.

[0083] In this embodiment, the change transmission efficiency data of ecological core elements and ecological derivative elements is extracted from the element transmission chain. Specifically, this transmission efficiency data is the ratio of the change magnitude of ecological derivative elements to the change magnitude of ecological core elements, reflecting the transmission effect of element changes. The change in transmission efficiency is calculated as the difference between the current transmission efficiency and the historical average transmission efficiency. When this change exceeds a preset threshold, the correlation strength between the two is adjusted according to preset rules. These rules include increasing the weight of the correlation strength when transmission efficiency increases and decreasing the weight of the correlation strength when transmission efficiency decreases. The adjustment range of the weight is set according to the magnitude of the change; the larger the change, the larger the adjustment range. The adjusted correlation strength weight is then updated in the corresponding connection relationships of the element association structure to ensure the dynamic adaptability of the model.

[0084] Step S1396: Analyze the practical value of historical element status data for subsequent correlation analysis, retain historical element status data that is valuable for subsequent correlation analysis, eliminate redundant data, construct element data retention rules when updating the structure, and collect data on the retention period, conditions, and storage methods.

[0085] In this embodiment, information gain analysis is used to analyze historical element state data, calculating the information gain value of each historical element state data for subsequent correlation analysis. This information gain value reflects the degree of contribution of the data to the correlation analysis results. When the information gain value is higher than a preset gain threshold, the historical element state data is determined to have practical value for subsequent correlation analysis and is retained; when the information gain value is lower than or equal to the preset gain threshold, it is determined to be redundant data and is discarded. Based on the analysis results, element data retention rules are constructed for structural updates. The rules include data retention period data, i.e., retaining historical element state data within the most recent preset period; data retention condition data, i.e., retaining data with an information gain value higher than the gain threshold; and data retention storage method data, i.e., storing the retained data in a distributed database to ensure data security and accessibility.

[0086] Step S1397: After completing the update of the element association structure, execute the preset consistency verification process. The consistency verification process verifies the updated element association relationship according to the preset rule matching conditions, parameter value range and data format specifications, and handles the problems identified in the verification according to the preset process. The steps, standards and exception handling methods of the consistency verification process have been preset.

[0087] In this embodiment, after updating the element association structure, a preset consistency verification process is immediately executed. This process includes three steps: rule matching verification, parameter value verification, and data format verification. Rule matching verification checks whether the updated element associations conform to ecological mechanisms and logical rules based on preset rule matching conditions. Parameter value verification checks whether the attribute parameters of the updated connection relationships are within a reasonable range based on preset parameter value ranges. Data format verification checks whether the data format of the updated element association structure meets the requirements based on preset data format specifications. During the verification process, if a problem is identified, it is handled according to preset anomaly handling methods. For example, if a parameter value is found to be out of range, the parameter value is automatically adjusted to a reasonable value within the range; if a data format error is found, the format is automatically converted; if a rule matching error is found, a manual review process is triggered. The verification results and anomaly handling records are stored for subsequent traceability and analysis.

[0088] Step S1398: Record the time of each update, the identifier of the adjusted element, the name and value of the changed parameter, the reason for the update, and the operation performed, according to the preset log recording specifications.

[0089] In this embodiment, each update operation of the element association structure is fully recorded according to the preset log recording specifications. The record content includes the specific timestamp of the update, the unique identifier of the adjusted core ecological element or ecological derivative element, the name of the changed parameter (such as the association strength weight value, trigger condition threshold, etc.) and its value before and after the adjustment, the reason for the update (such as gradual change trigger, abrupt change trigger, change in transmission efficiency trigger, etc.), and the specific operation performed (such as adding a connection relationship, deleting a connection relationship, adjusting parameter values, etc.). The log records are stored in a standardized text format, and each log record contains multiple fields, each field corresponding to one record content. The log records are arranged in chronological order to facilitate subsequent querying and auditing.

[0090] Step S1399: During the update of the feature association structure, when data loss, data format error or parameter value conflict is detected, a preset exception handling process is triggered, and backup data supplementation or operation rollback is performed according to the detected status type.

[0091] In this embodiment, during the update process of the feature association structure, the integrity, format correctness, and parameter consistency of the data are monitored in real time. When data loss is detected, a backup data supplementation process is triggered, specifically by searching for corresponding data in a pre-stored backup dataset. If the corresponding data is also not found in the backup dataset, a manual data entry process is triggered. When a data format error is detected, a format conversion process is triggered, automatically converting the incorrectly formatted data to a preset standard format. When a parameter value conflict is detected, an operation rollback process is triggered, restoring the feature association structure to its state before the update, recording the conflicting parameter information, and triggering a manual review process. The process and results of anomaly handling are recorded to facilitate subsequent problem investigation and optimization.

[0092] Step S13910: Based on the update cycle, triggering conditions, update timing, intensity adjustment rules, data retention strategy, consistency verification process, log recording specifications, and exception handling process, define a set of dynamic update rules for the element association structure.

[0093] In this embodiment, the update cycle, triggering conditions, update timing, intensity adjustment rules, data retention strategy, consistency verification process, log recording specifications, and exception handling process are integrated to define a set of dynamic update rules for the element association structure. This set includes multiple rule modules, each corresponding to a rule item. Each rule item contains attribute information such as the rule's triggering conditions, execution operation, execution timing, and exception handling method.

[0094] Step S1310: The element association structure containing the action conditions and update rules is encapsulated into a model and converted into a computable and runnable model architecture to generate an ecological element dynamic association model that describes the dynamic interaction relationship between elements.

[0095] In this embodiment, a graph neural network model architecture is used to encapsulate the element association structure containing the action conditions and update rules. This graph neural network model consists of four parts: an input layer, a graph convolutional layer, a fully connected layer, and an output layer. The input layer receives input data from the video stream scene interaction sequence and converts the input data into feature vectors of graph nodes. The graph convolutional layer processes the graph data of the element association structure, captures the relationships between nodes through graph convolution operations, and dynamically adjusts the graph structure and node features according to the action conditions and update rules. The fully connected layer integrates the feature vectors output by the graph convolutional layer and outputs intermediate results containing changes in the state of ecological elements. The output layer converts the intermediate results into interpretable structured data and outputs information on the morphological change trajectory of the grassland human settlement ecological space. The model training process employs supervised learning. Training data comes from an ecological element dataset and video stream scene interaction sequences. The input is scene element interaction information, and the output is state change data of core ecological elements and derived ecological elements. The optimizer uses the Adam optimizer with a preset initial learning rate, a preset batch size, and a preset number of training epochs. The mean absolute error (MAE) is used as the loss function for model training, and the mean absolute percentage error (MASE) is used as the basis for early stopping. In the model application phase, the video stream scene interaction sequences are converted into the required input format and fed into the input layer. After processing through graph convolutional layers and fully connected layers, the output layer obtains the state change data of ecological elements, thereby generating the morphological change trajectory of the grassland human settlement ecological space.

[0096] Step S140: Input the video stream scene interaction sequence into the ecological element dynamic association model, and output the morphological change trajectory of the grassland human settlement ecological space through the operation of the ecological element dynamic association model to generate the preliminary ecological space identification result.

[0097] Step S141: Parse the timeline information in the video stream scene interaction sequence, extract the start time, end time, and time interval of the video stream scene interaction sequence, determine the time span and time node distribution contained in the video stream scene interaction sequence, and establish a timeline index structure.

[0098] In this embodiment, timeline information is extracted from the metadata of the video stream scene interaction sequence. The metadata includes information such as the start time, end time, and frame rate of the video stream unit. The time interval is calculated, specifically the acquisition time difference between two adjacent time nodes, which is the reciprocal of the frame rate. The time span of the video stream scene interaction sequence is determined, specifically the time difference between the end time and the start time. The distribution of time nodes is determined, specifically the number of time nodes evenly distributed according to time intervals from the start time to the end time, and the specific acquisition time of each time node. Based on the specific acquisition time of each time node, a timeline index structure is established. This timeline index structure adopts the form of a hash table, where the key of the hash table is the acquisition time of the time node, and the value is the index position of the video stream scene interaction sequence corresponding to that time node, facilitating subsequent rapid retrieval and location.

[0099] Step S142: Extract the scene element interaction information corresponding to each time node from the video stream scene interaction sequence in chronological order. Based on the time axis index structure, obtain the interaction form, participating elements, and scope of influence for each time node one by one, and generate a scene element interaction information sequence sorted by time.

[0100] In this embodiment, based on the timeline index structure, the index position of the video stream scene interaction sequence corresponding to each time node is retrieved sequentially according to the acquisition time of the time nodes. The corresponding scene element interaction information is extracted from this index position, specifically including the interaction form, the identifiers of the participating scene components, and the scope of interaction impact. The extracted information is arranged in time node order to generate a time-sorted sequence of scene element interaction information. Each element in this sequence corresponds to the scene element interaction information of a time node, and the element's attributes include the time node acquisition time, interaction form type, a list of participating scene component identifiers, and a numerical value for the scope of interaction impact.

[0101] Step S143: Input the scene element interaction information of each time node into the ecological element dynamic association model in sequence. Input the data step by step according to the time process so that the ecological element dynamic association model processes the scene element interaction information in the actual time sequence and simulates the real time evolution process.

[0102] In this embodiment, each element in the time-ordered sequence of scene element interaction information is sequentially input into the dynamic association model of ecological elements. The input process strictly follows the chronological order of the time nodes to ensure that the model processes the scene element interaction information in the actual time sequence. When inputting information at each time node, it is converted into a tensor format required by the model, and the tensor's dimensions match the dimensions of the model's input layer. When processing information at each time node, the model updates the state of the element association structure based on the processing results of previous time nodes and the information at the current time node, simulating the real temporal evolution process of the grassland ecosystem and ensuring that the model's output conforms to the dynamic change law of the ecosystem.

[0103] Step S144: Call the corresponding element association rules through the dynamic association model of ecological elements. Based on the input scene element interaction information, match the preset association rules in the dynamic association model of ecological elements, simulate the change of element feature parameters and distribution range adjustment of ecological core elements under the interaction of scene elements, and generate ecological core element status change data.

[0104] In this embodiment, after receiving scene element interaction information at each time node, the dynamic association model of ecological elements first calls the preset element association rules to match the input scene element interaction information with the scene element interaction actions in the element association structure to find the corresponding ecological core element change type. Based on the matched change type and triggering conditions, it simulates changes in the feature parameters of the ecological core elements. Specifically, according to the preset parameter change rules, it calculates the amount of change and the changed value of the feature parameters of the ecological core elements under the current scene element interaction. It also simulates distribution range adjustments, specifically adjusting the distribution boundary coordinates and coverage area of ​​the ecological core elements based on the interaction influence range and changes in the feature parameters. The simulated feature parameter values ​​and distribution range information are organized by time node to generate ecological core element state change data, which includes information such as the time node collection time, the unique identifier of the ecological core element, the feature parameter value, the distribution boundary coordinates, and the coverage area.

[0105] Step S145: Based on the status change data of the core ecological elements, the element transmission link is invoked through the dynamic correlation model of ecological elements to deduce the corresponding changes of the derived ecological elements. According to the correlation between the core ecological elements and the derived ecological elements, the characteristic parameter change values ​​of the derived ecological elements are calculated to generate the status change data of the derived ecological elements.

[0106] In this embodiment, the dynamic correlation model of ecological elements, based on the state change data of core ecological elements, invokes the connection relationships in the element transmission chain and infers the corresponding changes of ecological derivative elements according to the correlation between core ecological elements and derived ecological elements. Specifically, it determines the time node when the ecological derivative elements change based on the transmission time difference information in the element transmission chain; it calculates the change value of the characteristic parameters of the ecological derivative elements based on the transmission efficiency information, where the change value is the product of the change in the characteristic parameters of the core ecological elements and the transmission efficiency. The inferred characteristic parameter change values ​​and distribution range information of the ecological derivative elements are organized according to time nodes to generate state change data of ecological derivative elements, which includes information such as the time node collection time, the unique identifier of the ecological derivative element, the change value of the characteristic parameters, the distribution boundary coordinates, and the coverage area.

[0107] Step S146: Record the changes in the core ecological elements and derived ecological elements at each time point, integrate the state change data, and construct a time-series dataset of element states in chronological order.

[0108] In this embodiment, the status change data of core ecological elements and derived ecological elements are integrated according to the collection time of each time node. The integrated data corresponding to each time node contains the status change information of all core ecological elements and derived ecological elements at that time node, specifically including the element's unique identifier, feature parameter value, feature parameter change value, distribution boundary coordinates, and coverage area. The integrated data is arranged in chronological order to construct a time-series dataset of element status, which is stored in the form of a multi-dimensional array. The first dimension of the array is the time node index, the second dimension is the element's unique identifier, and the third dimension is the element's status change parameter value.

[0109] Step S147: Analyze the time series dataset of element status, mine the correlation between changes in ecological elements and grassland spatial morphology based on preset association rules, and infer two specific changes in grassland spatial morphology: adjustment of ecological area range and migration of functional zoning boundaries, based on the changes in ecological element status data, and generate a spatial morphology change analysis report.

[0110] Step S1471: Extract the ecological core element status data for each time node from the element status time series dataset, filter out three key data categories: grassland vegetation type distribution, soil water retention capacity distribution, and protozoan distribution range, classify and organize them according to time node and element type, and generate an ecological core element status dataset.

[0111] In this embodiment, ecological core element status data for each time point is extracted from the element status time-series dataset. Three key data categories are selected: grassland vegetation type distribution, soil water retention capacity distribution, and protozoan distribution range. These data include information such as element distribution boundary coordinates, coverage area, and characteristic parameter values. The selected data is then categorized and organized according to the time point collection time and element type. The time point collection time is used as the first classification dimension, and the element type as the second classification dimension. The categorized data is stored in a multi-dimensional array. The first dimension of the array is the time point index, the second dimension is the element type identifier, and the third dimension is the element status parameter value, generating an ecological core element status dataset for subsequent distribution change analysis.

[0112] Step S1472: Compare the ecological core element status datasets at adjacent time points, analyze the expansion or contraction of the distribution range of ecological core elements element by element and region, calculate the area and proportion of change in distribution range, mark the specific location of the change, and generate a record of the distribution change of ecological core elements.

[0113] In this embodiment, the ecological core element status data of two adjacent time nodes in the element status time series dataset are compared element-by-element and region-by-region. Element-by-element comparison involves comparing three types of data: grassland vegetation type distribution, soil water retention capacity distribution, and protozoan distribution range. Region-by-region comparison involves comparing the distribution of each element in different geographical regions. The changed area of ​​distribution range is calculated as the difference between the distribution coverage area at the current time node and the distribution coverage area at the previous time node; the change ratio is calculated as the ratio of the changed area to the distribution coverage area at the previous time node. The specific location of the change is marked by comparing the distribution boundary coordinates of the two time nodes to determine the boundary coordinates and geographical location information of the areas where the distribution range has expanded or shrunk. The obtained changed area, change ratio, and changed location are recorded according to time node and element type to generate an ecological core element distribution change record, which is stored in tabular form, with each record corresponding to the distribution change information of one element at one time node.

[0114] Step S1473: Based on the records of changes in the distribution of core ecological elements and combined with the geographical characteristics of grassland space, determine the grassland space areas affected by changes in core ecological elements, output as candidate areas for ecological area range adjustment, collect three types of information of candidate areas: boundary coordinates, area size, and geographical location, and generate a list of candidate areas.

[0115] In this embodiment, based on the distribution change records of core ecological elements, elements whose distribution range changes exceed a preset threshold and their corresponding change area location information are extracted. Combined with the geographical characteristics of grassland space (such as topography, landforms, and hydrology), grassland spatial areas affected by changes in core ecological elements are determined. These areas are candidate areas for ecological area range adjustment. Boundary coordinate information for each candidate area is collected, specifically the minimum and maximum x and y coordinates; area size information is collected, specifically the area coverage value; and geographical location information is collected, specifically the grassland geographical division name and latitude and longitude range corresponding to the area. The above information is organized according to the unique identifier of the candidate area to generate a candidate area list. Each candidate area contains attribute information such as a unique identifier, boundary coordinates, area size, and geographical location.

[0116] Step S1474: Extract the state data of ecological derivative elements at each time point, obtain three key data types: vegetation growth rate distribution, soil moisture change distribution, and animal activity range distribution, and organize them in a structured manner according to time point and element type to generate an ecological derivative element state dataset.

[0117] In this embodiment, ecologically derived element state data for each time node is extracted from the element state time-series dataset. Three key data categories are obtained: vegetation growth rate distribution, soil moisture change distribution, and animal activity range distribution. This data includes information such as the distribution boundary coordinates of the elements and the distribution range of characteristic parameter values. The acquired data is structured according to the collection time of the time node and the element type. The collection time of the time node is used as the first dimension, and the element type is used as the second dimension. The structured data is stored in a multi-dimensional array. The first dimension of the array is the time node index, the second dimension is the element type identifier, and the third dimension is the element's state parameter value, generating an ecologically derived element state dataset for subsequent boundary change analysis.

[0118] Step S1475: Compare the state datasets of ecological derivative elements at adjacent time points, identify the direction and distance of movement of the distribution boundary of ecological derivative elements, calculate the rate of boundary movement and cumulative movement distance, mark the spatial areas involved in the boundary movement, and generate a record of boundary changes of ecological derivative elements.

[0119] In this embodiment, the state data of ecological derivative elements at two adjacent time points in the element state time series dataset are compared to identify the movement direction of the distribution boundary of the ecological derivative elements. Specifically, by comparing the distribution boundary coordinates of the two time points, the horizontal and vertical movement directions of the boundary are determined; the movement distance is calculated, specifically the Euclidean distance between the distribution boundary coordinates of the current time point and the distribution boundary coordinates of the previous time point; the boundary movement rate is calculated, specifically the ratio of the movement distance to the time difference between the two time points; and the cumulative movement distance is calculated, specifically the sum of the boundary movement distances from the starting time point to the current time point. The spatial areas involved in the boundary movement are marked, specifically by comparing the distribution coverage areas of the two time points to determine the boundary coordinates and geographical location information of the areas traversed by the boundary movement. The identified movement direction, movement distance, movement rate, cumulative movement distance, and involved spatial areas are recorded according to time points and element types to generate an ecological derivative element boundary change record, which is stored in tabular form, with each record corresponding to the boundary change information of one element at one time point.

[0120] Step S1476: Based on the boundary change records of ecological derivative elements and combined with the delineation principles of grassland functional zoning, determine the migration direction and migration distance of grassland functional zoning boundaries. The functional zoning boundaries are delineated based on the distribution characteristics of ecological derivative elements. Collect the coordinate positions of the migrated boundaries and the difference data between the original boundaries to generate functional zoning boundary migration records.

[0121] In this embodiment, based on the boundary change records of ecologically derived elements, elements whose distribution boundary changes exceed a preset threshold and their corresponding boundary movement information are extracted. Combined with the grassland functional zoning principles, which are based on the distribution characteristics of ecologically derived elements (e.g., areas with high vegetation growth rates are classified as vegetation restoration zones, and areas with stable soil moisture changes are classified as water conservation zones), the migration direction and distance of the grassland functional zoning boundaries are determined. The migration direction is consistent with the movement direction of the ecologically derived element distribution boundaries, and the migration distance is a preset proportion of the movement distance of the ecologically derived element distribution boundaries. The coordinate position information of the migrated boundaries is collected, specifically the minimum and maximum abscissa and ordinate values ​​of the migrated functional zoning boundaries. Difference data between the migrated and original boundaries is collected, specifically the overlapping area and non-overlapping area of ​​the migrated and original boundaries. This information is organized according to the unique identifier and time node of the functional zoning to generate functional zoning boundary migration records. Each functional zoning migration record includes attribute information such as a unique identifier, time node collection time, migration direction, migration distance, post-migration boundary coordinates, and difference data from the original boundary.

[0122] Step S1477: Mark the specific locations of the candidate areas for ecological area range adjustment and the migration of grassland functional zoning boundaries, and generate a spatial change marker map.

[0123] In this embodiment, a Geographic Information System (GIS) tool is used to mark the candidate areas for ecological region range adjustment and the specific locations of grassland functional zone boundary migrations on a basic geographic base map of the grassland space. The base map contains geographical feature information such as grassland topography, geomorphology, and hydrology. When marking candidate areas, the boundary outlines of the candidate areas are drawn on the base map using preset colors and fill styles. When marking functional zone boundary migrations, the original boundaries and the migrated boundaries are drawn on the base map using preset line styles and colors, and the migration direction and migration distance are marked. After marking is completed, a spatial change marker map is generated and stored in vector format for easy subsequent coordinate transformation and overlay operations.

[0124] Step S1478: Overlay the spatial change marker map with the corresponding video frame scene composition information, convert the coordinate system of the spatial change marker map into a pixel coordinate system consistent with the video frame, and after confirming the authenticity of the spatial change through visual comparison, extract the outline shape, boundary coordinates, and area size information of the overlaid spatial change area to determine the specific range of ecological area adjustment and the specific path of functional zone boundary migration, and generate a detailed record of spatial change.

[0125] For example, step S14781: extract the boundary coordinates and change type identifiers of the change areas in the spatial change marker map, obtain the set of boundary coordinate points and boundary shape parameters of each change area, the change type identifier is used to distinguish between ecological area range adjustment and functional zoning boundary migration, collect the encoding rule data and meaning data of the change type identifier, and generate the marker map extraction result.

[0126] In this embodiment, the boundary coordinates of the changed areas are extracted from the vector data of the spatial change marker map. Specifically, this involves the set of coordinates of all discrete points on the outline edge of each changed area, stored in the form of latitude and longitude in a geographic coordinate system. Simultaneously, change type identifiers are extracted. These identifiers employ preset coding rules, such as numeric or alphabetic coding, with different codes corresponding to different change types. Coding rule data and the meaning data corresponding to each code are collected; for example, one code represents an adjustment of ecological area boundaries, while another represents a migration of functional zone boundaries. The boundary shape parameters of each changed area are calculated, including the perimeter, area, and shape index. The shape index is calculated as the ratio of perimeter to area, reflecting the complexity of the boundary. The set of boundary coordinate points, boundary shape parameters, change type identifiers, and coding meaning data of the changed areas are organized according to the unique identifier of the changed area to generate the marker map extraction results.

[0127] Step S14782: Extract the video frame corresponding to the spatial change marker map from the video stream scene interaction sequence. Based on the time node correlation, obtain the complete scene composition information and pixel coordinate system of the video frame, including the origin position, coordinate axis direction, and pixel resolution of the pixel coordinate system.

[0128] In this embodiment, based on the time node information of the spatial change marker map, a single-frame video corresponding to the time node is extracted from the video stream scene interaction sequence. This video frame contains complete scene composition information at the time of acquisition, specifically including the boundary coordinates, color features, and texture features of the scene components. The pixel coordinate system parameters of this video frame are obtained. The origin of the pixel coordinate system is usually the top-left pixel of the video frame, with the X-axis horizontal and the Y-axis vertical. The pixel resolution is the number of pixels per unit length. This parameter is used for subsequent conversion from geographic coordinates to pixel coordinates. The scene composition information and pixel coordinate system parameters of the video frame are associated and stored to generate a video frame scene composition result set, which includes the unique identifier of the video frame, the time node acquisition time, scene component information, pixel coordinate system parameters, etc.

[0129] Step S14783: Based on the parameter mapping relationship between the two coordinate systems, perform point-by-point transformation on the boundary coordinates of the change area, convert the coordinate system of the spatial change marker map into a pixel coordinate system consistent with the video frame, and overlay the transformed spatial change marker map onto the corresponding video frame according to the principle of pixel coordinate alignment, align the boundary of the change area of ​​the spatial change marker map with the actual scene position in the video frame, and adjust the transparency and color of the spatial change marker map.

[0130] In this embodiment, a parameter mapping relationship between the geographic coordinate system of the spatial change marker map and the pixel coordinate system of the video frame is established. This mapping relationship is pre-calculated using parameters such as the installation location of the monitoring equipment, shooting angle, and lens focal length. A conversion formula from latitude and longitude coordinates to pixel coordinates is established. Each coordinate point in the set of boundary coordinate points of the change area is converted point by point, transforming the latitude and longitude coordinates into the corresponding pixel coordinates, ensuring that the coordinate accuracy meets the requirements during the conversion process. Following the principle of pixel coordinate alignment, the converted spatial change marker map is superimposed onto the corresponding video frame. Specifically, the boundary pixels of each change area in the spatial change marker map are matched and aligned with the corresponding pixels in the video frame. The transparency of the spatial change marker map is adjusted so that it does not obscure the scene composition information of the video frame after being superimposed, while clearly displaying the change area. The fill color and boundary line color of the change area are adjusted to clearly distinguish the change areas of ecological area range adjustment and functional zone boundary migration.

[0131] Step S14784: Using image analysis tools, observe the superimposed video frames, compare the changes in the spatial change marker map with the actual scene changes in the video frames region by region, check the consistency of boundary positions, range size, and morphological features, and generate a visual comparison record.

[0132] In this embodiment, a professional image analysis tool is used to open the superimposed video frames. The changed regions in the spatial change marker map are compared region by region with the actual scene changes in the video frames. The comparison is performed region by region based on the unique identifier of each changed region. During the comparison, the degree of consistency between the boundary positions of the changed regions and the boundary positions of the actual scene changes in the video frames is checked, specifically by calculating the overlap ratio of boundary pixels; the degree of consistency between the size of the changed regions and the size of the actual scene changes in the video frames is checked, specifically by calculating the difference ratio of region areas; and the degree of consistency between the morphological features of the changed regions and the morphological features of the actual scene changes in the video frames is checked, specifically by calculating the difference ratio of shape indices. The comparison results for each changed region are recorded to generate a visual comparison record. This record includes the unique identifier of the changed region, the overlap ratio of its boundary positions, the difference ratio of its size, the difference ratio of its morphological features, and the comparison conclusion.

[0133] Step S14785: When the boundary of the changed area coincides with the boundary of the actual scene in the video frame, the size of the area is consistent, and the morphological characteristics are consistent, record the spatial change as a real change, and store the confirmation result and related supporting data.

[0134] In this embodiment, based on the results of visual comparison recording, when the overlap ratio of the boundary positions of a certain changed area is higher than a preset overlap threshold, the difference ratio of the area size is lower than a preset difference threshold, and the difference ratio of the morphological features is lower than a preset difference threshold, the spatial change of the changed area is determined to be a real change. The confirmation result of this spatial change is recorded, including the unique identifier of the changed area, the change type identifier, and the actual state identifier; relevant supporting data is stored, including the corresponding visual comparison records, superimposed video frame screenshots, coordinate transformation parameters, etc.

[0135] Step S14786: When the boundary of the changed area does not coincide with the boundary of the actual scene in the video frame, or the size of the area differs or the morphological features do not match, record the range, degree of difference, and specific location of the non-overlapping area to generate a difference record.

[0136] In this embodiment, based on the results of visual comparison recording, when the overlap ratio of the boundary positions of a certain changed area is lower than or equal to a preset overlap threshold, the difference ratio of the area size is higher than or equal to a preset difference threshold, or the difference ratio of the morphological features is higher than or equal to a preset difference threshold, it is determined that the spatial change of the changed area does not match the actual scene. The range of the non-overlapping area is recorded, specifically the set of boundary coordinate points of the part of the changed area that does not overlap with the actual scene change; the degree of difference is recorded, specifically the specific values ​​of the overlap ratio of the boundary positions, the difference ratio of the area size, and the difference ratio of the morphological features; the specific location is recorded, specifically the video frame pixel coordinates and geographic coordinate information corresponding to the non-overlapping area. The above information is organized according to the unique identifier of the changed area to generate a difference record.

[0137] Step S14787: When there are differences in the boundary position, range size, or morphological features, the difference is processed according to the preset difference analysis rules: if the difference features meet the preset coordinate transformation error mode, the coordinate transformation parameters are adjusted according to the preset strategy, the boundary coordinates are corrected, and the superposition verification is performed again; if the difference features meet the preset element state data analysis error mode, the relevant data is reprocessed, the spatial change marker map is generated again, and the superposition verification is performed.

[0138] In this embodiment, when differences in boundary location, size, or morphological features are detected, processing is performed according to preset difference analysis rules. These rules include processing strategies corresponding to various difference features. If the difference features conform to a preset coordinate transformation error pattern, characterized by a systematic distribution of boundary position deviations within the allowable range of coordinate transformation errors, then the coordinate transformation parameters are adjusted according to the preset strategy. For example, the lens focal length and installation angle parameters of the monitoring equipment are adjusted, and the boundary coordinates of the changed area are re-transformed. After correcting the boundary coordinates, the overlay verification is performed again until the degree of difference meets the requirements. If the difference features conform to a preset element state data analysis error pattern, characterized by a mismatch between the size deviation and the trend of element state changes, and an inconsistency between the morphological feature deviation and the element distribution characteristics, then the process returns to the element state time series dataset analysis step. The relevant data is reprocessed, such as recalculating the distribution changes of core ecological elements and redetermining the migration direction of functional zone boundaries. A spatial change marker map is then generated again and overlay verification is performed until the degree of difference meets the requirements.

[0139] Step S14788: If the analysis determines that the discrepancy is caused by an error in the analysis of the feature state time series dataset, return to the feature state time series dataset analysis step, reprocess the relevant data, correct the analysis results, and then regenerate the spatial change marker map for overlay verification.

[0140] In this embodiment, if the discrepancy is determined to be caused by an error in the analysis of the element state time series dataset, such as an error in calculating the distribution changes of core ecological elements or an error in identifying the boundary changes of derived ecological elements, the process returns to the element state time series dataset analysis step. The element state time series data is then reprocessed to correct the analysis results. For example, the changed area and proportion of core ecological elements are recalculated, and the boundary movement direction and distance of derived ecological elements are re-identified. Based on the corrected analysis results, a spatial change marker map is generated again. The process of coordinate transformation, overlay, and comparison is repeated until the degree of discrepancy meets the requirements, ensuring the accuracy of the spatial change marker map.

[0141] Step S14789: Record the verified real spatial change information, integrate the boundary coordinates, range size, change type and verification basis of the real spatial changes, and generate a verified grassland spatial morphology change record.

[0142] In this embodiment, verified real spatial change information is organized, integrating boundary coordinates (pixel coordinates and geographic coordinates), size (area value), change type (ecological area adjustment or functional zone boundary migration), and verification basis (visual comparison records, overlaid video frame screenshots, coordinate transformation parameters, etc.). This information is then organized according to the unique identifier of the changed area and time node to generate verified grassland spatial morphology change records. Each record includes attribute information such as the unique identifier of the changed area, the time node collection time, boundary coordinates, size, change type, and verification basis.

[0143] Step S1479: Integrate information on ecological area range adjustment, functional zoning boundary migration, and spatial change details to generate a grassland spatial morphology change analysis report.

[0144] In this embodiment, candidate area information for ecological area range adjustment, information on functional zone boundary migration, and detailed records of spatial changes are integrated. During the integration process, the data is categorized by time point and change type, with time point serving as the first classification dimension and change type as the second. The integrated content includes the specific scope of ecological area range adjustment at each time point, the specific path of functional zone boundary migration, and detailed parameters of spatial changes. This information is organized using a combination of text descriptions, charts, and data tables to generate a grassland spatial morphology change analysis report. The report is stored in PDF format for easy reference and use later.

[0145] Step S148: Convert the spatial morphological change analysis report into a morphological change trajectory of the grassland human settlement ecological space, which includes information on the direction, magnitude, and rate of morphological change.

[0146] In this embodiment, the information in the grassland spatial morphology change analysis report is converted into a morphological change trajectory of the grassland human settlement ecological space. This morphological change trajectory is represented in the form of a time series, with each time node corresponding to a spatial morphological state. The directional information of morphological change includes the direction of expansion or contraction of the ecological area and the migration direction of functional zone boundaries; the amplitude information of morphological change includes the changed area of ​​the ecological area and the migration distance of the functional zone boundaries; the rate information of morphological change includes the ratio of the changed area of ​​the ecological area to the time difference and the ratio of the migration distance of the functional zone boundaries to the time difference. The above information is organized in chronological order to generate a morphological change trajectory, which includes a list of time nodes and information such as the direction, amplitude, and rate of morphological change corresponding to each time node.

[0147] Step S149: Extract the stable form stage and dynamic change stage from the morphological change trajectory, and collect the start time node data and end time node data of the stable form stage and the dynamic change stage. The stable form stage refers to the time interval in which the spatial form remains unchanged, and the dynamic change stage refers to the time interval in which the spatial form continuously changes.

[0148] In this embodiment, a sliding window analysis method is used to analyze the morphological change trajectory. The size of the sliding window is a preset time interval length. By traversing the time series of the morphological change trajectory through the sliding window, the average value of the morphological change amplitude within the window is calculated. When the average value of the morphological change amplitude within the window is lower than a preset stability threshold, the time interval corresponding to the window is determined to be a stable morphological stage; when the average value of the morphological change amplitude within the window is higher than or equal to the preset stability threshold, it is determined to be a dynamic change stage. Start and end time node data for both the stable morphological stage and the dynamic change stage are collected. The start time node data is the start time of the window, and the end time node data is the end time of the window. The above stage information is associated and stored with the morphological change trajectory to facilitate the subsequent generation of ecological space identification results.

[0149] Step S1410: Based on the start time node data and the end time node data, integrate the spatial morphological characteristics of the stable morphological stage and the changing trends of the dynamic change stage to generate preliminary ecological space identification results that reflect the ecological space status of the grassland human settlement environment.

[0150] In this embodiment, based on the start and end time node data of the stable morphological stage and the dynamic change stage, the spatial morphological features of the stable morphological stage are extracted, specifically including the outline shape of the grassland ecological area, the division method of functional zones, and the spatial layout structure within this stage. The change trends of the dynamic change stage are also extracted, specifically including the direction of change of the ecological area's extent, the migration direction of functional zone boundaries, and the rate of change. Integrating the spatial morphological features of the stable morphological stage and the change trends of the dynamic change stage generates a preliminary ecological space identification result reflecting the ecological spatial status of the grassland human settlement environment. This result includes information on the stable morphological stage, the dynamic change stage, spatial morphological features, and change trends, facilitating subsequent model calibration and the generation of the final identification result.

[0151] Step S150: Based on the difference information between the preliminary ecological space identification result and the video stream scene interaction sequence, update the function parameters of the element association in the dynamic association model of ecological elements to obtain the calibrated dynamic association model of ecological elements. Then, input the video stream scene interaction sequence into the calibrated dynamic association model of ecological elements again to generate the final identification result of the grassland human settlement environment ecological space.

[0152] Step S151: Extract the ecological space morphological features and element association features from the preliminary ecological space identification results, and generate a feature extraction result set. The ecological space morphological features include the outline shape of the ecological region, the division method of functional zones, and the spatial layout structure. The element association features include the association strength, efficiency, and transmission delay between the core ecological elements and the derived ecological elements.

[0153] In this embodiment, a feature extraction tool is used to extract ecological space morphological features and element association features from the preliminary ecological space identification results. Ecological space morphological features include the outline shape parameters of the ecological region (such as perimeter, area, shape index, etc.), functional zoning parameters (such as the number of zones, area ratio of zones, etc.), and spatial layout structure parameters (such as the relative position and distance of each zone). Element association features include the association strength between core ecological elements and derived ecological elements (such as the correlation coefficient value of feature parameters), efficiency of action (such as the rate of change transmission), and transmission delay (such as the time difference of change transmission). The extracted feature parameters are organized to generate a feature extraction result set, which is stored in the form of a multi-dimensional array. The first dimension of the array is the feature type identifier (morphological feature or association feature), the second dimension is the identifier of the specific feature parameter, and the third dimension is the value of the feature parameter.

[0154] Step S152: Compare the ecological space morphology features with the scene composition information of the corresponding time and region in the video stream scene interaction sequence, analyze the degree of matching between the two in terms of spatial range, boundary position, and constituent element category, extract the mismatch between the spatial morphology and the actual scene presentation, and the discrepancies between the element association features and the actual interaction situation, and generate a set of difference information.

[0155] In this embodiment, the ecological spatial morphological features in the feature extraction result set are compared with the scene composition information of the corresponding time node and corresponding geographical region in the video stream scene interaction sequence. The corresponding time node is the video stream scene interaction sequence information that is consistent with the time node of the preliminary ecological space identification result, and the corresponding geographical region is the video stream scene interaction sequence information that corresponds to the ecological region and functional zone. During the comparison, the degree of matching between the two in terms of spatial range, boundary position, and constituent element category is analyzed. The degree of matching in spatial range is determined by calculating the overlap ratio between the ecological region area of ​​the identification result and the corresponding region area of ​​the video stream scene composition information; the degree of matching in boundary position is determined by calculating the Hausdorff distance between the ecological region boundary of the identification result and the corresponding region boundary of the video stream scene composition information; the degree of matching in constituent element category is determined by calculating the overlap ratio between the ecological element category of the identification result and the scene component category of the video stream scene composition information. When the degree of matching is lower than the preset matching threshold, the corresponding mismatch or inconsistency is extracted. The above information is organized to generate a set of difference information. Each set of difference information includes attribute information such as difference type, corresponding time node, corresponding geographical region, difference degree parameter, and specific difference description.

[0156] Step S153: Compare the set of difference information with the preset data validity rules and sequence parsing standards. If no data format error or sequence parsing logic error is found, the difference is determined to be caused by the function parameters of the element association in the dynamic association model of ecological elements, and the parameter association determination result is generated.

[0157] In this embodiment, the set of discrepancies is compared with preset data validity rules and sequence parsing standards. Data validity rules include data format specifications, data value ranges, and data integrity requirements; sequence parsing standards include the parsing process of video stream scene interaction sequences, feature extraction methods, and matching rules. During the comparison, it is checked whether the discrepancies are caused by data format errors, data values ​​exceeding the range, missing data, or sequence parsing logic errors. If no such errors are found, the discrepancies are determined to be caused by the function parameters of the element association in the dynamic association model of ecological elements, such as unreasonable trigger threshold settings or inaccurate transmission efficiency coefficient values. The determination result is recorded to generate a parameter association determination result, which includes the unique identifier of the discrepancy information, the reason for the determination, and the type of parameter that needs to be adjusted.

[0158] Step S154: Based on the parameter association determination results, screen the key parameters that affect the corresponding differences in the dynamic association model of ecological elements, and generate a list of key parameters. The key parameters include trigger threshold, change transmission efficiency, and association strength coefficient.

[0159] In this embodiment, based on the parameter association determination results, key parameters affecting corresponding differences in the dynamic association model of ecological elements are screened. These key parameters include the trigger threshold for scene element interaction actions to trigger changes in core ecological elements, the efficiency coefficient of change transmission from core ecological elements to derived ecological elements, and the association strength coefficient between core ecological elements and derived ecological elements. During the screening process, corresponding key parameters are determined according to the type of difference and the reason for the determination. For example, when the difference is inaccurate identification of the ecological area range, the corresponding key parameters might be the trigger threshold for changes in grassland vegetation type distribution and the efficiency coefficient of change transmission of vegetation growth rate. The screened key parameters are then organized to generate a key parameter list, which is stored in tabular form. Each record corresponds to the identifier, description, current value, and possible adjustment direction of a key parameter.

[0160] Step S155: Based on the preset range of the difference value and the preset type of difference information, execute the corresponding parameter update process, update the values ​​of the related factor parameters, and record the values ​​before and after the parameter update and the basis for the update.

[0161] For example, step S1551: Quantify the difference information, convert the degree of difference into a calculable difference value, and generate a quantified difference dataset.

[0162] In this embodiment, each difference information in the difference information set is quantified, converting the degree of difference into a calculable difference value. Specifically, the difference value is calculated based on the type of difference and the degree of matching parameters. For example, when the difference is a spatial range mismatch, the difference value is 1 minus the overlap ratio of the spatial range; when the difference is a boundary position mismatch, the difference value is the ratio of the Hausdorff distance of the boundary position to the preset maximum allowable distance; when the difference is a constituent element category mismatch, the difference value is 1 minus the overlap ratio of the constituent element categories. The quantified difference values ​​are then organized to generate a quantified difference dataset, which contains the unique identifier, difference type, and difference value of the difference information, facilitating subsequent difference level classification and parameter adjustment range determination.

[0163] Step S1552: Difference levels are divided according to the preset difference threshold range, and each difference level is associated with a preset parameter adjustment range standard.

[0164] In this embodiment, the difference values ​​in the quantified difference dataset are divided into multiple difference levels based on a preset difference threshold range. These levels include slight difference, moderate difference, and severe difference, with each level corresponding to a range of difference values. Each difference level is associated with a preset parameter adjustment range standard. The adjustment range for slight difference is a preset small percentage, for moderate difference it is a preset medium percentage, and for severe difference it is a preset large percentage. The adjustment range standard is expressed as a percentage; for example, the adjustment range for slight difference is a preset percentage range, for moderate difference it is another preset percentage range, and for severe difference it is a third preset percentage range. The correspondence between difference levels and adjustment range standards is recorded to generate a difference level-adjustment range mapping table.

[0165] Step S1553: Based on the links in which the differences occur, the types of elements involved, and the forms of expression, analyze the types of difference information, distinguish the differences caused by errors in the association of core ecological elements, the differences caused by errors in the transmission of ecological derivative elements, and the differences caused by errors in the interaction and matching of scene elements, and collect the judgment criteria data for each type.

[0166] In this embodiment, the types of difference information are analyzed based on the stages in which the differences arise, the types of elements involved, and their manifestations. The stages in which differences arise include the extraction of scene element interaction information, the association of core ecological elements, and the transmission of derived ecological elements. The types of elements involved include core ecological elements, derived ecological elements, and scene components. The manifestations include spatial mismatch, boundary location mismatch, and inconsistent element association strength. Based on the analysis results, the difference information is categorized into three types: differences caused by errors in the association of core ecological elements, differences caused by errors in the transmission of derived ecological elements, and differences caused by errors in scene element interaction matching. Judgment criteria data for each type are collected. The judgment criterion for errors in the association of core ecological elements is that the difference arises in the association stage of core ecological elements, involving the association relationship between core ecological elements and scene element interaction information. The judgment criterion for errors in the transmission of derived ecological elements is that the difference arises in the transmission stage of derived ecological elements, involving the transmission relationship between core ecological elements and derived ecological elements. The judgment criterion for errors in scene element interaction matching is that the difference arises in the extraction of scene element interaction information, involving the recognition of interactive actions in scene components. The above judgment criterion data is organized to generate a list of difference type judgment criteria, which is stored in tabular form.

[0167] Step S1554: For the differences caused by the incorrect association of the core ecological elements, locate the mapping relationship parameters of the interaction information between the core ecological elements and the scene elements, adjust the trigger threshold and association priority of the mapping relationship according to the adjustment range standard associated with the difference level, and record the parameter values ​​and adjustment basis before and after the adjustment.

[0168] In this embodiment, for discrepancies caused by incorrect association of core ecological elements, the corresponding mapping relationship parameters are located based on the interaction information between the core ecological elements and scene elements corresponding to the discrepancy information. These parameters include the trigger threshold and association priority of the mapping relationship. The trigger threshold is adjusted according to the adjustment range standard associated with the discrepancy level. If the discrepancy is minor, the trigger threshold is fine-tuned according to the adjustment range for minor discrepancies; if the discrepancy is moderate, the trigger threshold is adjusted according to the adjustment range for moderate discrepancies; and if the discrepancy is severe, the trigger threshold is significantly adjusted according to the adjustment range for severe discrepancies. The association priority is adjusted, and the association priority is expressed numerically, with higher values ​​indicating higher priority. The association priority value is increased or decreased according to the severity of the discrepancy. The parameter values ​​before and after adjustment and the basis for adjustment are recorded. The adjustment basis includes the unique identifier of the discrepancy information, the discrepancy level, and the adjustment range standard.

[0169] Step S1555: For the differences caused by the transmission errors of ecological derivative elements, locate the transmission path parameters of the corresponding ecological core elements and ecological derivative elements, determine the adjustment range according to the difference level, adjust the transmission rate coefficient and transmission delay threshold, and record the parameter values ​​before and after the adjustment and the basis for the adjustment.

[0170] In this embodiment, for discrepancies caused by errors in the transmission of ecological derivative elements, the corresponding transmission path parameters are located based on the ecological core elements and ecological derivative elements corresponding to the discrepancy information. These parameters include transmission rate coefficients and transmission delay thresholds. The adjustment range is determined according to the level of discrepancy: minor discrepancies correspond to smaller adjustment ranges, moderate discrepancies to moderate adjustment ranges, and severe discrepancies to larger adjustment ranges. The transmission rate coefficient is adjusted: if the discrepancy is due to changes in ecological derivative elements lagging behind changes in ecological core elements, the transmission rate coefficient is increased; if the discrepancy is due to changes in ecological derivative elements preceding changes in ecological core elements, the transmission rate coefficient is decreased. The transmission delay threshold is adjusted: if the discrepancy has an excessively long transmission delay, the transmission delay threshold is decreased; if the discrepancy has an excessively short transmission delay, the transmission delay threshold is increased. The parameter values ​​before and after adjustment, as well as the basis for adjustment, are recorded. The basis for adjustment includes the unique identifier of the discrepancy information, the discrepancy level, and the adjustment range standard.

[0171] Step S1556: For the differences caused by scene element interaction matching errors, locate the corresponding scene element interaction information and extract parameters. Determine the adjustment range according to the difference level, adjust the two types of sensitivity parameters of distance judgment threshold and direction consistency deviation allowable value for interaction action recognition, and record the parameter values ​​before and after adjustment and the basis for adjustment.

[0172] In this embodiment, for discrepancies caused by scene element interaction matching errors, the corresponding scene element interaction information extraction parameters are located based on the scene element interaction action type corresponding to the discrepancy information. These parameters include two types of sensitivity parameters: a distance judgment threshold for interaction action recognition and an allowable value for directional consistency deviation. The adjustment range is determined according to the discrepancy level: minor discrepancies correspond to smaller adjustment ranges, moderate discrepancies to moderate adjustment ranges, and severe discrepancies to larger adjustment ranges. The distance judgment threshold is adjusted by increasing it if the discrepancy is due to excessive interaction action recognition and decreasing it if the discrepancy is due to insufficient interaction action recognition. The allowable value for directional consistency deviation is adjusted by decreasing it if the discrepancy is due to excessive accompanying movement action recognition and increasing it if the discrepancy is due to insufficient accompanying movement action recognition. The parameter values ​​before and after adjustment, along with the adjustment basis, are recorded. The adjustment basis includes the unique identifier of the discrepancy information, the discrepancy level, and the adjustment range standard.

[0173] Step S1557: Adopt the adjustment record standard to record the parameter name, value before adjustment, value after adjustment, corresponding difference information, difference level, adjustment range, and adjustment time for each parameter adjustment.

[0174] In this embodiment, a preset adjustment record specification is used to record detailed information for each parameter adjustment, including the parameter name, value before adjustment, value after adjustment, unique identifier of the corresponding difference information, difference level, adjustment range, and adjustment time. The adjustment record specification defines the format and content requirements for the records.

[0175] Step S1558: Substitute the adjusted parameters into the dynamic correlation model of ecological elements, rerun the correlation process of the dynamic correlation model of ecological elements, simulate the interaction between scene elements and ecological elements based on the adjusted parameters, generate the adjusted simulation results, and extract the key feature parameters of the simulation results.

[0176] In this embodiment, the adjusted parameter values ​​are updated to the corresponding positions in the dynamic association model of ecological elements, and the element association process of the model is rerun. Based on the adjusted parameters, the interaction between scene elements and ecological elements is simulated. The simulation process processes the input data of the video stream scene interaction sequence in chronological order to generate adjusted simulation results, which include information such as the state change data of core ecological elements and ecological derivative elements, and the trajectory of ecological spatial morphological changes. Key feature parameters of the simulation results are extracted, including the overlap ratio of ecological area range, Hausdorff distance at the boundary position, and correlation coefficient of element association strength. These parameters are used for subsequent fit evaluation.

[0177] Step S1559: Compare the adjusted simulation results with the corresponding scene element interaction information in the verification dataset, calculate the degree of fit between the two in three dimensions: spatial morphology, element association, and temporal evolution, and generate fit evaluation data.

[0178] In this embodiment, the adjusted simulation results are compared with the corresponding scene element interaction information in the validation dataset. The validation dataset is a portion of data pre-selected from the scene interaction sequence of the video stream, including scene composition information, element interaction information, etc. During the comparison, the fit between the two in three dimensions—spatial morphology, element association, and temporal evolution—is calculated. The fit in the spatial morphology dimension is determined by calculating the overlap ratio between the ecological area range of the simulation results and the corresponding area range of the validation dataset; the fit in the element association dimension is determined by calculating the correlation coefficient between the element association strength of the simulation results and the element interaction information of the validation dataset; the fit in the temporal evolution dimension is determined by calculating the dynamic time warping distance between the element state change time series of the simulation results and the element state change time series of the validation dataset. The fit in the three dimensions is weighted and summed to obtain the comprehensive fit, with the weights set according to the importance of each dimension. The fit calculation results are organized to generate fit evaluation data, which includes information such as the unique identifier of the simulation results, the unique identifier of the validation dataset, spatial morphology fit, element association fit, temporal evolution fit, and comprehensive fit.

[0179] Step S15510: If the fit is lower than the preset fit threshold, then based on the new difference information and fit evaluation data, return to the step of performing the step of locating the mapping relationship parameters of the interaction information between the corresponding ecological core elements and scene elements for the difference caused by the error in the association of the ecological core elements, until the fit between the simulation result and the interaction information of the scene elements in the verification dataset reaches or exceeds the preset fit threshold.

[0180] In this embodiment, the overall fit in the fit evaluation data is compared with a preset fit threshold. If the overall fit is lower than the preset fit threshold, then based on the new difference information (i.e., the difference between the adjusted simulation results and the validation dataset) and the fit evaluation data, the steps of reverting to the steps of addressing the differences caused by the association error of the core ecological elements, locating the mapping relationship parameters of the interaction information between the corresponding core ecological elements and scene elements are performed. The process of parameter adjustment, simulation operation, and fit evaluation is repeated until the overall fit reaches or exceeds the preset fit threshold, ensuring the accuracy and reliability of the model.

[0181] Step S156: Update the adjusted factor association parameters to the ecological factor dynamic association model, replace the original parameters in the ecological factor dynamic association model, and generate the updated ecological factor dynamic association model.

[0182] In this embodiment, the updated parameter values ​​in the parameter update record are updated one by one to the corresponding positions in the dynamic association model of ecological elements, replacing the original parameter values ​​in the model. The update process is performed using a model parameter configuration tool to ensure the accuracy and completeness of the parameter updates. After the update is completed, a complete test run is performed on the model, inputting a small amount of video stream scene interaction sequence data to check whether the model's output results are normal. If the output results meet expectations, the updated dynamic association model of ecological elements is generated and stored in an executable file format for easy subsequent calling and running.

[0183] Step S157: Extract a portion of the data from the video stream scene interaction sequence as verification data. Extract a set proportion of sequence data evenly according to time distribution and scene type to generate a verification dataset.

[0184] In this embodiment, stratified sampling is used to extract a portion of data from the video stream scene interaction sequence as validation data. The stratification dimensions include time distribution and scene type. Time distribution stratification is based on the video stream's acquisition time period, ensuring the extracted validation data covers the entire acquisition time span. Scene type stratification is based on the type of interaction actions of scene elements, ensuring the extracted validation data covers all interaction action types. The extraction ratio is a preset value, such as extracting a preset proportion of data from the entire video stream scene interaction sequence. The extracted data is then organized to generate a validation dataset, which includes a unique identifier for the validation data, the identifier of the corresponding video stream unit, the acquisition time of the time node, scene element interaction information, etc., for subsequent model validation and calibration.

[0185] Step S158: Input the verification dataset into the updated dynamic association model of ecological elements, run the updated dynamic association model of ecological elements to process the verification data, obtain the verification and identification results output by the updated dynamic association model of ecological elements, and collect all feature parameters of the verification and identification results.

[0186] In this embodiment, the validation dataset is converted into the input format required by the updated dynamic association model of ecological elements. The input format is in tensor form, and the dimensions match the dimensions of the model's input layer. The input data is fed into the model, which processes the validation data. The model processes the input data at each time point in chronological order, simulating the interaction between scene elements and ecological elements, and outputs validation and identification results. These results include information such as the state change data of core ecological elements and derived ecological elements, and the trajectory of ecological spatial morphological changes. All feature parameters of the validation and identification results are collected, including the boundary coordinates and area of ​​the ecological region, the migration direction and distance of the functional zone boundaries, and the association strength and transmission efficiency between core ecological elements and derived ecological elements. These parameters are used for subsequent fit calculations and model calibration.

[0187] Step S159: Compare the verification recognition result with the video stream scene interaction sequence data corresponding to the verification dataset, calculate the fit index between the two, and when the fit does not reach the preset standard, return to the parameter adjustment step, repeat the process of adjusting the element association effect parameter until the fit reaches or exceeds the preset standard, and obtain the calibrated ecological element dynamic association model.

[0188] In this embodiment, the verification and identification results are compared with the video stream scene interaction sequence data corresponding to the verification dataset. The comparison process includes three aspects: spatial morphology comparison, element association comparison, and temporal evolution comparison. Spatial morphology comparison compares the consistency of the ecological area range and functional zoning boundaries; element association comparison compares the consistency of the association relationship between the core ecological elements and the derived ecological elements; and temporal evolution comparison compares the consistency of the temporal sequence of ecological element state changes. A fit index is calculated between the two, which is a comprehensive fit index, a weighted sum of the fit in the three comparison aspects, with the weights set according to the importance of each aspect. The calculated fit index is compared with a preset fit standard. If the fit does not reach the preset standard, the parameter adjustment step is returned, and the process of adjusting the element association parameters is repeated until the fit reaches or exceeds the preset standard. The model obtained at this point is the calibrated ecological element dynamic association model, which has higher accuracy and reliability.

[0189] Step S1510: Input the video stream scene interaction sequence into the calibrated ecological element dynamic association model, start the calibrated ecological element dynamic association model to process all data, the calibrated ecological element dynamic association model runs according to the optimized parameters and logic, and outputs the final identification result of grassland human settlement environment ecological space.

[0190] In this embodiment, all video stream scene interaction sequences are converted into the input format required by the calibrated ecological element dynamic association model. The input format is in tensor form, and the dimensions match the dimensions of the model's input layer. The input data is fed into the model, and the model is started to process all the data. The model runs according to the optimized parameters and logic. The processing process processes the input data at each time node in chronological order, simulating the interaction between scene elements and ecological elements, and outputting the final identification result of the grassland human settlement environment ecological space. This result includes information such as the precise range of ecological areas, the accurate boundaries of functional zones, the real correlation relationships of ecological elements, and the changing trajectory of ecological space morphology.

[0191] Based on the same inventive concept, please refer to Figure 2The diagram shows a schematic block diagram of the grassland human settlement ecological space identification system 100 provided in the embodiments of this application. The grassland human settlement ecological space identification system 100 may include a communication unit 110, a machine-readable storage medium 120 and a processor 130.

[0192] In this embodiment, the machine-readable storage medium 120 can also be integrated into the processor 130 and can communicate and interact with external systems through the communication unit 110. The machine-readable storage medium 120 stores machine-executable instructions for executing the scheme of this application, and the processor 130 executes the machine-executable instructions stored in the machine-readable storage medium 120 to implement the grassland human settlement ecological space identification method provided in the aforementioned method embodiments.

[0193] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A method for identifying the ecological space of grassland human settlements, characterized in that, The method includes: Acquire a continuous video stream of the grassland, which includes uninterrupted video stream units collected from different functional zones and different time spans of the grassland. Each video stream unit fully records the dynamic changes of the scene and the interaction process of elements in the corresponding area during the collection period. Dynamic scene analysis is performed on each video stream unit in the continuous grassland video stream to identify the interactive actions and state changes of scene elements within each video stream unit, and to generate the video stream scene interaction sequence corresponding to each video stream unit. Based on the scene element interaction information in the video stream scene interaction sequence, the core ecological elements and derived elements in the grassland human settlement environment are associated to construct an ecological element dynamic association model that describes the dynamic interaction relationship between elements. The video stream scene interaction sequence is input into the ecological element dynamic association model. Through the calculation of the ecological element dynamic association model, the morphological change trajectory of the grassland human settlement ecological space is output, and the preliminary ecological space identification result is generated. Based on the difference information between the preliminary ecological space identification results and the video stream scene interaction sequence, the function parameters of the element association in the dynamic association model of ecological elements are updated to obtain the calibrated dynamic association model of ecological elements. The video stream scene interaction sequence is then input into the calibrated dynamic association model of ecological elements again to generate the final identification result of the grassland human settlement environment ecological space.

2. The grassland human settlement ecological space identification method according to claim 1, characterized in that, The process of dynamically analyzing each video stream unit in the continuous grassland video stream, capturing the interactive actions and state changes of scene elements within each video stream unit, and generating a video stream scene interaction sequence corresponding to each video stream unit includes: The independent video frames contained in each video stream unit of the grassland continuous video stream are separated frame by frame in chronological order, and the pixel information and color channel information of each independent video frame are completely preserved. The separated independent video frames are arranged in the original chronological order to form an ordered video frame sequence corresponding to each video stream unit. The scene composition information of each video frame in the ordered video frame sequence is extracted, each pixel of the video frame is traversed, and different scene components within the video frame are identified based on a preset image feature difference threshold. The pixel distribution range, color feature value, and contour edge information of each scene component are recorded to generate a structured scene composition information dataset. The scene components include three categories: artificial facility form, natural vegetation appearance, and land cover. By comparing the scene composition information datasets of two adjacent video frames, the pixel distribution range changes, color feature value fluctuations, and contour edge offsets of the scene components are compared dimension by dimension. Dynamic change signs of the scene components are captured, and the starting and ending pixel coordinates of each dynamic change sign are marked to define the spatial range of the change. The dynamic change signs include three categories: position movement, shape change, and appearance / disappearance. Mark the scene components corresponding to each dynamic change sign, associate the same scene components in the previous and next frames, record three types of quantitative information of the scene component in adjacent video frames: position offset, morphological deformation degree, and color change amplitude, and organize them into a single frame change record in a unified format. Integrate the single-frame change records of all adjacent video frames in an ordered video frame sequence, arrange them sequentially according to the time order of the video frames, and construct a continuous change trajectory data chain; Identify the interaction actions between scene components in the change trajectory data chain, and based on the positional relationship, change sequence, and morphological correlation of scene components, distinguish the interaction forms between three types of scene elements: occlusion, contact, and accompanying movement, and collect the judgment basis parameters and feature data for each interaction form. Record the identifiers of the constituent parts of each interactive action, the starting video frame number, the ending video frame number, the duration of the action, and the state change data of each participating part, and organize them in chronological order to form a scene interaction record; Align the change trajectory data chain with the scene interaction record on the timeline, integrate the aligned change trajectory data chain and scene interaction record in chronological order, organize the dynamic change information and interaction information into a scene dynamic dataset in a unified format, and perform serialization processing on the scene dynamic dataset to generate the video stream scene interaction sequence corresponding to each video stream unit.

3. The grassland human settlement ecological space identification method according to claim 2, characterized in that, The interaction actions between scene components in the identified trajectory data chain, based on the positional relationships, temporal sequence of changes, and morphological correlation of the scene components, distinguishes the interaction forms between three types of scene elements: occlusion, contact, and accompanying movement. It collects the judgment criteria parameters and feature presentation data for each interaction form, including: Extract the position coordinate information of all scene components in the change trajectory data chain, establish a unified coordinate reference system based on the pixel coordinate system of video frames, obtain the boundary coordinates and center coordinates of each scene component, and classify and store them according to the scene component identifier and time node; Track the position coordinate changes of each scene component in consecutive video frames, and generate an independent movement trajectory for each scene component based on the time series of coordinate information. This independent movement trajectory contains three types of information: position coordinates, movement direction, and movement distance at each time point. By comparing the independent movement trajectories of any two scene components, the boundary coordinate distance and center coordinate distance between the two in the same video frame are calculated, and the sequence of distance data changes over time is recorded to generate a distance change dataset. The observation position coordinate distance change trend is processed. When the distance shrinks to the preset first distance threshold and the number of frames reaches the preset first frame number threshold, the overlapping signs of the morphological features of the scene components are combined to determine that the two scene components have made contact. The contact action determination threshold parameters and the number of frames are collected. When the boundary coordinates of one scene component completely cover the boundary coordinates of another scene component, and the covered part cannot be observed by pixel feature detection in subsequent consecutive video frames, the occlusion action is determined by combining the positional relationship and movement trajectory of the two scene components, and the coverage range determination parameters and observation disappearance duration parameters of the occlusion action are collected. When the angle between the movement directions of the independent movement trajectories of the two scene components is less than a preset angle threshold, and the fluctuation amplitude of the position coordinate distance is less than a preset distance fluctuation threshold, the duration of the corresponding state is verified to reach a preset second frame number threshold, it is determined that an accompanying movement action has occurred, and the allowable deviation parameters of direction consistency and the fluctuation threshold parameters of distance stability are collected. Record the starting and ending video frame numbers for each interaction action. Based on the time nodes of the action's occurrence and termination, calculate the duration of the action, accurate to the smallest time unit of the video frame. Extract three types of parameters—contour shape, color distribution, and texture details—of the components of two scenes when they interact. Compare the values ​​of these three types of parameters before and after the interaction to generate a dataset of morphological feature changes. Analyze the changes in morphological features during the interaction process, calculate the magnitude and rate of change of morphological feature parameters, determine the degree of influence of the action on the morphology of the scene components, classify the level of influence, and collect the level standard parameters. By integrating interaction action types, time intervals, identifiers of participating scene components, degree of morphological influence, and judgment criteria parameters, a standardized format is used to construct a record of the interaction forms between scene elements.

4. The grassland human settlement ecological space identification method according to claim 1, characterized in that, The method involves associating core and derived ecological elements in the grassland human settlement environment with scene element interaction information from the video stream scene interaction sequence, and constructing a dynamic association model of ecological elements describing the dynamic interaction relationships between elements, including: Data on the core ecological elements and related elements derived from the core elements that maintain grassland ecological stability were collected. The core ecological elements include grassland vegetation type, soil water retention capacity, and distribution of protozoa. The derived ecological elements include vegetation growth rate, soil moisture change, and animal activity range. Data were collected through three methods: grassland ecological survey literature retrieval, field ecological condition observation and recording, and acquisition through ecological monitoring data sharing platform. Traverse each data node of the video stream scene interaction sequence, extract the interaction form, interaction frequency, and interaction impact range between the scene components, and generate a set of scene element interaction information; By feature matching, the correspondence between the core ecological elements and the interaction information set of scene elements is determined. The feature parameters of each core ecological element are compared with the attribute parameters of the interaction information of scene elements one by one. The scene element interaction actions with a matching degree higher than the preset threshold are judged as actions that directly affect the state of the corresponding core ecological elements, and a preliminary list of associated correspondences is generated. Based on the initial list of associated elements, the change type of the core ecological elements is labeled for each scene element interaction action. The characteristic index data and judgment standard parameters of the change type are collected, and the interaction action is mapped one by one to the change type to generate a structured mapping table. This study analyzes the relationship between ecological derivative elements and ecological core elements. By combining ecological mechanism analysis and data statistical analysis, it examines the process by which changes in ecological core elements lead to changes in ecological derivative elements. It collects data on the transmission of changes between the two and establishes a correlation between the change characteristics of ecological core elements and the change characteristics of ecological derivative elements based on the statistical analysis results, generating a correlation analysis report. Based on the correlation analysis report, we track the specific links and processes by which changes in each core ecological element lead to changes in corresponding derived ecological elements, mark key node data and influencing factor data in the path, and generate element transmission links. By integrating the mapping table and the element transmission link, an element association structure is formed, which organizes the connection relationships between core ecological elements, derived ecological elements, and interactive information of scene elements in a structured manner. Based on the basic attributes of ecological elements and actual data of scene interactions, we collect scene condition data and element status data of each connection relationship of the triggering element association structure, and record specific trigger condition description data and judgment standard parameters. Based on the changing trends of ecological elements and the dynamic adjustment needs of quantitative indicators and relationships, we determine the update timing data, update scope data, and update method data for adaptive adjustment of the element relationship structure after the element status changes, and generate a set of update rules. The element association structure, which includes the conditions for action and the update rules, is modeled and encapsulated, and transformed into a computable and runnable model architecture to generate an ecological element dynamic association model that describes the dynamic interaction relationships between elements.

5. The grassland human settlement ecological space identification method according to claim 4, characterized in that, Based on the dynamic adjustment needs of ecological element change trends, quantitative indicators, and correlations, the update timing data, update range data, and update method data are determined to adaptively adjust the element correlation structure after element state changes, generating an update rule set, including: Statistical analysis is performed on historical change data of core ecological elements. Based on a preset threshold for change per unit time, state changes are divided into gradual changes and abrupt changes. State changes with a change per unit time below the threshold are judged as gradual changes, and state changes with a change per unit time above or equal to the threshold are judged as abrupt changes. For gradual changes, based on the range of changes per unit time corresponding to the determined gradual changes, a corresponding update cycle is set to determine the timing of updates for core ecological elements, and specific duration data and adjustment condition data of the update cycle are collected. In response to abrupt changes, an instant update triggering process is established to monitor the status changes of core ecological elements in real time. When abrupt changes are detected in core ecological elements, the structural update process is initiated, and the detection frequency data, mutation judgment threshold parameters, and update initiation process data of the triggering process are collected. Based on historical data analysis, the response time of ecological derivative elements after changes in ecological core elements is analyzed to obtain response delay duration data of ecological derivative elements. The update timing of ecological derivative elements is determined according to the response delay duration, and update time point data corresponding to different response delay durations are collected. Based on the change transmission efficiency data of core ecological elements and derived ecological elements, when the change in the change transmission efficiency exceeds a preset threshold, the correlation strength between the two is adjusted according to preset rules. Analyze the practical value of historical element status data for subsequent correlation analysis, retain historical element status data that is valuable for subsequent correlation analysis, eliminate redundant data, construct element data retention rules when updating the structure, and collect data on data retention period, conditions, and storage methods. After the element association structure is updated, a preset consistency verification process is executed. This consistency verification process verifies the updated element association relationship based on preset rule matching conditions, parameter value ranges, and data format specifications, and handles the problems identified in the verification according to the preset process. The steps, standards, and exception handling methods of the consistency verification process have been preset. According to the preset log recording specifications, record the time of each update, the identifier of the adjusted element, the name and value of the changed parameter, the reason for the update, and the operation performed; During the update of the feature association structure, when data loss, data format error or parameter value conflict is detected, a preset exception handling process is triggered, and backup data supplementation or operation rollback is performed according to the detected status type. Based on the update cycle, triggering conditions, update timing, intensity adjustment rules, data retention strategy, consistency verification process, log recording specifications, and exception handling process, a set of dynamic update rules for the element association structure is defined.

6. The method for identifying the ecological space of grassland human settlements according to claim 1, characterized in that, The step involves inputting the video stream scene interaction sequence into the ecological element dynamic association model, and through the calculation of the ecological element dynamic association model, outputting the morphological change trajectory of the grassland human settlement ecological space, generating preliminary ecological space identification results, including: The timeline information in the video stream scene interaction sequence is analyzed, the start time, end time, and time interval of the video stream scene interaction sequence are extracted, the time span and time node distribution of the video stream scene interaction sequence are determined, and a timeline index structure is established. Extract scene element interaction information corresponding to each time node from the video stream scene interaction sequence in chronological order. Based on the time axis index structure, obtain the interaction form, participating elements, and scope of influence of each time node one by one to generate a scene element interaction information sequence sorted by time. The scene element interaction information at each time point is sequentially input into the ecological element dynamic association model. Data is input step by step according to the time process, so that the ecological element dynamic association model processes the scene element interaction information according to the actual time sequence, simulating the real time evolution process. By calling the corresponding element association rules through the dynamic association model of ecological elements, and based on the input scene element interaction information, the system matches the preset association rules in the dynamic association model of ecological elements, simulates the changes in the feature parameters and distribution range of the core ecological elements under the interaction of scene elements, and generates data on the state changes of the core ecological elements. Based on the status change data of core ecological elements, the dynamic correlation model of ecological elements is used to call the element transmission link, and the corresponding changes of ecological derivative elements are inferred. According to the correlation between core ecological elements and ecological derivative elements, the characteristic parameter change values ​​of ecological derivative elements are calculated, and the status change data of ecological derivative elements are generated. Record the changes in the core ecological elements and derived ecological elements at each time point, integrate the state change data, and construct a time-series dataset of element states in chronological order. Analyze the time series dataset of element status, mine the correlation between changes in ecological elements and grassland spatial morphology based on preset association rules, and infer two specific changes in grassland spatial morphology: adjustment of ecological area range and migration of functional zoning boundaries based on the changes in ecological element status data, and generate a spatial morphology change analysis report. The spatial morphological change analysis report is converted into a morphological change trajectory of the grassland human settlement ecological space, which includes information on the direction, magnitude, and rate of morphological change. Extract the stable morphological phase and the dynamic change phase from the morphological change trajectory, and collect the start and end time node data of the stable morphological phase and the dynamic change phase. The stable morphological phase refers to the time interval in which the spatial morphology remains unchanged, and the dynamic change phase refers to the time interval in which the spatial morphology continuously changes. Based on the start and end time point data, the spatial morphological characteristics of the stable morphological stage and the changing trends of the dynamic change stage are integrated to generate preliminary ecological space identification results that reflect the ecological space status of the grassland human settlement environment.

7. The grassland human settlement ecological space identification method according to claim 6, characterized in that, The analyzed time-series dataset of ecological elements is used to explore the correlation between changes in ecological elements and grassland spatial morphology. It identifies two specific changes caused by changes in ecological elements: adjustments to the ecological range of grassland spatial morphology and migration of functional zoning boundaries. A spatial morphology change analysis report is generated, including: Ecological core element status data for each time point is extracted from the element status time series dataset. Three key data categories are selected: grassland vegetation type distribution, soil water retention capacity distribution, and protozoan distribution range. The data are then classified and organized according to time point and element type to generate an ecological core element status dataset. By comparing the status datasets of ecological core elements at adjacent time points, analyzing the expansion or contraction of the distribution range of ecological core elements element by element and region, calculating the area and proportion of change in the distribution range, marking the specific location of the change, and generating a record of the distribution change of ecological core elements. Based on the records of changes in the distribution of core ecological elements and combined with the geographical characteristics of grassland space, grassland spatial areas affected by changes in core ecological elements are identified, and candidate areas for ecological area range adjustment are output. Three types of information, namely boundary coordinates, area size, and geographical location of candidate areas, are collected to generate a list of candidate areas. Extract the state data of ecological derivative elements at each time point, obtain three key data: vegetation growth rate distribution, soil moisture change distribution, and animal activity range distribution, and organize them in a structured manner according to time point and element type to generate an ecological derivative element state dataset. By comparing the state datasets of ecological derivative elements at adjacent time points, the direction and distance of movement of the distribution boundary of ecological derivative elements are identified, the rate of boundary movement and cumulative distance of movement are calculated, the spatial areas involved in the boundary movement are marked, and a record of the boundary changes of ecological derivative elements is generated. Based on the boundary change records of ecological derivative elements and combined with the delineation principles of grassland functional zoning, the migration direction and migration distance of grassland functional zoning boundaries are determined. The functional zoning boundaries are delineated based on the distribution characteristics of ecological derivative elements. The coordinate positions of the migrated boundaries and the difference data between the original boundaries are collected to generate functional zoning boundary migration records. Mark the candidate areas for ecological area range adjustment and the specific locations of grassland functional zone boundary migration, and generate a spatial change marking map; The spatial change marker map is overlaid with the scene composition information of the corresponding video frame. The coordinate system of the spatial change marker map is converted into the pixel coordinate system consistent with the video frame. After confirming the authenticity of the spatial change through visual comparison, the outline shape, boundary coordinates and area size of the overlaid spatial change area are extracted to determine the specific range of ecological area adjustment and the specific path of functional zone boundary migration, and generate a detailed record of spatial change. By integrating information on ecological area adjustments, functional zoning boundary migrations, and detailed records of spatial changes, a report analyzing grassland spatial morphological changes is generated.

8. The method for identifying the ecological space of grassland human settlements according to claim 1, characterized in that, Based on the difference information between the preliminary ecological space identification results and the video stream scene interaction sequence, the function parameters of the element association in the dynamic association model of ecological elements are updated to obtain a calibrated dynamic association model of ecological elements. The video stream scene interaction sequence is then input into the calibrated dynamic association model of ecological elements again to generate the final identification result of the grassland human settlement environment ecological space, including: The ecological space morphological features and element association features are extracted from the preliminary ecological space identification results to generate a feature extraction result set. The ecological space morphological features include the outline shape of the ecological region, the division method of functional zoning, and the spatial layout structure. The element association features include the association strength, efficiency, and transmission delay between the core ecological elements and the derived ecological elements. The ecological space morphology features are compared with the scene composition information of the corresponding time and region in the video stream scene interaction sequence. The degree of matching between the two in terms of spatial range, boundary position and constituent element category is analyzed. The mismatch between the spatial morphology and the actual scene presentation, and the inconsistency between the element association features and the actual interaction situation are extracted to generate a set of difference information. The set of discrepancy information is compared with the preset data validity rules and sequence parsing standards. If no data format errors or sequence parsing logic errors are found, the discrepancy is determined to be caused by the function parameters of the element association in the dynamic association model of ecological elements, and parameter association determination results are generated. Based on the parameter association determination results, key parameters affecting the corresponding differences in the dynamic association model of ecological elements are screened, and a list of key parameters is generated. The key parameters include trigger threshold, change transmission efficiency, and association strength coefficient. Based on the preset range of the difference values ​​and the preset type of difference information, execute the corresponding parameter update process, update the values ​​of the related factor parameters, and record the values ​​before and after the parameter update and the basis for the update. The adjusted factor association parameters are updated to the dynamic association model of ecological elements, replacing the original parameters in the dynamic association model of ecological elements, and the updated dynamic association model of ecological elements is generated. Extract a portion of the data from the video stream scene interaction sequence as verification data, and extract a set proportion of sequence data evenly according to time distribution and scene type to generate a verification dataset; Input the verification dataset into the updated dynamic association model of ecological elements, run the updated dynamic association model of ecological elements to process the verification data, obtain the verification and identification results output by the updated dynamic association model of ecological elements, and collect all feature parameters of the verification and identification results. Compare the verification and identification results with the video stream scene interaction sequence data corresponding to the verification dataset, calculate the fit index between the two, and when the fit does not reach the preset standard, return to the parameter adjustment step, repeat the process of adjusting the element association parameters until the fit reaches or exceeds the preset standard, and obtain the calibrated ecological element dynamic association model. The video stream scene interaction sequence is input into the calibrated ecological element dynamic association model. The calibrated ecological element dynamic association model is started to process all data. The calibrated ecological element dynamic association model runs according to the optimized parameters and logic, and outputs the final identification result of the grassland human settlement environment ecological space.

9. A grassland human settlement ecological space identification system, characterized in that, include: processor; A machine-readable storage medium for storing machine-executable instructions of the processor; The processor is configured to execute the grassland human settlement ecological space identification method according to any one of claims 1 to 8 by executing the machine-executable instructions.

10. A computer program product, characterized in that, The computer program product includes machine-executable instructions stored in a computer-readable storage medium. A processor of a computer device reads the machine-executable instructions from the computer-readable storage medium and executes the machine-executable instructions, causing the computer device to perform the grassland human settlement ecological space identification method as described in any one of claims 1 to 8.

Images