Plateau lake basin vegetation degradation early warning method based on spatio-temporal graph convolution network

Abstract

The application provides a highland lake basin vegetation degradation early warning method based on a space-time graph convolution network, comprising the following steps: acquiring multi-source remote sensing data to construct a space-time graph convolution network model, and matching model node information with a space-time position of data; inputting historical vegetation degradation data into the model, and making the data feature dimension consistent with the model node feature; generating a space-time evolution feature map according to a prediction result output by the model and sending the space-time evolution feature map to an early warning platform; storing early warning information after verification and sending the early warning information to a basin ecological management system; if no confirmation signal is received, adding the early warning information to a reissue queue, reissuing the early warning information according to a preset interval sequence, and deleting the early warning information from the queue after successful reissue. The application can improve the timeliness and scientificity of basin ecological management.

Description

Technical Field

[0001] This invention relates to the field of ecological environment monitoring and intelligent early warning technology, and more specifically, to a method for early warning of vegetation degradation in plateau lake basins based on spatiotemporal graph convolutional networks. Background Technology

[0002] In plateau lake basins, vegetation degradation is becoming increasingly serious, threatening the ecological environment and biodiversity. Traditional methods for monitoring vegetation degradation mainly rely on ground surveys and simple remote sensing data analysis, which have many limitations. Ground surveys are time-consuming and labor-intensive, unable to cover large areas, and cannot monitor vegetation dynamics in real time. While simple remote sensing data analysis can provide vegetation information over a wide area, it lacks in-depth analysis of the spatiotemporal evolution characteristics of vegetation degradation, making it difficult to accurately predict the trend and extent of vegetation degradation. Furthermore, existing technologies are inadequate in the transmission and confirmation of early warning information regarding vegetation degradation. Failure to deliver or confirm early warning information in a timely manner may result in missing the optimal opportunity for ecological protection and restoration.

[0003] The existing technologies have at least the following problems or defects: First, existing monitoring methods cannot effectively integrate vegetation cover and environmental factor information from multi-source remote sensing data, making it difficult to construct models that can accurately reflect the spatiotemporal evolution characteristics of vegetation degradation; second, there is a lack of effective management and re-issuance mechanisms for vegetation degradation early warning information, which may result in loss or failure to be confirmed in a timely manner during the transmission of early warning information, thus compromising the reliability and timeliness of early warnings; third, existing technologies are insufficient in the visualization of vegetation degradation early warning information, failing to intuitively display the spatiotemporal evolution process of vegetation degradation, which is detrimental to the formulation and implementation of ecological protection decisions. Summary of the Invention

[0004] This invention provides a method for early warning of vegetation degradation in plateau lake basins based on spatiotemporal graph convolutional networks, comprising: Multi-source remote sensing monitoring data of plateau lake basins are acquired. Based on the vegetation coverage information and environmental factor information included in the multi-source remote sensing monitoring data, a spatiotemporal graph convolutional network model is constructed. The spatiotemporal graph node information included in the constructed spatiotemporal graph convolutional network model is the same as the spatiotemporal location information included in the multi-source remote sensing monitoring data. Historical vegetation degradation monitoring data is input into a spatiotemporal graph convolutional network model. The spatiotemporal feature information included in the historical vegetation degradation monitoring data has the same spatiotemporal graph node feature dimension as that included in the spatiotemporal graph convolutional network model. In response to the vegetation degradation prediction results output by the spatiotemporal graph convolutional network model, a spatiotemporal evolution feature map corresponding to the vegetation degradation prediction results is generated based on the vegetation degradation prediction results. Send the spatiotemporal evolution feature map to the vegetation degradation early warning platform; In response to the vegetation degradation early warning verification information fed back by the vegetation degradation early warning platform, and if the vegetation degradation early warning verification information meets the preset early warning threshold conditions, the vegetation degradation early warning verification information is stored and processed, and the verification success signal of the corresponding vegetation degradation early warning verification information is sent to the vegetation degradation early warning platform. The received vegetation degradation early warning verification information is sent to the watershed ecological management system. In response to the confirmation signal that no corresponding vegetation degradation early warning verification information has been received from the watershed ecological management system within the preset monitoring period, the vegetation degradation early warning verification information is added to the early warning re-issuance queue as early warning information to be reissued. Based on the early warning re-issuance queue and the preset re-issuance time interval sequence, the vegetation degradation early warning verification information is re-issued.

[0005] Furthermore, based on the early warning re-issuance queue and the preset re-issuance time interval sequence, the vegetation degradation early warning verification information is re-issued, including: In response to the determination that the pre-warning information to be reissued, corresponding to the vegetation degradation pre-warning verification information in the pre-warning reissue queue, meets the preset reissue conditions, the preset reissue time interval that meets the preset time order condition in the preset reissue time interval sequence is determined as the target reissue time interval. Based on the determined target retransmission interval, the following retransmission steps are performed for vegetation degradation early warning verification information: the early warning information to be retransmitted corresponding to vegetation degradation early warning verification information is sent to the watershed ecological management system to retransmit the vegetation degradation early warning verification information. The determined target retransmission time interval is deleted from the preset retransmission time interval sequence to obtain the updated preset retransmission time interval sequence as the preset retransmission time interval sequence. Determine whether a confirmation signal for the corresponding vegetation degradation early warning verification information sent by the watershed ecological management system has been received within the preset monitoring period; In response to the confirmation signal received from the watershed ecological management system within the preset monitoring period, the pre-issued early warning information will be removed from the early warning reissue queue.

[0006] Furthermore, the reissue process also includes: In response to the confirmation signal that no corresponding vegetation degradation early warning verification information was received from the watershed ecological management system within the preset monitoring period, a preset retransmission time interval that meets the preset time order condition is selected from the preset retransmission time interval sequence as the target retransmission time interval, and the retransmission step is executed again for the vegetation degradation early warning verification information.

[0007] Furthermore, based on vegetation cover information and environmental factor information included in multi-source remote sensing monitoring data, a spatiotemporal graph convolutional network model is constructed, including: Spatiotemporal alignment processing is performed on multi-source remote sensing monitoring data to obtain a standardized spatiotemporal dataset; In response to the determination that the standardized spatiotemporal dataset meets the preset data quality conditions, a spatiotemporal graph convolutional network model containing a spatiotemporal feature extraction module is constructed based on the standardized spatiotemporal dataset and the spatiotemporal graph convolutional network architecture.

[0008] Furthermore, the method also includes: In response to the determination that the standardized spatiotemporal dataset does not meet the preset data quality conditions, the abnormal sensor numbers included in the multi-source remote sensing monitoring data are determined. Preset data anomaly alarm information is sent to the monitoring device terminal corresponding to the anomaly sensor number.

[0009] Furthermore, based on the vegetation degradation prediction results, a spatiotemporal evolution feature map corresponding to the vegetation degradation prediction results is generated, including: obtaining the current monitoring time window information; Based on the obtained current monitoring time window information, generate time series coding information; The randomly generated graph convolution kernel parameter initialization information is determined as the random initialization parameter; Preset spatial topology information, time series encoding information, and random initialization parameters are determined as the elements for constructing the spatiotemporal graph. Feature fusion processing is performed on the elements constructing the spatiotemporal map to obtain a fused spatiotemporal feature vector; The fused spatiotemporal feature vectors are processed by graph convolution to obtain spatiotemporal feature map data; The spatiotemporal feature map data is visualized and rendered to obtain the rendered spatiotemporal feature map as a spatiotemporal evolution feature map.

[0010] Furthermore, the reissue process also includes: In response to the determination that no confirmation signal of the corresponding vegetation degradation early warning verification information sent by the watershed ecological management system has been received within the preset monitoring period, and the detection of communication anomaly information of the watershed ecological management system, a corresponding communication failure alarm information of the watershed ecological management system is generated based on the communication anomaly information. Based on the preset ecological early warning level standards, determine the ecological early warning level corresponding to the abnormal communication information; The communication failure alarm information and ecological early warning level are sent to the monitoring equipment terminal with the corresponding abnormal sensor number.

[0011] Furthermore, the spatiotemporal feature map data is visualized and rendered to obtain the rendered spatiotemporal feature map as a spatiotemporal evolution feature map, including: Based on the preset color coding rules for vegetation degradation levels, determine the color coding information corresponding to each spatiotemporal map node in the spatiotemporal feature map data; Spatial overlay processing is performed on color encoding information to generate an overlaid spatiotemporal feature layer; The spatiotemporal feature layer is coordinate registered with the preset plateau lake basin base map data to obtain the registered spatiotemporal feature map. Based on the preset ecological early warning level labeling template, ecological early warning level labeling information is added to the registered spatiotemporal feature map to obtain a spatiotemporal evolution feature map.

[0012] Furthermore, the multi-source remote sensing monitoring data undergoes spatiotemporal alignment processing to obtain a standardized spatiotemporal dataset, including: Extract timestamp and geographic coordinate information from multi-source remote sensing monitoring data; Based on the preset time window division rules, different timestamp information is mapped to a unified time series grid; Based on the preset geographic grid division parameters, different geographic coordinate information is converted to a unified spatial coordinate system; Missing value interpolation is performed on the mapped unified time series grid and the transformed unified spatial coordinate system to generate a standardized spatiotemporal dataset.

[0013] Furthermore, determining the target transmission interval information includes: The weighting factor of each time interval in the preset resend interval sequence is calculated based on historical resend success rate data. Remove the preset retransmission time intervals with weight factors below a preset threshold from the sequence to generate an optimized preset retransmission time interval sequence. Based on the optimized preset retransmission time interval sequence, the preset retransmission time interval that satisfies the preset time order condition is determined by arranging the time intervals in ascending order.

[0014] The embodiments of the present invention have at least the following beneficial effects: 1. A spatiotemporal graph convolutional network model can be constructed based on multi-source remote sensing monitoring data. Through spatiotemporal alignment and feature fusion processing, it can effectively capture the spatiotemporal evolution pattern of vegetation degradation in plateau lake basins, solve the problem that traditional methods cannot take into account spatiotemporal correlation, and improve the accuracy of degradation prediction.

[0015] 2. The reliable transmission of early warning information can be achieved through the early warning retransmission queue and the preset retransmission time interval sequence. The retransmission mechanism is automatically triggered when no confirmation signal is received from the watershed ecological management system, which solves the problem of possible loss or delay of early warning information and ensures the timeliness of ecological management decisions.

[0016] 3. The spatiotemporal distribution characteristics of vegetation degradation prediction results can be intuitively displayed through the visualization rendering of spatiotemporal feature maps and the labeling of ecological early warning levels. This solves the problem of insufficient visualization in traditional monitoring methods, making it easier for managers to quickly identify high-risk degradation areas and take targeted measures. Attached Figure Description

[0017] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent from the following detailed description taken in conjunction with the accompanying drawings. Several embodiments of the invention are illustrated in the drawings by way of example and not limitation, wherein: Figure 1 This is a flowchart illustrating an embodiment of the present invention of a method for early warning of vegetation degradation in plateau lake basins based on spatiotemporal graph convolutional networks. Detailed Implementation

[0018] The principles and spirit of the invention will now be described with reference to several exemplary embodiments. It should be understood that these embodiments are provided merely to enable those skilled in the art to better understand and implement the invention, and are not intended to limit the scope of the invention in any way. Rather, these embodiments are provided to make the invention more thorough and complete, and to fully convey the scope of the invention to those skilled in the art.

[0019] Those skilled in the art will recognize that embodiments of the present invention can be implemented as a system, apparatus, device, method, or computer program product. Therefore, the present invention can be specifically implemented in the following forms: entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0020] It should be noted that the number of any elements in the accompanying drawings is for illustrative purposes only and not as a limitation, and any naming is for distinction only and has no limiting meaning.

[0021] The following is for reference. Figure 1 , Figure 1 This is a flowchart illustrating a method for early warning of vegetation degradation in plateau lake basins based on spatiotemporal graph convolutional networks, provided in an embodiment of the present invention. Figure 1 As shown, a method for early warning of vegetation degradation in plateau lake basins based on spatiotemporal graph convolutional networks includes: S1. Obtain multi-source remote sensing monitoring data of plateau lake basins. Based on the vegetation coverage information and environmental factor information included in the multi-source remote sensing monitoring data, construct a spatiotemporal graph convolutional network model. The spatiotemporal graph node information included in the constructed spatiotemporal graph convolutional network model is the same as the spatiotemporal location information included in the multi-source remote sensing monitoring data. S2. Input the historical vegetation degradation monitoring data into the spatiotemporal graph convolutional network model. The spatiotemporal feature information included in the historical vegetation degradation monitoring data and the spatiotemporal graph node feature dimension included in the spatiotemporal graph convolutional network model are the same. S3. In response to the vegetation degradation prediction results output by the spatiotemporal graph convolutional network model, generate a spatiotemporal evolution feature map corresponding to the vegetation degradation prediction results. S4. Send the spatiotemporal evolution feature map to the vegetation degradation early warning platform; S5. In response to the vegetation degradation early warning verification information fed back by the vegetation degradation early warning platform, and if the vegetation degradation early warning verification information meets the preset early warning threshold conditions, store and process the vegetation degradation early warning verification information, and send the verification success signal of the corresponding vegetation degradation early warning verification information to the vegetation degradation early warning platform. S6. Send the received vegetation degradation early warning verification information to the watershed ecological management system. S7. In response to the confirmation signal that no corresponding vegetation degradation early warning verification information has been received from the watershed ecological management system within the preset monitoring period, the vegetation degradation early warning verification information is added to the early warning re-issuance queue as early warning information to be reissued. S8. Based on the early warning re-issuance queue and the preset re-issuance time interval sequence, the vegetation degradation early warning verification information is re-issued.

[0022] It should be noted that when acquiring multi-source remote sensing monitoring data for plateau lake basins, this data includes vegetation cover information and environmental factor information. Multi-source remote sensing monitoring data refers to remote sensing data from different sensors and platforms, such as satellites and drones. This data provides information on various environmental factors, including vegetation cover, soil moisture, temperature, and precipitation. Using this data, a spatiotemporal graph convolutional network model can be constructed, where the spatiotemporal graph node information matches the spatiotemporal location information in the data. The spatiotemporal graph convolutional network is a graph-based deep learning model capable of handling data with complex spatiotemporal relationships and is suitable for spatiotemporally dependent ecological problems such as vegetation degradation. When historical vegetation degradation monitoring data is input into the spatiotemporal graph convolutional network model, the spatiotemporal characteristics of this historical data need to match the feature dimensions of the model's spatiotemporal graph nodes so that the model can effectively learn the spatiotemporal evolution patterns of vegetation degradation. The vegetation degradation prediction results output by the model are used to generate a spatiotemporal evolution feature map, which visually displays the spatiotemporal distribution and evolution trend of vegetation degradation. Subsequently, the feature map is sent to the vegetation degradation early warning platform, which receives and processes early warning information and provides verification information. When the verification information meets the preset early warning threshold conditions, it is stored and a successful verification signal is sent back to the early warning platform. Finally, the early warning verification information is sent to the watershed ecological management system. If no confirmation signal is received within the preset monitoring period, the information is added to the early warning re-issuance queue as a pending early warning information and re-issued according to the preset re-issuance time interval sequence.

[0023] Specifically, vegetation cover information in multi-source remote sensing monitoring data refers to the ratio of vegetation cover area to total land area calculated from remote sensing images, typically used to assess vegetation growth and coverage. Environmental factor information includes natural environmental factors such as temperature, precipitation, soil moisture, and wind speed, which directly affect vegetation growth and degradation. Spatiotemporal graph nodes in a spatiotemporal graph convolutional network model refer to data points with specific locations and attributes in time and space, such as vegetation cover and environmental factor values ​​at a specific geographic location at a given time. The model's input data needs to undergo spatiotemporal alignment processing to ensure that data from different timestamps and geographic coordinates are unified within the same spatiotemporal framework. Preset warning threshold conditions refer to thresholds set based on the severity of vegetation degradation; when the model's predicted degradation level exceeds this threshold, a warning is triggered. The watershed ecological management system is a comprehensive management platform responsible for receiving and processing vegetation degradation warning information and responding and confirming it based on actual conditions. The warning re-issuance queue is a data structure used to store warning information that was not successfully sent or confirmed for subsequent re-issuance.

[0024] Preferably, the construction process of the spatiotemporal graph convolutional network model is as follows: First, spatiotemporal alignment processing is performed on multi-source remote sensing monitoring data to extract timestamp information and geographic coordinate information from the data, and these are mapped to a unified time series grid and spatial coordinate system. Then, missing value imputation processing is performed on the mapped data to generate a standardized spatiotemporal dataset. After confirming that the data quality meets preset conditions, a model containing a spatiotemporal feature extraction module is constructed based on the standardized spatiotemporal dataset and the spatiotemporal graph convolutional network architecture. This module extracts spatiotemporal features from the data through graph convolution operations; these features can reflect the spatiotemporal evolution patterns of vegetation degradation.

[0025] Furthermore, during model training, historical vegetation degradation monitoring data is input, and model parameters are adjusted through optimization algorithms to enable the model to accurately predict the trend and extent of vegetation degradation. In the retransmission of early warning information, appropriate time intervals are selected for retransmission based on a preset retransmission time interval sequence. For example, weight factors for each time interval can be calculated based on historical retransmission success rate data, and time intervals with weight factors below a preset threshold are removed to generate an optimized retransmission time interval sequence. During the retransmission process, used time intervals are deleted from the sequence after each retransmission, and whether to continue retransmission is determined based on whether a confirmation signal is received.

[0026] In some embodiments, vegetation degradation early warning verification information is re-issued according to an early warning re-issuance queue and a preset re-issuance time interval sequence, including: In response to the determination that the pre-warning information to be reissued, corresponding to the vegetation degradation pre-warning verification information in the pre-warning reissue queue, meets the preset reissue conditions, the preset reissue time interval that meets the preset time order condition in the preset reissue time interval sequence is determined as the target reissue time interval. Based on the determined target retransmission interval, the following retransmission steps are performed for vegetation degradation early warning verification information: the early warning information to be retransmitted corresponding to vegetation degradation early warning verification information is sent to the watershed ecological management system to retransmit the vegetation degradation early warning verification information. The determined target retransmission time interval is deleted from the preset retransmission time interval sequence to obtain the updated preset retransmission time interval sequence as the preset retransmission time interval sequence. Determine whether a confirmation signal for the corresponding vegetation degradation early warning verification information sent by the watershed ecological management system has been received within the preset monitoring period; In response to the confirmation signal received from the watershed ecological management system within the preset monitoring period, the pre-issued early warning information will be removed from the early warning reissue queue.

[0027] It should be noted that the method described in this paper, which involves re-issuing vegetation degradation early warning verification information based on the early warning re-issuance queue and the preset re-issuance time interval sequence, is a mechanism designed to address potential unconfirmed or lost information during transmission. The early warning re-issuance queue is a data structure used to store early warning information to be re-issued, while the preset re-issuance time interval sequence is a list of time intervals used to determine the time interval for each re-issuance. When the early warning information to be re-issued in the early warning re-issuance queue meets the preset re-issuance conditions, the system selects a time interval from the preset re-issuance time interval sequence that meets the preset time order condition as the target re-issuance time interval. The preset re-issuance condition here refers to the absence of a confirmation signal from the watershed ecological management system within a preset monitoring period, while the preset time order condition refers to selecting time intervals according to the order of the time interval sequence. After determining the target re-issuance time interval, the system executes the re-issuance step, sending the early warning information to be re-issued to the watershed ecological management system, and deleting the used time interval from the time interval sequence after successful transmission to avoid reuse. This process ensures that early warning information is delivered in a timely and reliable manner, improving the effectiveness and stability of the early warning system.

[0028] Specifically, the early warning re-issuance queue is a first-in, first-out (FIFO) data structure used to store early warning information that has not received a confirmation signal within a preset monitoring period. Each early warning to be resent contains detailed information on vegetation degradation early warning verification, such as the degree, location, and time of vegetation degradation. The preset resending time interval sequence is a list of time intervals, such as [5 minutes, 10 minutes, 20 minutes]. These time intervals are preset based on historical resending success rate data and are used to determine the time interval for each resending. When an early warning to be resent meets the preset resending conditions, the system selects a time interval from this sequence as the target resending time interval. For example, if the preset time order condition is to select the first time interval in the sequence, then the first resending time interval is 5 minutes. During the resending process, the system records the time and result of each resending to determine whether further resending is needed in subsequent steps. If a confirmation signal is received from the watershed ecological management system within the preset monitoring period, the early warning to be resent is deleted from the early warning re-issuance queue, indicating that the early warning has been successfully delivered and confirmed.

[0029] Preferably, in the retransmission step, the system selects the target retransmission time interval according to preset logic. For example, the system can calculate a weight factor for each time interval based on historical retransmission success rate data. The weight factor reflects the probability of success for that time interval during the retransmission process. If the weight factor of a certain time interval is lower than a preset threshold, such as 0.5, then that time interval will be removed from the sequence, generating an optimized preset retransmission time interval sequence. This ensures that the system prioritizes time intervals with higher success rates during the retransmission process, improving retransmission efficiency.

[0030] Furthermore, during the retransmission process, the system dynamically updates the preset retransmission time interval sequence. After each successful retransmission, the used time interval is removed from the sequence, and the system decides whether to continue retransmission based on whether an acknowledgment signal is received. For example, if no acknowledgment signal is received after the first retransmission, the system will select the next time interval from the optimized sequence to continue retransmission. This process continues until an acknowledgment signal is received or all time intervals have been used up. In this way, the system can flexibly respond to different network conditions and system response delays, ensuring that early warning information can be delivered to the watershed ecological management system in a timely and effective manner.

[0031] In some embodiments, the reissue step further includes: In response to the confirmation signal that no corresponding vegetation degradation early warning verification information was received from the watershed ecological management system within the preset monitoring period, a preset retransmission time interval that meets the preset time order condition is selected from the preset retransmission time interval sequence as the target retransmission time interval, and the retransmission step is executed again for the vegetation degradation early warning verification information.

[0032] It should be noted that the retransmission step mentioned in this method refers to a mechanism automatically initiated by the system when the early warning information fails to be confirmed by the watershed ecological management system within a preset monitoring period. The core of this mechanism is to ensure that important early warning information can be reliably delivered and confirmed. Specifically, when the system determines that it has not received a confirmation signal from the watershed ecological management system for the corresponding early warning verification information within the preset monitoring period, it selects a time interval from the preset retransmission time interval sequence that meets the preset time order condition as the target retransmission time interval and executes the retransmission step again. The preset time order condition here refers to selecting time intervals sequentially according to the order of the time interval sequence, such as selecting the next time interval in the sequence. In this way, the system can continuously attempt retransmission until a confirmation signal is received, thereby improving the reliability of early warning information transmission and the robustness of the system.

[0033] Specifically, the resending step refers to the process by which the system resends the warning information when no acknowledgment signal is received. In this process, the preset resending time interval sequence is a list containing multiple time intervals, such as [5 minutes, 15 minutes, 30 minutes]. These time intervals are preset based on historical resending success rate data and are used to determine the time interval for each resending. The target resending time interval is a specific time interval selected from this sequence to determine the time point for the next resending. For example, if no acknowledgment signal is received on the first resending, the system will select the next time interval from the sequence, such as 15 minutes, as the target resending time interval and resend the warning information after that time interval. Furthermore, the preset monitoring period refers to the time window during which the system waits for an acknowledgment signal, such as 30 minutes. If no acknowledgment signal is received within this period, the system will trigger the resending mechanism. This process continues until an acknowledgment signal is received or all time intervals have been used up.

[0034] Preferably, to improve the efficiency and reliability of the retransmission mechanism, the system can further optimize the retransmission steps. For example, when selecting the target retransmission time interval, the system can calculate a weight factor for each time interval based on historical retransmission success rate data. The weight factor reflects the probability of success for that time interval during the retransmission process. If the weight factor of a certain time interval is lower than a preset threshold, such as 0.5, it is removed from the sequence, generating an optimized preset retransmission time interval sequence. This ensures that the system prioritizes time intervals with higher success rates during the retransmission process. During the retransmission process, the system records the time and result of each retransmission to dynamically adjust the retransmission strategy. For example, if no acknowledgment signal is received after multiple consecutive retransmissions, the system can automatically adjust the preset monitoring period or increase the retransmission frequency.

[0035] Furthermore, the system can also introduce an anomaly detection mechanism during the resending process. For example, when a communication anomaly of the watershed ecological management system is detected, a communication failure alarm is automatically triggered, and the alarm level is determined according to the preset ecological early warning level standard, thereby further improving the system's reliability and emergency response capability.

[0036] In some embodiments, a spatiotemporal graph convolutional network model is constructed based on vegetation cover information and environmental factor information included in multi-source remote sensing monitoring data, including: Spatiotemporal alignment processing is performed on multi-source remote sensing monitoring data to obtain a standardized spatiotemporal dataset; In response to the determination that the standardized spatiotemporal dataset meets the preset data quality conditions, a spatiotemporal graph convolutional network model containing a spatiotemporal feature extraction module is constructed based on the standardized spatiotemporal dataset and the spatiotemporal graph convolutional network architecture.

[0037] It should be noted that constructing a spatiotemporal graph convolutional network model based on vegetation cover and environmental factor information from multi-source remote sensing monitoring data, as mentioned in this method, is one of the core steps in the entire vegetation degradation early warning method. Multi-source remote sensing monitoring data refers to remote sensing data from different sensors and platforms, such as satellites and drones. This data can provide information on various environmental factors, including vegetation cover, soil moisture, temperature, and precipitation. Vegetation cover information refers to the ratio of vegetation cover area to total land area calculated from remote sensing images, used to assess the growth status and coverage of vegetation. Environmental factor information includes natural environmental factors such as temperature, precipitation, soil moisture, and wind speed, which directly affect vegetation growth and degradation. The spatiotemporal graph convolutional network model is a graph-based deep learning model capable of handling data with complex spatiotemporal relationships, and is suitable for spatiotemporally dependent ecological problems such as vegetation degradation. By performing spatiotemporal alignment on these data to obtain a standardized spatiotemporal dataset, and constructing a spatiotemporal graph convolutional network model that includes a spatiotemporal feature extraction module, the spatiotemporal features of vegetation degradation can be effectively extracted, providing a foundation for subsequent early warning.

[0038] Specifically, multi-source remote sensing monitoring data refers to remote sensing data acquired from different sources, including satellite sensors, drones, or other monitoring equipment. For example, satellite remote sensing data can provide information on vegetation cover over a wide area, while drone monitoring data can provide higher-resolution information on local areas. Vegetation cover information refers to the ratio of vegetation cover area to total land area calculated from remote sensing imagery, and is typically used to assess vegetation growth and coverage. Environmental factor information includes natural environmental factors such as temperature, precipitation, soil moisture, and wind speed, which directly affect vegetation growth and degradation. When constructing a spatiotemporal graph convolutional network model, spatiotemporal alignment of this data is required, unifying data with different timestamps and geographic coordinates into the same spatiotemporal framework. This includes extracting timestamp and geographic coordinate information from the data, mapping different timestamp information to a unified time series grid according to a preset time window division rule, and transforming different geographic coordinate information to a unified spatial coordinate system according to preset geographic grid division parameters. The final standardized spatiotemporal dataset is a dataset with missing value imputation, used as input to the model. Spatiotemporal graph convolutional network model is a deep learning model whose core is spatiotemporal graph nodes. These nodes correspond to spatiotemporal location information in the data and can capture the temporal and spatial variation patterns of vegetation cover and environmental factors.

[0039] Preferably, the process of constructing a spatiotemporal graph convolutional network model can be further refined into the following steps: First, preprocess the multi-source remote sensing monitoring data, including data cleaning, format conversion, and standardization. For example, for vegetation cover data, normalization can be used to convert it into values ​​between 0 and 1 for easier model processing. For environmental factor data, unit conversion and standardization can be performed to make it comparable. Second, spatiotemporal alignment is performed, extracting timestamp and geographic coordinate information from the data and mapping them to a unified time series grid and spatial coordinate system. For example, timestamp information can be mapped to a time grid in hours, and geographic coordinate information can be converted to a unified geographic coordinate system, such as the WGS-84 coordinate system. Then, missing value imputation is performed on the mapped data, for example, using linear interpolation or neighborhood average to imput missing data, generating a standardized spatiotemporal dataset. Next, based on the standardized spatiotemporal dataset and the spatiotemporal graph convolutional network architecture, a model containing a spatiotemporal feature extraction module is constructed. This module extracts spatiotemporal features from the data through graph convolution operations, and these features can reflect the spatiotemporal evolution of vegetation degradation. During model training, historical vegetation degradation monitoring data is input, and model parameters are adjusted through optimization algorithms to enable the model to accurately predict the trend and extent of vegetation degradation. Finally, the spatiotemporal evolution feature map is generated based on the vegetation degradation prediction results output by the model, providing visualization support for subsequent early warning.

[0040] In some embodiments, the method further includes: In response to the determination that the standardized spatiotemporal dataset does not meet the preset data quality conditions, the abnormal sensor numbers included in the multi-source remote sensing monitoring data are determined. Preset data anomaly alarm information is sent to the monitoring device terminal corresponding to the anomaly sensor number.

[0041] It should be noted that the method described in this article, which involves determining the abnormal sensor numbers included in the multi-source remote sensing monitoring data in response to the determination that the standardized spatiotemporal dataset does not meet the preset data quality conditions, and sending preset data anomaly alarm information to the monitoring equipment terminal corresponding to the abnormal sensor number, is a data quality monitoring mechanism. Before constructing the spatiotemporal graph convolutional network model, the multi-source remote sensing monitoring data needs to be checked for data quality. If the data is found to not meet the preset quality conditions, such as excessive data loss or excessive data noise, the system will further analyze and determine which sensors provide data with anomalies. The numbers of these sensors will be recorded to accurately locate the source of the problematic data. Subsequently, the system will send preset data anomaly alarm information to the corresponding monitoring equipment terminal to notify equipment maintenance personnel or relevant management personnel to check and repair. This mechanism can promptly detect and handle data quality problems, ensuring that the data input to the model is reliable and accurate, thereby improving the model's performance and the accuracy of early warnings.

[0042] Specifically, a standardized spatiotemporal dataset refers to a dataset that has undergone spatiotemporal alignment, where the data has been unified within the same temporal and spatial framework and has undergone preprocessing steps such as imputation of missing values. Data quality conditions refer to indicators such as data completeness, accuracy, and consistency; for example, the data missing rate should not exceed 10%, and the data noise level should be within an acceptable range. Anomaly sensor IDs are unique identifiers for sensors that provide data that does not meet quality requirements, allowing traceability to specific sensor devices. Predefined data anomaly alarm messages are predefined notification messages containing the anomaly sensor ID, anomaly type (e.g., missing data, excessive data noise), and suggested handling measures. Monitoring device terminals refer to the software or hardware systems installed on sensor devices, used to receive and process alarm information and notify relevant personnel for intervention. When the system detects that data does not meet quality conditions, it triggers an alarm mechanism, sending alarm information to the corresponding monitoring device terminal so that timely measures can be taken to repair or replace the anomaly sensor.

[0043] Preferably, to achieve more efficient data quality monitoring and anomaly handling, the system can adopt the following detailed steps: First, define clear data quality indicators, such as data integrity indicators (e.g., missing value ratio), data accuracy indicators (e.g., noise level), and data consistency indicators (e.g., deviation from other sensor data). During the data preprocessing stage, perform quality assessment on the multi-source remote sensing monitoring data and calculate these indicator values ​​for each sensor's data. If the data from a certain sensor does not meet preset quality conditions, such as a missing value ratio exceeding 10% or a noise level exceeding a threshold, the system will mark the sensor as an abnormal sensor and record its number. Then, the system will generate detailed alarm information, including the abnormal sensor number, anomaly type, the time and spatial range of the abnormal data, and suggested handling measures. Alarm information can be sent to the monitoring equipment terminal in various ways, such as via network communication protocols like TCP / IP to the equipment's management software, or via SMS or email to notify equipment maintenance personnel.

[0044] Furthermore, the system can also log alarm events for subsequent analysis and tracing. Through these refined steps, the system can more accurately identify and handle data quality issues, ensuring the quality of data input into the model, thereby improving the reliability and effectiveness of the entire vegetation degradation early warning system.

[0045] In some embodiments, a spatiotemporal evolution feature map corresponding to the vegetation degradation prediction result is generated based on the vegetation degradation prediction result, including: obtaining the current monitoring time window information; Based on the obtained current monitoring time window information, generate time series coding information; The randomly generated graph convolution kernel parameter initialization information is determined as the random initialization parameter; Preset spatial topology information, time series encoding information, and random initialization parameters are determined as the elements for constructing the spatiotemporal graph. Feature fusion processing is performed on the elements constructing the spatiotemporal map to obtain a fused spatiotemporal feature vector; The fused spatiotemporal feature vectors are processed by graph convolution to obtain spatiotemporal feature map data; The spatiotemporal feature map data is visualized and rendered to obtain the rendered spatiotemporal feature map as a spatiotemporal evolution feature map.

[0046] It should be noted that the generation of spatiotemporal evolution feature maps corresponding to the vegetation degradation prediction results, mentioned in this method, is a key step in transforming the model's output prediction results into intuitive visualizations. Here, the vegetation degradation prediction results refer to the information such as the degree, extent, and future trends of vegetation degradation calculated by the spatiotemporal graph convolutional network model based on input multi-source remote sensing monitoring data. The spatiotemporal evolution feature map is a visual graph that can intuitively display the distribution and evolution of vegetation degradation in time and space, facilitating ecological managers to quickly understand the specific situation of vegetation degradation. The process of generating this feature map includes multiple steps such as obtaining the current monitoring time window information, generating time series encoding information, initializing graph convolution kernel parameters, determining the spatiotemporal map construction elements, performing feature fusion processing, graph convolution operation processing, and visualization rendering processing. These steps work together to transform complex spatiotemporal data into easily understandable graphical information.

[0047] Specifically, the vegetation degradation prediction result is the output of the spatiotemporal graph convolutional network model, containing information such as the degree of vegetation degradation (e.g., changes in vegetation cover), the extent of degradation (e.g., affected geographical areas), and potential future trends. This information is obtained through deep learning and analysis of multi-source remote sensing monitoring data. The spatiotemporal evolution feature map is a visualization tool that graphically displays the spatiotemporal information of vegetation degradation. Generating this map involves several key steps: First, the current monitoring time window information refers to the time range to be analyzed, such as data from the most recent month or quarter; time series encoding information converts the time information into a format that the model can process, such as converting dates and times into continuous numerical sequences. Next, the random initialization parameters refer to the initial weight parameters randomly generated in the graph convolutional network, which are continuously optimized during model training. The spatiotemporal map construction elements include pre-defined spatial topology information, such as the division of geographical regions, time series encoding information, and random initialization parameters, which together constitute the basic framework of the spatiotemporal map. Feature fusion processing refers to integrating data from different sources, such as vegetation cover and environmental factors, to generate a fused spatiotemporal feature vector. Graph convolution processing utilizes graph convolutional networks to compute fused spatiotemporal feature vectors, extracting more representative spatiotemporal features. Visualization rendering then transforms the extracted spatiotemporal feature map data into graphics, such as using color coding to represent the degree of vegetation degradation, generating the final spatiotemporal evolution feature map.

[0048] Preferably, the process of generating the spatiotemporal evolution feature map can be further refined as follows: First, based on the current monitoring time window information, time series encoding information is generated. For example, if the monitoring time window is from January 2024 to March 2024, each day or week within this period can be converted into a time series encoding value. Then, graph convolution kernel parameter initialization information is randomly generated; these parameters will be used for subsequent graph convolution operations. Next, combined with preset spatial topology information, such as geographic grid division, time series encoding information, and random initialization parameters, the spatiotemporal map construction elements are determined. In the feature fusion processing stage, multi-source data such as vegetation cover information and environmental factor information are fused to generate a fused spatiotemporal feature vector. For example, the values ​​of vegetation cover and soil moisture can be weighted and summed to obtain a comprehensive feature value. In the graph convolution operation processing stage, a graph convolution network is used to calculate the fused spatiotemporal feature vector to extract more representative spatiotemporal features. For example, convolution operations can be used to capture the correlation of vegetation degradation in adjacent areas and time points. Finally, in the visualization and rendering stage, according to the preset color coding rules for vegetation degradation levels (e.g., green for normal, yellow for mild degradation, and red for severe degradation), the color coding information corresponding to each spatiotemporal node in the spatiotemporal feature map data is spatially overlaid to generate an overlaid spatiotemporal feature layer. This layer is then coordinate-registered with the preset plateau lake basin base map data to ensure the geographic accuracy of the graphics. Finally, ecological warning level annotations are added to the registered spatiotemporal feature map, such as marking severely degraded areas as "high-risk," generating the final spatiotemporal evolution feature map. This process not only visually displays the spatiotemporal evolution of vegetation degradation but also provides strong support for ecological management and decision-making.

[0049] In some embodiments, the reissue step further includes: In response to the determination that no confirmation signal of the corresponding vegetation degradation early warning verification information sent by the watershed ecological management system has been received within the preset monitoring period, and the detection of communication anomaly information of the watershed ecological management system, a corresponding communication failure alarm information of the watershed ecological management system is generated based on the communication anomaly information. Based on the preset ecological early warning level standards, determine the ecological early warning level corresponding to the abnormal communication information; The communication failure alarm information and ecological early warning level are sent to the monitoring equipment terminal with the corresponding abnormal sensor number.

[0050] It should be noted that the retransmission step mentioned in this method involves generating a communication failure alarm message and determining the ecological warning level when no confirmation signal is received from the watershed ecological management system and a communication anomaly is detected. The communication anomaly message here refers to network failures, signal interruptions, or receiver system failures that occur during the transmission of warning information. These anomalies may prevent the warning information from being successfully delivered or confirmed. The communication failure alarm message is a notification message used to remind relevant personnel that there is a problem with the communication link and that it needs to be addressed promptly. Simultaneously, based on preset ecological warning level standards, the system will assess the potential impact of the communication anomaly on the ecological warning and determine the corresponding warning level in order to take appropriate emergency measures. This mechanism not only ensures the reliable transmission of warning information but also enables timely notification of relevant personnel when a communication failure occurs, improving the system's emergency response capability.

[0051] Specifically, communication anomaly information refers to any communication problems detected during the transmission of early warning information, such as excessive network latency, packet loss, or no response from the receiving system. These anomalies can be detected by the system's built-in communication status monitoring module and recorded as specific anomaly types and timestamps. Communication fault alarm information is a structured notification message containing a detailed description of the communication anomaly, such as network interruption or excessive packet loss rate, the time of occurrence, the affected early warning information number, and suggested handling measures. The preset ecological early warning level standard is a set of rules used to determine the early warning level based on the severity of the communication anomaly and its potential impact on the ecological early warning. For example, if a communication anomaly causes an early warning information delay of more than one hour, it may be rated as a high early warning level; if it is only a brief signal fluctuation, it may be rated as a low early warning level. The classification of early warning levels can help relevant personnel prioritize the most urgent issues to ensure the safety and stability of the ecosystem.

[0052] Preferably, in the resending step, the system can further optimize the generation and processing of communication failure alarms. First, when a communication anomaly is detected, the system immediately records detailed information about the anomaly, including the anomaly type, occurrence time, duration, and the affected warning information number. Then, based on preset ecological warning level standards, the system assesses the severity of the communication anomaly. For example, if a communication anomaly prevents the delivery of warning information within a preset monitoring period, and the warning information involves a severely degraded vegetation area, the system will classify it as a high warning level and generate detailed communication failure alarm information. The alarm information will be sent to relevant personnel through various channels, such as SMS, email, or instant messaging tools, to the maintenance personnel and emergency response team of the watershed ecological management system. Simultaneously, the system continuously monitors the communication status. If the communication anomaly cannot be resolved within a short time, the system can automatically adjust the resending strategy for the warning information, such as shortening the resending interval or increasing the number of resending attempts, to ensure that the warning information is delivered as quickly as possible.

[0053] Furthermore, the system can record the processing results of each communication anomaly for subsequent fault analysis and system optimization. Through these detailed steps, the system can respond quickly when communication failures occur, ensuring the reliability and effectiveness of the ecological early warning system.

[0054] In some embodiments, the spatiotemporal feature map data is subjected to visualization rendering processing to obtain the rendered spatiotemporal feature map as a spatiotemporal evolution feature map, including: Based on the preset color coding rules for vegetation degradation levels, determine the color coding information corresponding to each spatiotemporal map node in the spatiotemporal feature map data; Spatial overlay processing is performed on color encoding information to generate an overlaid spatiotemporal feature layer; The spatiotemporal feature layer is coordinate registered with the preset plateau lake basin base map data to obtain the registered spatiotemporal feature map. Based on the preset ecological early warning level labeling template, ecological early warning level labeling information is added to the registered spatiotemporal feature map to obtain a spatiotemporal evolution feature map.

[0055] It should be noted that the visualization rendering process of the spatiotemporal feature map data mentioned in this method, resulting in a rendered spatiotemporal evolution feature map, is a key step in transforming the spatiotemporal feature map data into an intuitive visual graphic. Here, the spatiotemporal feature map data refers to the dataset containing the spatiotemporal characteristics of vegetation degradation, obtained after graph convolution operations. The visualization rendering process refers to displaying this data graphically so that users can intuitively understand the spatiotemporal evolution of vegetation degradation. The final generated spatiotemporal evolution feature map is a visual graphic that clearly displays the degree, extent, and trend of vegetation degradation through color coding, spatial overlay, and annotation, providing visual support for ecological management and decision-making.

[0056] Specifically, the spatiotemporal feature map data is the output data after processing by a graph convolutional network, containing feature information of vegetation degradation at different times and spatial locations. This data typically exists in the form of vectors or matrices, with each element corresponding to the feature value of a spatiotemporal node. Visualization rendering refers to displaying this data graphically, specifically including the following steps: First, a preset vegetation degradation level color coding rule is a mapping rule used to map the degree of vegetation degradation, such as mild degradation, moderate degradation, and severe degradation, to different colors; for example, green represents normal, yellow represents mild degradation, and red represents severe degradation. Next, spatial overlay processing involves overlaying feature layers from different time points to generate a comprehensive spatiotemporal feature layer to display the evolution of vegetation degradation over time. Then, coordinate registration processing involves aligning the generated spatiotemporal feature layer with preset plateau lake basin base map data in terms of geographic coordinates to ensure the geographic accuracy of the graphics. Finally, ecological warning level labeling information refers to the text or symbols added to the spatiotemporal evolution feature map to indicate the ecological warning level. For example, marking severely degraded areas with the word "high risk" can help users quickly identify key areas of concern.

[0057] Preferably, the specific steps of the visualization rendering process can be further refined as follows: First, according to the preset color coding rules for vegetation degradation levels, assign corresponding color coding information to each spatiotemporal map node. For example, if the vegetation degradation level of a certain node is slightly degraded, then its color coding is yellow. Then, these color coding information are spatially overlaid to generate an overlaid spatiotemporal feature layer. For example, feature layers at different time points are overlaid together in chronological order to form a dynamic spatiotemporal evolution layer. Next, this layer is coordinate registered with the preset plateau lake basin base map data to ensure that the geographical location of each node is consistent with the geographical coordinates on the base map. For example, coordinate transformation and alignment operations are performed using Geographic Information System (GIS) software. Finally, ecological warning level labeling information is added to the registered spatiotemporal feature map. For example, red "high risk" is added to severely degraded areas, and yellow "low risk" is added to slightly degraded areas.

[0058] Furthermore, additional auxiliary information, such as timestamps, scale bars, and legends, can be added according to user needs to enhance the readability and usability of the graphics. Through these refinement steps, the generated spatiotemporal evolution feature map can not only intuitively display the spatiotemporal evolution of vegetation degradation, but also provide strong support for ecological management and decision-making.

[0059] In some embodiments, spatiotemporal alignment processing is performed on multi-source remote sensing monitoring data to obtain a standardized spatiotemporal dataset, including: Extract timestamp and geographic coordinate information from multi-source remote sensing monitoring data; Based on the preset time window division rules, different timestamp information is mapped to a unified time series grid; Based on the preset geographic grid division parameters, different geographic coordinate information is converted to a unified spatial coordinate system; Missing value interpolation is performed on the mapped unified time series grid and the transformed unified spatial coordinate system to generate a standardized spatiotemporal dataset.

[0060] It should be noted that the spatiotemporal alignment processing of multi-source remote sensing monitoring data mentioned in this method to obtain a standardized spatiotemporal dataset is a fundamental step in the entire vegetation degradation early warning method. Multi-source remote sensing monitoring data typically originates from different sensors and platforms, with varying temporal and spatial resolutions. Therefore, spatiotemporal alignment processing is necessary to ensure data consistency and comparability. Spatiotemporal alignment processing involves extracting timestamp and geographic coordinate information from the data and mapping this information to a unified time-series grid and spatial coordinate system. In this way, a standardized spatiotemporal dataset can be generated, providing a reliable data foundation for subsequent model building and analysis.

[0061] Specifically, multi-source remote sensing monitoring data refers to vegetation cover and environmental factor data from different sensors, such as satellite remote sensing and drone monitoring. The timestamp information of this data refers to the specific time point of data collection, while the geographic coordinate information refers to the spatial location of the data collection. The purpose of spatiotemporal alignment processing is to unify this data into a standardized spatiotemporal framework. For example, preset time window division rules can map timestamp information to a time grid with hours or days as units, so that data collected at different times can be processed uniformly. Preset geographic grid division parameters transform geographic coordinate information into a unified spatial coordinate system, such as using latitude and longitude grids to divide the entire monitoring area into several fixed-size grid units. During the alignment process, missing value imputation processing is also required for the mapped unified time series grid and the transformed unified spatial coordinate system, for example, through spatial interpolation or time series imputation methods, to generate a complete standardized spatiotemporal dataset. This process ensures the continuity and integrity of the data in time and space, providing high-quality data support for subsequent analysis.

[0062] Preferably, the specific steps of the spatiotemporal alignment processing can be further refined as follows: First, extract timestamp information and geographic coordinate information from multi-source remote sensing monitoring data. For example, extract the specific date and time of data collection from satellite remote sensing data, and extract the geographic coordinates of flight trajectories from UAV monitoring data. Then, according to preset time window division rules, map the information of different timestamps to a unified time series grid. For example, divide the timestamps according to hourly or daily time intervals to generate a time series grid. Next, according to preset geographic grid division parameters, transform the geographic coordinate information into a unified spatial coordinate system. For example, divide the entire monitoring area into 1 km × 1 km grid cells, and map the geographic coordinates of each data point to the corresponding grid cell. After mapping, imputation processing is performed on missing values. For example, for missing data in the time series, linear interpolation or moving average methods can be used for imputation; for spatially missing data, kriging interpolation or nearest neighbor interpolation methods can be used for imputation. Finally, the generated standardized spatiotemporal dataset will contain complete temporal and spatial information, providing a high-quality data foundation for subsequent vegetation degradation analysis and model construction.

[0063] In some embodiments, determining the target transmission interval information includes: The weighting factor of each time interval in the preset resend interval sequence is calculated based on historical resend success rate data. Remove the preset retransmission time intervals with weight factors below a preset threshold from the sequence to generate an optimized preset retransmission time interval sequence. Based on the optimized preset retransmission time interval sequence, the preset retransmission time interval that satisfies the preset time order condition is determined by arranging the time intervals in ascending order.

[0064] It should be noted that determining the target transmission interval information mentioned in this method is a key step in optimizing the time interval selection in the early warning information retransmission mechanism. In the retransmission mechanism, rationally selecting the retransmission time interval is crucial for improving the success rate of early warning information delivery. This step analyzes historical retransmission success rate data, calculates the weight factor of each time interval in the preset retransmission time interval sequence, and eliminates inefficient time intervals based on the weight factors, thereby generating an optimized retransmission time interval sequence. Finally, the preset retransmission time intervals that meet the preset time order conditions are determined by arranging the time intervals in ascending order, ensuring the efficiency and reliability of the retransmission operation.

[0065] Specifically, the target transmission interval information refers to the time interval selected in the retransmission mechanism for the next retransmission. This time interval selection is based on the optimization of a preset retransmission time interval sequence. This sequence is a list containing multiple time intervals, such as [5 minutes, 15 minutes, 30 minutes], which are preset based on historical retransmission experience. The effectiveness of each time interval in the retransmission process can be evaluated by calculating a weight factor. The weight factor is calculated based on historical retransmission success rate data, reflecting the probability of successful retransmission within a specific time interval. If the weight factor of a certain time interval is lower than a preset threshold, such as 0.5, the retransmission success rate of that time interval is considered low, and it should be removed from the sequence. The final optimized preset retransmission time interval sequence is a filtered, more efficient list of time intervals. Arranging the time intervals in ascending order to determine those meeting the preset time order condition ensures that the retransmission operation is performed in ascending order, thereby minimizing retransmission delay while maintaining a high success rate.

[0066] Preferably, the process of determining the target transmission interval information can be further refined as follows: First, collect and analyze historical retransmission success rate data, which records the success rate of retransmission operations at different time intervals. For example, statistical analysis reveals that the success rate of retransmission at a 5-minute time interval is 80%, while the success rate at a 15-minute time interval is 60%. Next, calculate the weighting factor for each time interval based on these success rate data. For example, the retransmission success rate can be directly used as the weighting factor, or the weighting factor can be obtained through a transformation function, such as normalization. Then, set a threshold for the weighting factor, such as 0.5, and remove time intervals with weighting factors lower than this threshold from the preset retransmission time interval sequence. For example, if the weighting factor for a 15-minute time interval is 0.4, which is lower than the threshold of 0.5, it is removed from the sequence. Finally, arrange the remaining time intervals in ascending order to generate an optimized preset retransmission time interval sequence. In the retransmission operation, the time intervals in this sequence are selected sequentially as the target transmission interval information, thereby ensuring that each retransmission is performed within the most effective time interval. This process not only improves the success rate of reissues, but also optimizes the overall efficiency of the reissue mechanism.

[0067] The above embodiments of the present invention have the following beneficial effects: 1. A spatiotemporal graph convolutional network model can be constructed based on multi-source remote sensing monitoring data. Through spatiotemporal alignment and feature fusion processing, it can effectively capture the spatiotemporal evolution pattern of vegetation degradation in plateau lake basins, solve the problem that traditional methods cannot take into account spatiotemporal correlation, and improve the accuracy of degradation prediction.

[0068] 2. The reliable transmission of early warning information can be achieved through the early warning retransmission queue and the preset retransmission time interval sequence. The retransmission mechanism is automatically triggered when no confirmation signal is received from the watershed ecological management system, which solves the problem of possible loss or delay of early warning information and ensures the timeliness of ecological management decisions.

[0069] 3. The spatiotemporal distribution characteristics of vegetation degradation prediction results can be intuitively displayed through the visualization rendering of spatiotemporal feature maps and the labeling of ecological early warning levels. This solves the problem of insufficient visualization in traditional monitoring methods, making it easier for managers to quickly identify high-risk degradation areas and take targeted measures.

[0070] Furthermore, the storage medium in the embodiments of this application stores program instructions capable of implementing all the above methods. These program instructions can be stored in the storage medium in the form of a software product, including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, or terminal devices such as computers, servers, mobile phones, and tablets.

[0071] The above description is merely an explanation of some preferred embodiments of the present invention and the technical principles employed. Those skilled in the art should understand that the scope of the invention as described in the embodiments of the present invention is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of the present invention.

Claims

1. A method for early warning of vegetation degradation in plateau lake basins based on spatiotemporal graph convolutional networks, comprising: Multi-source remote sensing monitoring data of plateau lake basins are acquired. Based on the vegetation coverage information and environmental factor information included in the multi-source remote sensing monitoring data, a spatiotemporal graph convolutional network model is constructed. The spatiotemporal graph node information included in the constructed spatiotemporal graph convolutional network model is the same as the spatiotemporal location information included in the multi-source remote sensing monitoring data. Historical vegetation degradation monitoring data is input into a spatiotemporal graph convolutional network model. The spatiotemporal feature information included in the historical vegetation degradation monitoring data has the same spatiotemporal graph node feature dimension as that included in the spatiotemporal graph convolutional network model. In response to the vegetation degradation prediction results output by the spatiotemporal graph convolutional network model, a spatiotemporal evolution feature map corresponding to the vegetation degradation prediction results is generated based on the vegetation degradation prediction results. Send the spatiotemporal evolution feature map to the vegetation degradation early warning platform; In response to the vegetation degradation early warning verification information fed back by the vegetation degradation early warning platform, and if the vegetation degradation early warning verification information meets the preset early warning threshold conditions, the vegetation degradation early warning verification information is stored and processed, and the verification success signal of the corresponding vegetation degradation early warning verification information is sent to the vegetation degradation early warning platform. The received vegetation degradation early warning verification information is sent to the watershed ecological management system. In response to the confirmation signal that no corresponding vegetation degradation early warning verification information has been received from the watershed ecological management system within the preset monitoring period, the vegetation degradation early warning verification information is added to the early warning re-issuance queue as early warning information to be reissued. The method reissues vegetation degradation early warning verification information according to the early warning reissue queue and the preset reissue time interval sequence; wherein, the method further includes: in response to determining that the early warning information to be reissued corresponding to the vegetation degradation early warning verification information in the early warning reissue queue meets the preset reissue conditions, the preset reissue time interval that meets the preset time order condition in the preset reissue time interval sequence is determined as the target reissue time interval, including: The weighting factor of each time interval in the preset resend interval sequence is calculated based on historical resend success rate data. Remove the preset retransmission time intervals with weight factors below a preset threshold from the sequence to generate an optimized preset retransmission time interval sequence. Based on the optimized preset retransmission time interval sequence, the preset retransmission time interval that satisfies the preset time order condition is determined by arranging the time intervals in ascending order.

2. The method according to claim 1, characterized in that, Based on the early warning re-issuance queue and the preset re-issuance time interval sequence, the vegetation degradation early warning verification information is re-issued, including: Based on the determined target retransmission interval, the following retransmission steps are performed for vegetation degradation early warning verification information: the early warning information to be retransmitted corresponding to vegetation degradation early warning verification information is sent to the watershed ecological management system to retransmit the vegetation degradation early warning verification information. The determined target retransmission time interval is deleted from the preset retransmission time interval sequence to obtain the updated preset retransmission time interval sequence as the preset retransmission time interval sequence. Determine whether a confirmation signal for the corresponding vegetation degradation early warning verification information sent by the watershed ecological management system has been received within the preset monitoring period; In response to the confirmation signal received from the watershed ecological management system within the preset monitoring period, the pre-issued early warning information will be removed from the early warning reissue queue.

3. The method according to claim 2, characterized in that, The replacement process also includes: In response to the confirmation signal that no corresponding vegetation degradation early warning verification information was received from the watershed ecological management system within the preset monitoring period, a preset retransmission time interval that meets the preset time order condition is selected from the preset retransmission time interval sequence as the target retransmission time interval, and the retransmission step is executed again for the vegetation degradation early warning verification information.

4. The method according to claim 1, characterized in that, Based on vegetation cover and environmental factor information included in multi-source remote sensing monitoring data, a spatiotemporal graph convolutional network model is constructed, including: Spatiotemporal alignment processing is performed on multi-source remote sensing monitoring data to obtain a standardized spatiotemporal dataset; In response to the determination that the standardized spatiotemporal dataset meets the preset data quality conditions, a spatiotemporal graph convolutional network model containing a spatiotemporal feature extraction module is constructed based on the standardized spatiotemporal dataset and the spatiotemporal graph convolutional network architecture.

5. The method according to claim 4, characterized in that, The method also includes: In response to the determination that the standardized spatiotemporal dataset does not meet the preset data quality conditions, the abnormal sensor numbers included in the multi-source remote sensing monitoring data are determined. Preset data anomaly alarm information is sent to the monitoring device terminal corresponding to the anomaly sensor number.

6. The method according to claim 1, characterized in that, Based on the vegetation degradation prediction results, a spatiotemporal evolution feature map corresponding to the vegetation degradation prediction results is generated, including: obtaining the current monitoring time window information; Based on the obtained current monitoring time window information, generate time series coding information; The randomly generated graph convolution kernel parameter initialization information is determined as the random initialization parameter; Preset spatial topology information, time series encoding information, and random initialization parameters are determined as the elements for constructing the spatiotemporal graph. Feature fusion processing is performed on the elements constructing the spatiotemporal map to obtain a fused spatiotemporal feature vector; The fused spatiotemporal feature vectors are processed by graph convolution to obtain spatiotemporal feature map data; The spatiotemporal feature map data is visualized and rendered to obtain the rendered spatiotemporal feature map as a spatiotemporal evolution feature map.

7. The method according to claim 5, characterized in that, The replacement process also includes: In response to the determination that no confirmation signal of the corresponding vegetation degradation early warning verification information sent by the watershed ecological management system has been received within the preset monitoring period, and the detection of communication anomaly information of the watershed ecological management system, a corresponding communication failure alarm information of the watershed ecological management system is generated based on the communication anomaly information. Based on the preset ecological early warning level standards, determine the ecological early warning level corresponding to the abnormal communication information; The communication failure alarm information and ecological early warning level are sent to the monitoring equipment terminal with the corresponding abnormal sensor number.

8. The method according to claim 6, characterized in that, The spatiotemporal feature map data is visualized and rendered to obtain the rendered spatiotemporal feature map as a spatiotemporal evolution feature map, including: Based on the preset color coding rules for vegetation degradation levels, determine the color coding information corresponding to each spatiotemporal map node in the spatiotemporal feature map data; Spatial overlay processing is performed on color encoding information to generate an overlaid spatiotemporal feature layer; The spatiotemporal feature layer is coordinate registered with the preset plateau lake basin base map data to obtain the registered spatiotemporal feature map. Based on the preset ecological early warning level labeling template, ecological early warning level labeling information is added to the registered spatiotemporal feature map to obtain a spatiotemporal evolution feature map.

9. The method according to claim 4, characterized in that, Spatiotemporal alignment of multi-source remote sensing monitoring data yields a standardized spatiotemporal dataset, including: Extract timestamp and geographic coordinate information from multi-source remote sensing monitoring data; Based on the preset time window division rules, different timestamp information is mapped to a unified time series grid; Based on the preset geographic grid division parameters, different geographic coordinate information is converted to a unified spatial coordinate system; Missing value interpolation is performed on the mapped unified time series grid and the transformed unified spatial coordinate system to generate a standardized spatiotemporal dataset.

Images