A multi-level forest resource monitoring method and system

Abstract

The present invention relates to the fields of computer technology and forest resource intelligent monitoring technology, and specifically discloses a multi-level forest resource monitoring method and system. The method includes: collecting computing power, storage and communication status information of edge nodes and normalizing them, and constructing a task resource mapping in combination with a forest resource monitoring task set; constructing a scheduling optimization model based on this mapping, performing multi-constraint scoring and iterative optimization, and generating a task-node matching structure; compressing a central graph neural network model according to the matching results and deploying it to the edge node; selecting a communication mode in combination with the communication status, and configuring channel resources; executing local inference and asynchronously transmitting data back; finally extracting features to complete data aggregation, feeding back task weights and node scores, and updating the scheduling strategy in a closed loop. The present invention effectively improves the efficiency of global monitoring scheduling and system stability in environments with uneven resource distribution and communication fluctuations.

Description

Technical Field

[0001] The present invention relates to the fields of computer technology and forest resource intelligent monitoring technology, and in particular to a multi-level forest resource monitoring method and system. Background Art

[0002] With the increasing requirements for ecological protection and the deepening of digital management of forest resources, traditional forest monitoring methods have gradually exposed the following problems: First, the monitoring system mostly relies on central node processing. Faced with the forest environment with wide geographical distribution and strong resource heterogeneity, it cannot effectively adapt to the reality of unbalanced computing power, storage and communication capabilities among edge nodes, resulting in high overall task response delay; Second, the existing scheduling methods generally adopt static or rule-driven strategies, lacking real-time perception and response capabilities to dynamic state changes of edge nodes, making it difficult to achieve global optimal allocation of monitoring tasks under resource-constrained conditions; Third, the high bandwidth occupancy and poor stability of the data backhaul link further restrict the real-time feedback and closed-loop processing capabilities of monitoring results.

[0003] The existing technology still lacks a method and system that can dynamically construct task-resource mapping relationships under multi-dimensional resource constraints, and combine edge node state partial derivative analysis with system functional control mechanism to achieve optimal matching allocation between forest resource monitoring tasks and nodes. Summary of the Invention

[0004] The present invention provides a multi-level forest resource monitoring method and system to solve the problem of how to construct a multi-dimensional task resource mapping based on the computing power status, storage capacity and communication conditions of edge nodes, and by introducing a multi-constraint scheduling scoring function, a node state partial derivative mechanism and a system energy functional optimization model, a dynamic and stable forest resource monitoring task-node optimal matching structure is achieved, thereby improving the global monitoring scheduling efficiency and feedback closed-loop capability in a heterogeneous resource environment.

[0005] In order to solve the above technical problems, the present invention provides a multi-level forest resource monitoring method, comprising:

[0006] By normalizing the status information of each edge node and the forest resource monitoring task set, a task resource mapping is constructed;

[0007] By obtaining task resource mapping, building a scheduling optimization model and performing iterative matching, a matching allocation structure is generated;

[0008] The expression of the scheduling optimization model is:

[0009]

[0010] in, For the task With edge nodes Comprehensive fit score; The task-node matching set formed by the current scheduling strategy; is the scheduling conflict cost function; is the conflict penalty factor; is the comprehensive matching score of the scheduling strategy;

[0011] The match assignment structure includes a model deployment priority score:

[0012]

[0013] in, Score the deployment priority of task-node pairs; For nodes Current scheduling deviation risk indicator;

[0014] Compress the central GNN model and deploy it to the edge node according to the matching allocation structure, build the model deployment structure, and complete the registration tag and version binding;

[0015] According to the model deployment structure and communication status information, select the communication mode and configure the communication channel to complete the synchronization and storage of the communication configuration structure;

[0016] Execute local inference tasks of the model deployment structure, complete the encoding and asynchronous return of inference results, and generate the return data structure;

[0017] Perform feature extraction and data aggregation on the returned data structure to generate task weight strategy and node resource score, which are fed back to the task resource mapping and update the configuration strategy.

[0018] Furthermore, constructing a task resource mapping includes:

[0019] Normalize the computing power status information, storage status information, and communication status information of each edge node to generate edge node status information;

[0020] According to task granularity, task priority and computational complexity, the forest resource monitoring task set is structured and processed to generate task set information.

[0021] Construct task resource mapping based on edge node status information and task set information.

[0022] Furthermore, building a scheduling optimization model includes:

[0023] Obtain task resource mapping and build a scheduling optimization model based on edge node status information and task set information;

[0024] Execute task scheduling matching operations and generate scheduling intermediate results;

[0025] Iterative optimization and constraint rearrangement are performed based on the intermediate scheduling results to update the scheduling results.

[0026] Furthermore, generating a matching allocation structure includes:

[0027] Extract the optimal allocation relationship based on the scheduling results after iterative optimization and build a matching allocation structure;

[0028] Generate edge node task execution list based on matching allocation structure.

[0029] Furthermore, building a model deployment structure includes:

[0030] Obtain the matching allocation structure, perform compression processing on the central graph neural network model, and generate a lightweight model;

[0031] Deploy the lightweight model to the edge nodes according to the matching distribution structure and build the model deployment structure;

[0032] Register the model deployment structure and bind the version information.

[0033] Furthermore, configuring the communication channel includes:

[0034] Obtain model deployment structure and communication status information, and select a communication mode based on the communication quality level;

[0035] Build communication configuration structure and allocate communication channel resources;

[0036] Complete the edge node synchronization processing of the communication configuration structure and store the configuration results in the communication configuration table.

[0037] Furthermore, performing local reasoning tasks includes:

[0038] Obtain the model deployment structure deployed on the edge node and call the lightweight model to perform local inference tasks;

[0039] Obtain inference result data and encode the data according to the communication configuration structure;

[0040] Asynchronous transmission of inference results is performed through the communication configuration structure to generate the transmission data structure.

[0041] Furthermore, the update configuration strategy includes:

[0042] Obtain the returned data structure, perform global feature extraction and multi-node data aggregation;

[0043] Generate task weight strategy and node resource score;

[0044] Feedback the task weight strategy and node resource score to the task resource mapping to complete the strategy update.

[0045] Furthermore, the graph neural network model is compressed through knowledge distillation, and the communication configuration structure is constructed using a time-sharing multiple access mechanism.

[0046] Furthermore, a multi-level forest resource monitoring system is applied to any of the multi-level forest resource monitoring methods described above, comprising:

[0047] The task resource construction module is used to obtain edge node status information and forest resource monitoring task sets and build task resource mapping;

[0048] The task scheduling module is used to perform scheduling optimization based on task resource mapping and generate matching allocation structure;

[0049] Model compression and deployment module, used to generate lightweight models based on the matching allocation structure and deploy them to edge nodes;

[0050] The communication configuration module is used to obtain the model deployment structure and communication status information, build the communication configuration structure and complete the channel configuration;

[0051] The local reasoning and return module is used to perform local reasoning tasks, complete the return of reasoning results, and generate the return data structure;

[0052] The data feedback module is used to generate task weight strategies and node resource scores based on the returned data structure, and update the task resource mapping.

[0053] The key innovations of the present invention include:

[0054] (1) The system integrates the computing power, storage capacity, and communication status of edge nodes into the scheduling scoring structure, forming a cross-dimensional fusion scheduling input feature system to improve the accuracy of task-node adaptation.

[0055] (2) Explicit modeling and iterative reconstruction of scheduling stability are achieved, solving the scheduling instability problem caused by rapid fluctuations in resource status.

[0056] (3) Integrate scoring, deployment, communication and feedback into a unified structure, open up the entire link of "scheduling-deployment-communication-feedback", and have a highly scalable and standardized implementation path.

[0057] The following are its main beneficial effects:

[0058] (1) This paper introduces a multi-dimensional task resource mapping mechanism that integrates computing power, storage, and communication status. By constructing an adaptation scoring function to comprehensively score task-node combinations, it achieves high-precision adaptation in a heterogeneous resource environment. Compared with traditional coarse-grained scheduling methods based on task priority or geographic distance, this method can effectively reduce task scheduling delays and improve the real-time response capability of forest resource monitoring tasks.

[0059] (2) This invention constitutes a new iterative optimization scheduling framework, in which partial derivatives can identify high-volatility nodes in real time, while functional objectives can suppress scheduling jitter caused by unstable resources, effectively avoiding problems such as node overload, data congestion, and increased failure rates. The system maintains stable operation in dynamic environments, providing structural protection for edge intelligent monitoring.

[0060] (3) The deployment priority scoring function introduced in the matching allocation structure directly links the scheduling results with the subsequent model deployment and communication configuration, promotes the formation of a linkage closed loop among model compression deployment, channel resource allocation and backhaul strategy, effectively improves the overall efficiency of collaborative monitoring between heterogeneous nodes, and ensures that the system has end-to-end data consistency and scheduling execution consistency. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 A schematic diagram of a multi-level forest resource monitoring method provided in an embodiment of the present application;

[0062] Figure 2 This is a structural block diagram of a multi-level forest resource monitoring system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0063] Example 1: Reference Figure 1 , is a flow chart of a multi-level forest resource monitoring method provided by an embodiment of the present invention. The flow chart may include at least steps S100-S600:

[0064] S100: constructing a task resource mapping by normalizing the status information of each edge node and the forest resource monitoring task set.

[0065] S200: Obtain task resource mapping, build a scheduling optimization model, perform iterative matching, and generate a matching allocation structure.

[0066] S300: Compress the central GNN model and deploy it to the edge node according to the matching allocation structure, build the model deployment structure, and complete the registration tag and version binding.

[0067] S400: Select a communication mode and configure a communication channel based on the model deployment structure and communication status information, and complete synchronization and storage of the communication configuration structure.

[0068] S500: Execute the local reasoning task of the model deployment structure, complete the encoding and asynchronous return of the reasoning results, and generate the return data structure.

[0069] S600: Extract features and aggregate data from the returned data structure, generate task weight strategy and node resource score, feed back to task resource mapping and update configuration strategy.

[0070] Step S100 at least includes steps S110-S130:

[0071] S110: Obtain computing power status information, storage status information, and communication status information of each edge node, perform unified normalization processing, and generate edge node status information.

[0072] Specifically, edge node status information includes, but is not limited to, processor type, main frequency, memory capacity, remaining storage capacity, communication bandwidth, connection stability, current operating load level, and energy consumption evaluation indicators. The system first collects status information for all online edge nodes through edge monitoring terminals or a management center, relying on the hardware configuration list uploaded during node registration and periodic operating status reports as preliminary raw data.

[0073] Among them, the collection of computing power status information includes dynamic sampling of the node's available processing resources, combining historical task response time and peak scheduling load to construct the computing power index value in the current cycle; storage status information includes the node's current remaining storage capacity, data cache area size and write rate; communication status information includes the average signal-to-noise ratio, packet loss rate, average bandwidth occupancy and the number of recent communication failures between the communication base station.

[0074] After collecting raw state data, the system performs a unified normalization process. This normalization method uses interval mapping and a time window sliding average to uniformly map all indicators to the closed interval [0, 1], eliminating the impact of inconsistent data dimensions on scheduling scores. The normalization parameter template is initially set by the system or dynamically updated through historical model feedback.

[0075] After normalization is complete, a structured edge node status information structure is generated. This structure contains the multiple dimensional state attributes mentioned above and is bound to a unique node identifier. The edge node status information structure will serve as one of the input elements of the subsequent task resource mapping construction operation and participate in the task adaptation scoring process.

[0076] S120: Obtain a forest resource monitoring task set, perform structured processing based on task granularity, task priority, and computational complexity, and generate task set information.

[0077] Specifically, the forest resource monitoring task set comes from multiple data scheduling sources, including inspection tasks regularly issued by the central forestry management platform, high-priority response tasks triggered by sudden abnormal alarms, and temporary tasks generated based on remote sensing satellite or drone telemetry analysis.

[0078] Each monitoring task includes a task identifier, monitoring target location, task type (such as image acquisition, audio monitoring, temperature and humidity monitoring), execution period, data return requirements, and business urgency level. The system breaks down this information into tasks, extracting the smallest executable task unit and assigning granularity-level tags to these tasks according to predefined rules.

[0079] The task granularity label is derived based on the task space scope and sampling density assessment; the task priority label is assigned based on the task source type and response time limit assessment; and the computational complexity assessment is based on the inference time, collected data volume, processing path structure, and other aspects of similar task models in the past.

[0080] The system then performs structured processing on the task set to generate a standardized task set information structure, which includes fields such as task type, target location code, priority score, and computational complexity score. This task set information structure will be subsequently coupled with the edge node status information structure to participate in the task resource mapping process.

[0081] S130: Construct a task resource mapping based on the edge node status information and the task set information.

[0082] After obtaining edge node status information and task set information, this step executes the task resource matching structure construction operation, which specifically includes: multi-dimensional feature matching, schedulability judgment, and task scoring function calculation.

[0083] First, the system spatially matches the geographical distribution of edge nodes based on the task location parameters recorded in the task set information, and eliminates invalid nodes whose distance from the target task location exceeds the deployment threshold range.

[0084] Next, for each remaining node-task pair, computing power adaptability matching is performed. The system calculates a computing power matching score based on the node's normalized computing power index and the task's computational complexity label. Storage adaptability is scored based on the node's storage capacity and the task's data volume. Communication adaptability is also assessed based on communication status indicators and task data backhaul requirements.

[0085] The three scoring dimensions described above form the core of the task-resource matching evaluation. The system calculates the overall matching score for each node-task pair based on a weighted function and marks the feasibility flag in the scoring structure. This scoring structure is stored in a table structure as the output object of the task-resource mapping.

[0086] Finally, the completed task resource mapping contains fields such as task number, node number, comprehensive score value, schedulability flag, etc., forming standardized input data as the key input for scheduling optimization model construction and task matching calculation.

[0087] By implementing this step, the system can achieve multi-dimensional structured modeling between task sets and edge node capabilities, forming a complete task-resource mapping relationship. This mapping result significantly improves the perception accuracy of subsequent scheduling models for resource heterogeneity, providing a high-quality scoring basis. At the same time, it achieves unified quantification and schedulability expression of three core resources: computing power, storage, and communication. This provides key support for the efficient allocation and execution of multi-level forest resource monitoring tasks, effectively reduces overall scheduling delays, and improves the system's resource scheduling robustness in large-scale node environments.

[0088] Step S200 at least includes steps S210-S230:

[0089] S210: Obtain the task resource mapping, build an enhanced scheduling optimization model, and perform task scheduling matching operations.

[0090] The system first obtains the task resource mapping constructed by S130. This mapping includes the task number, node number, compute adaptability score, storage adaptability score, communication adaptability score, and schedulability flag. Based on this mapping, the system constructs a task-node bipartite graph structure and expresses the multidimensional resource adaptability relationship in the form of edge weights, thus constructing an enhanced scheduling optimization model.

[0091] Formula ①: Comprehensive adaptation scoring function

[0092]

[0093] in:

[0094] :Task With node Comprehensive fit score;

[0095] :Task Adaptability score in the computing power dimension;

[0096] :Task Adaptation score in the storage dimension;

[0097] :Task Fit score in the communication dimension;

[0098] : Adaptation weight coefficient, satisfying , predefined by the system or adjusted by historical feedback.

[0099] Based on this, the system generates a comprehensive scoring matrix, which forms the edge weight set of the scheduling graph structure and serves as the input basis for subsequent scheduling model matching. The system uses this scoring matrix as input to construct the enhanced scheduling matching objective function.

[0100] Formula ②: Strengthened scheduling matching objective function

[0101] In order to optimize the overall resource utilization efficiency of the system under multiple constraints, the system constructs an enhanced scheduling optimization model, specifically the following scheduling objective function:

[0102]

[0103] in:

[0104] : The task-node matching set formed by the current scheduling strategy;

[0105] : Scheduling conflict cost function, which measures the degree of scheduling violations such as repeated task allocation and node overload;

[0106] : conflict penalty factor, reflecting the current scheduling stability priority;

[0107] : Comprehensive matching score of the scheduling strategy.

[0108] After executing the above enhanced objective function, the system completes the first round of task scheduling initial matching and forms an initial scheduling result set for use in the next step.

[0109] S220 , performing iterative optimization and constraint rearrangement on the intermediate results of the task scheduling and matching operation to obtain an optimal allocation structure.

[0110] In order to solve the situation of sudden changes in edge node computing power and dynamic communication failure, this step introduces a stability partial derivative mechanism and a system functional reconstruction method to iteratively adjust the initial matching structure generated in S210.

[0111] After obtaining the initial scheduling result outputted by S210 , the system further identifies unstable nodes in the scheduling structure and performs constraint rearrangement optimization.

[0112] Specifically, the system performs the following operations for each matched node: , according to its normalized state vector The computing power, storage and communication status in the comprehensive scoring function Perform partial derivative operation to obtain the scheduling offset index , as shown in formula ③:

[0113] Formula ③: Calculation of node scheduling partial derivatives

[0114]

[0115] in:

[0116] :node Current scheduling deviation risk indicator;

[0117] :node The state vector includes three dimensions: computing power, storage, and communication;

[0118] : The sensitive partial derivative function of the adaptation score to the node state;

[0119] : Vector norm operation, indicating the absolute influence of the partial derivative.

[0120] This partial derivative indicator is used to identify scheduling sensitive nodes. Exceeding the stability threshold When , it is considered as a high-volatility node and needs to be rearranged and replaced. Furthermore, in order to achieve global optimal scheduling stability control, the system constructs a node energy functional function, as defined in formula ④:

[0121] Formula ④: Definition of system scheduling energy functional

[0122]

[0123] in:

[0124] : Scheduling strategy The system energy functional value of

[0125] : Single node scheduling pressure function;

[0126] : total number of nodes;

[0127] : Node offset index calculated by formula ③;

[0128] : Matching score from formula ①.

[0129] Furthermore, the system is based on this functional To minimize the target, while ensuring that the comprehensive score is not lower than the preset threshold ( ). The system replaces high Nodes are matched to tasks until the functional value converges and satisfies the scoring constraints, thus obtaining the optimal matching structure. .

[0130] S230: Generate a matching allocation structure based on the optimal allocation structure.

[0131] After completing the functional optimization iteration of the scheduling matching structure, the system will finally optimize the matching structure Converted into a matching allocation structure. The structure records the task number, node number, final adaptation score, node current state vector and deployment priority score of each task-node pair, which is used to guide subsequent model deployment and communication configuration.

[0132] Formula 5: Model deployment priority score

[0133]

[0134] in:

[0135] : Deployment priority score of task–node pair;

[0136] :Task With node Comprehensive fit score;

[0137] :node Current scheduling deviation risk indicator.

[0138] The matching allocation structure contains the task number, node number, final score, deployment priority score, and state vector information. This structure will serve as the input decision basis for subsequent S300 module compression deployment and will be mapped synchronously with the communication channel priority to ensure the closed-loop and stable operation of the system resource dynamic allocation mechanism.

[0139] Through the design and implementation of the S200 module, this paper constructs an enhanced scheduling optimization mechanism that integrates graph matching, partial derivative detection, and system functional control. Compared with the traditional Hungarian algorithm, this mechanism achieves the following technical performance while ensuring resource adaptation accuracy:

[0140] Optimize the time complexity of task-node scheduling to improve scheduling real-time performance;

[0141] The multi-constraint scoring structure enables joint scheduling and evaluation of computing power, storage, and communication resources;

[0142] The matching allocation structure output can directly serve the subsequent model deployment and communication configuration strategy, providing a mathematical and structural basis for realizing the closed-loop system of "task scheduling-model deployment-data feedback".

[0143] Step S300 at least includes steps S310-S330:

[0144] S310: Obtain the matching allocation structure, compress the central GNN model, and generate a lightweight model.

[0145] First, the matching allocation structure generated in S230 is called. The structure contains the task number, node number, comprehensive score, deployment priority score, and node state vector. The system sorts the task-node pairs according to the deployment priority score and determines the compression processing priority queue.

[0146] Specifically, a knowledge distillation strategy is employed to lightweight the central graph neural network model. The compression process includes the following core steps: first, the central GNN model structure and its pre-trained parameters on the original dataset are obtained; then, an auxiliary teacher-student network structure is constructed, where the teacher model is the complete structure and the student model is a compressed version of the target. Subsequently, the training subset associated with the task number in the matching assignment structure is used as the distillation input data. A joint cross-entropy and KL divergence loss function is used for training, and the student model weight parameters are adjusted to optimize its generalization ability for edge task scenarios.

[0147] During the compression process, the model's number of layers, feature dimensions, and activation structure are dynamically adjusted based on the deployment node's state vector (including computing power, storage, and communication dimensions). The total number of model parameters after compression is strictly controlled to be below a preset storage threshold. After compression, multiple lightweight model instances are generated, each with a unique mapping and binding relationship to the target deployment node. The resulting lightweight model structure and parameters serve as direct input to step S320.

[0148] S320: Deploy the lightweight model to the edge node according to the matching allocation structure to build a model deployment structure.

[0149] The model distribution and deployment operation is completed based on the lightweight model set generated in S310 and the task-node binding information in the matching allocation structure. Specifically, based on the edge node identifier corresponding to each task, the system writes the corresponding lightweight model file to the node's local model directory via a secure transmission channel and initializes the model call interface.

[0150] Deployment relies on the device driver adaptation layer of the edge node. The system pre-parses the node's computing architecture (such as CPU architecture, GPU type, and tensor processing unit) before distribution, automatically selecting and converting model formats (such as ONNX, TensorRT, and TFLite). During deployment, the system performs integrity verification and sandbox loading tests on the transferred model files to ensure error-free invocation and low-latency response on the target edge node.

[0151] The system then constructs a model deployment structure, a data encapsulation designed for distributed environments. This structure includes a model identifier, a unique node identifier, a deployment timestamp, a compression ratio, an API call path, and a status flag. This structure is persistently stored locally on edge nodes and uploaded to the central scheduling platform for subsequent registration tag processing and unified call scheduling management.

[0152] S330: Register and mark the model deployment structure and bind it to a version.

[0153] After the edge node model is deployed, the registration markup and version binding operation of the deployed model is performed. Specifically, the system first calls the model identification code and node identification contained in the deployment structure, establishes a unique registration index for each model-node binding, and generates a registration token.

[0154] The registration process uses a model scheduling registration interface based on the central scheduling system. This interface receives structured information including the model hash digest, deployment path, version number, and resource binding status, and writes it to a central version control database. This database supports version tracking, historical model rollback, and inference performance logging, providing support for subsequent model maintenance.

[0155] Version binding is achieved by introducing a unified model version identification specification. Each lightweight model generates an initial version number after distillation and compression, and a version upgrade mechanism is automatically triggered after each subsequent round of online fine-tuning or policy update. Version upgrade events are synchronously written to the deployment structure and the central scheduling platform, and the model loader automatically loads the corresponding version during inference, ensuring the consistency and stability of inference results.

[0156] Finally, after registration and version binding, the model deployment structure becomes a callable deployment entity, participating in the local inference task execution process in S500. This structure also serves as one of the transmission priority bases in the subsequent S420 communication configuration, achieving multi-dimensional joint mapping with communication resource scheduling.

[0157] Through the implementation of the S300 module, the system achieves the following technical performance and structural support functions:

[0158] Driven by the matching allocation structure, customized knowledge distillation and compression are performed on the central GNN model, balancing model accuracy and resource load control;

[0159] Enables dynamic deployment of lightweight models across heterogeneous edge nodes, supporting automatic format conversion and fast loading under different node architectures;

[0160] Build a structured model deployment structure and complete registration tag and version binding operations to provide structural guarantees for distributed reasoning unified scheduling and version control;

[0161] The output results of the model deployment structure will serve as one of the communication configuration priority bases in the S400 module, while providing a structural interface for the S500 local reasoning and providing intermediate structural support for the system performance closed-loop scheduling mechanism.

[0162] Step S400 at least includes steps S410-S430:

[0163] S410: Acquire the model deployment structure and the communication status information, and select a communication mode according to the communication quality level.

[0164] The system first extracts fields such as the deployment node identifier, deployment timestamp, model compression ratio, interface call path, and running status flag from the model deployment structure output by the S330 module, and binds them to the normalized communication status information from the S110 module. This communication status information includes, but is not limited to, attributes such as communication link bandwidth, average signal-to-noise ratio between nodes, historical communication failure counts, maximum packet loss rate, and current channel occupancy.

[0165] Specifically, the system calculates the communication quality level based on the communication status evaluation results of the deployed nodes. The communication quality level is divided into three categories: high, medium, and low by the central scheduling system according to a standardized indicator grading strategy. For deployed nodes with a high communication quality level, the system prioritizes high-speed and high-frequency transmission modes (such as Wi-Fi 6 or 5G NR mode); for deployed nodes with a medium communication quality level, the system adopts a medium-speed controllable transmission mode (such as traditional 4G LTE or TDMA time-division multiplexing mode); and for deployed nodes with a low communication quality level, the system switches to a reliability-prioritized low-frequency communication mode (such as low-power, long-distance solutions such as LoRa or ZigBee) to ensure the integrity of the transmission task and low packet loss rate.

[0166] The selection of the above communication mode is based not only on the current communication performance of the node, but also on the model compression ratio and call path requirements recorded in the model deployment structure, to ensure that the communication mode meets the minimum requirement thresholds of the corresponding model running tasks in terms of throughput and response delay.

[0167] After the system selects the communication mode, it binds the communication mode identifier with the corresponding node identifier to form a preliminary communication configuration unit, which provides basic input for the subsequent S420 construction of the communication configuration structure.

[0168] S420: Based on the communication mode and the model deployment structure, a communication configuration structure is constructed, and communication channel resources are configured.

[0169] This sub-step builds a communication configuration structure and performs a refined configuration operation of the communication channel resources based on the communication mode result and model deployment structure parameters determined in S410.

[0170] Specifically, the system groups communication modes and establishes communication configuration templates for each. During this configuration, the system calculates the communication time slots, channel allocation priorities, retransmission mechanism parameters, data packet encapsulation format, and target encoding scheme based on the model call frequency, data transmission volume, and estimated inference return period recorded in the model deployment structure. If the communication mode is TDMA, a time slot allocation table is generated based on the model period and the predicted return window. If the communication mode is LoRa, dynamic adjustment calculations are performed for the spreading factor, coding rate, and transmit power.

[0171] At the same time, the system uses historical channel congestion and access frequency data from the communication status information obtained in S110 as communication channel conflict avoidance parameters. Based on these parameters, the system performs interference avoidance optimization on the channel allocation strategy for each edge node within the same communication area. If the system detects the presence of a multi-task centralized deployment node in the current area, its communication channel frequency configuration will prioritize frequency hopping and configure backoff timer parameters to reduce the probability of contention conflicts.

[0172] After the communication configuration structure is constructed, the system generates a communication configuration structure. This structure includes fields such as the deployment node identifier, the bound communication mode identifier, the channel number, the channel frequency band, the time slot start and end times, the packet encapsulation format, the retransmission strategy parameters, and the channel priority tag. This structure is also bound to the model identifier in the model deployment structure, achieving a three-dimensional configuration linkage structure: task, model, and communication.

[0173] S430: Complete the edge node synchronization processing of the communication configuration structure and store the configuration result in the communication configuration table.

[0174] After completing the construction of the communication configuration structure, the system immediately starts the communication configuration synchronization mechanism to execute the distribution of configuration instructions and the status registration operation of the edge nodes.

[0175] Specifically, the system sends each communication configuration structure to the corresponding deployment node in the form of a structured command. After receiving the configuration instructions, the node performs parameter updates and communication interface reloads in accordance with the configuration content. The system deploys a confirmation mechanism module on the node side. After the configuration update is completed, a configuration confirmation signal is generated and transmitted back to the central system as a confirmation that the configuration has taken effect.

[0176] After edge node synchronization is complete, the system writes all communication configuration structures into a unified communication configuration table. This table is a structured communication resource configuration list maintained by the scheduling system, containing a complete record of each task's corresponding model-communication parameter combination. The communication configuration table supports real-time query and dynamic updates, and serves as the basis for transmission path and protocol selection during the subsequent S500 module asynchronous return channel matching process.

[0177] Furthermore, the communication configuration table also records the statistical indicators of communication behavior between edge nodes, including performance parameters such as actual bandwidth occupancy, retransmission ratio, average delay, and error frame ratio, which are used as a reference for the S600 module task resource scoring and feedback update mechanism.

[0178] Complete the parameter binding between the communication configuration structure and the model deployment structure to form an end-to-end transmission control closed loop, and establish a structural dependency foundation for data return of the S500 module and data feedback of the S600 module.

[0179] Step 500 at least includes steps S510-S530:

[0180] S510: Obtain the model deployment structure deployed on each edge node and execute the corresponding local inference task.

[0181] Specifically, the system first retrieves the model deployment structure that has been registered and bound in step S330, which includes the edge node identifier, lightweight model identification code, deployment path, compression ratio, model call interface path and current running status flag.

[0182] Before executing local inference, the system initializes the edge node inference module based on the deployment path and loads the model file that matches the node architecture. The model loading process depends on the edge node's operating environment, including but not limited to the operating system type, available computing units (such as CPU, GPU, TPU), and their driver versions. After successful loading, the system verifies the callability of the model interface to ensure that the model can properly accept input data and output inference results.

[0183] The input data for inference tasks comes from information collected by forest resource monitoring sensors mounted on edge nodes. Specifically, for nodes deployed with video monitoring capabilities, the input data includes image frame sequences of the target monitoring area; for audio collection nodes, the input is noise samples or bird sound clips; and for environmental monitoring tasks, the input is multi-dimensional structured environmental data such as temperature and humidity, light intensity, or PM2.5.

[0184] The model deployment structure pre-configures the model input dimensions and formats. The system formats and converts the raw data into tensors based on these dimensions, performing standardization and normalization steps to generate data tensors that conform to the model input interface. The model inference interface is then called to perform a forward propagation on this input data, yielding preliminary analysis of forest resource status.

[0185] The inference results are returned as a structure, including but not limited to the inference label (such as fire risk level, illegal logging probability, pest and disease indicator), corresponding confidence score, output timestamp, and inference time estimate. This structure is cached in the node's local memory and serves as input data for step S520.

[0186] After the local inference task is completed, the system updates the running status flag field in the model deployment structure and records the inference task number, node ID, model version and execution time to support subsequent version comparison and performance statistics.

[0187] S520: Obtain inference result data of the local inference task, and encode the data according to the communication configuration structure.

[0188] After completing the local inference task, the system retrieves the inference result structure in the local cache of the edge node and calls the communication configuration structure output in S430 to perform encoding adaptation processing. The communication configuration structure contains fields such as the communication mode selected by the node, channel number, allocated time slot, data encapsulation format, retransmission strategy parameters, and channel priority tag.

[0189] First, the system determines the encoding format requirements for the current edge node based on the communication mode identifier. If the communication mode is high-speed and high-frequency (such as Wi-Fi 6), the encoding format can be binary packaging, which offers high compression ratios and low transmission latency. If the communication mode is low-frequency and low-power (such as LoRa), a redundant control encoding scheme is required, adding error correction fields to reduce bit error rates during long-distance transmission.

[0190] During the encoding process, the system maps each field in the inference result structure to the communication protocol's packet field table. It performs hashing or indexing on textual data and fixed-point quantization on numeric data to reduce packet size. Furthermore, it sets a send priority flag for each packet based on the channel priority tag assigned in the communication configuration structure. If the node's communication load exceeds a set threshold, the system places the packet in a delayed send queue and writes it to the configured return window start time.

[0191] Furthermore, if the communication configuration structure enables the packet retransmission mechanism, the system will split the encoded data structure into multiple sub-packets, and each sub-packet will be attached with a sequence number identifier and check redundancy bits for reassembly and error correction recovery at the receiving end.

[0192] After data encoding is complete, a standardized inference data packet structure is generated. The fields include: node ID, task number, inference label, confidence score, data packet number, encapsulation format type, priority flag, whether to enable the retransmission mechanism, etc. This structure will serve as the input object for the asynchronous return processing of S530.

[0193] S530: Asynchronously transmit the inference result data back through the communication configuration structure, and generate a transmission data structure.

[0194] After the data packet structure is encoded, the system calls the communication parameters recorded in the communication configuration table constructed in S430 and starts the asynchronous return module to perform data transmission operations. The asynchronous return module is a multi-threaded asynchronous task scheduling unit with functions such as automatic retransmission, delay queue management, channel conflict detection, and transmission completion confirmation.

[0195] First, the system binds the local communication interface of the edge node based on the channel number and communication frequency band recorded in the communication configuration structure, initializes the communication channel, and applies for bandwidth resources. If the system detects that the channel resources in the communication area where the current node is located are saturated, it automatically adds the task to the delay queue and starts the backoff timer.

[0196] The backhaul process utilizes an asynchronous, non-blocking transmission mechanism. The system writes each inference result packet into the transmission buffer in the order it was subpacketized and monitors the transmission completion flag. If an exception occurs during transmission (such as bandwidth interruption, packet loss, or retransmission failure), the system executes retransmission logic based on the retransmission policy parameters in the communication configuration structure. If the maximum number of retries exceeds the threshold, the transmission event is marked as a failure and logged to the error log for subsequent invocation by the S600 feedback mechanism.

[0197] After all data packets are sent, the system assembles the inference result data into a structured return data structure. The fields include: task number, edge node identifier, model version, inference label, confidence score, communication mode identifier, return success status flag, transmission time statistics, number of retransmissions, channel number, total number of data packets and reception confirmation timestamp, etc.

[0198] After being stored locally, the returned data structure is immediately uploaded to the return buffer queue of the central scheduling system, and the communication status statistical parameters of the corresponding node in the communication configuration table are updated. After the central system obtains the returned data structure, it can perform global feature extraction and data aggregation processing in S600.

[0199] Under the coordinated control of the model deployment structure and the communication configuration structure, local inference operations of edge nodes are completed, supporting the consistency of inference calls in heterogeneous model environments; a standardized return data structure is constructed to achieve data consistency throughout the entire process from model deployment to communication control to data aggregation, providing a trusted input basis for the scoring and policy updates of the S600 module.

[0200] Step S600 at least includes steps S610-S630:

[0201] S610: Obtain the returned data structure, and perform global feature extraction and multi-node data aggregation processing.

[0202] Specifically, the system first batch extracts the return data structure from multiple edge nodes from the return data cache module of the central scheduling platform. The return data structure is asynchronously generated and uploaded in step S530. The fields include task number, edge node identifier, model version, inference label, confidence score, communication mode identifier, return success status flag, transmission time statistics, number of retransmissions, channel number, total number of data packets, and reception confirmation timestamp.

[0203] After acquisition is complete, the system performs global feature extraction. This process extracts risk factors, target states, and abnormal distribution characteristics related to forest resource monitoring based on the inference labels and confidence score fields recorded in the returned data structure. For example, for fire risk tasks, the system aggregates the frequency of fire labels and their corresponding confidence distributions at different nodes to extract the variation pattern of the forest thermal index. For pest and disease monitoring tasks, the consistency index of abnormal plant features identified by multiple nodes in the same target area is calculated.

[0204] During feature extraction, the system simultaneously reads the deployed model version and inference duration fields to determine the adaptability of the current model performance to each node and identify bottlenecks in the model's responsiveness under specific resource configurations. Communication-related fields (such as the number of retransmissions, channel number, and return success status) are also analyzed to assess the stability of the communication path and the transmission reliability of each node.

[0205] After completing global feature extraction, the system categorizes the data returned by all edge nodes by task number and performs multi-node data aggregation. This aggregation process includes, but is not limited to, cross-node feature alignment, redundant data elimination, confidence-weighted fusion, and anomalous result removal. Based on the consistency difference between before and after aggregation, the system automatically identifies nodes with data anomalies or reasoning failures and marks them for subsequent policy control modules for priority adjustment.

[0206] After processing in this step, the system generates an aggregated data structure with fields including task number, aggregated inference label, multi-node weighted confidence, average inference time, average retransmission rate, communication success rate, model version distribution, and abnormal node identification list, providing basic data support for the scoring and strategy generation of the S620 module.

[0207] S620: Based on the multi-node data aggregation processing result, update the task weight strategy and node resource score.

[0208] After obtaining the aggregated data structure output by S610, the system enters the task control and node performance scoring phase. First, the system performs a task importance assessment based on the task dimension. This assessment is based on factors such as the aggregated reasoning confidence level, the consistency of multi-node reasoning results, communication quality level, and model version consistency indicators.

[0209] The system generates a weight label for each task, which reflects the priority level of the task in the next scheduling cycle. The task weight strategy includes fields such as the task number, the current cycle weight score, the difference from the previous cycle, the task type identifier, and the recommended scheduling frequency. If the task results show low multi-node consistency, decreased confidence, or reduced communication stability, the system automatically increases the task weight score to ensure priority matching for subsequent resources. Otherwise, the weight is appropriately reduced to relieve resource allocation pressure.

[0210] At the same time, the system performs resource scoring on edge nodes. The scoring logic is based on multiple metrics, including model execution time, inference accuracy, return success rate, number of retransmissions, and communication channel occupancy. During the scoring process, the system references the edge node status information generated in S110 and the return results recorded in S530 to construct a node performance model.

[0211] Each node generates a node resource score structure, which includes fields such as node identification, execution stability score, communication reliability score, model adaptability score, average inference time, abnormal behavior count, and score update timestamp. Nodes with excellent execution performance and stable communication quality are marked as "priority nodes" and given higher priority in the next round of S210 task scheduling. Nodes that frequently fail, have return latency exceeding a threshold, or experience large fluctuations in inference confidence are marked as "demoted nodes" and are either prioritized for elimination in the next round of task resource mapping or assigned only low-complexity tasks.

[0212] Through this step, the system forms a dynamic scoring result of the task-node dimension and outputs it to the S630 module in a structured manner to drive the parameter update process of the task resource mapping.

[0213] S630: Feedback the task weight strategy and the node resource score to the task resource mapping, and update the task and resource allocation strategy.

[0214] After the task weight strategy and node resource scoring structure are generated, the system immediately initiates the mapping feedback mechanism to complete the parameter writeback and configuration update process. This sub-step uses the task resource mapping constructed in S130 as the basic structure and combines the scoring results output in S620 to reconstruct the original task-node adaptation score and scheduling feasibility status.

[0215] Specifically, the system first maps the weight score field in the task weight policy to the corresponding task entry and updates its priority parameter in the resource mapping table. Simultaneously, the execution stability and communication score fields in the node resource scoring structure are mapped to the corresponding node information, replacing their original status indicator values. This process ensures that the task resource mapping can dynamically adapt based on execution feedback within each scheduling cycle, thus addressing node performance fluctuations and resource mutations in the actual operating environment.

[0216] The system then recalculates the scoring function, rescoring the updated task-node combinations and generating a new task-resource mapping scoring matrix, which serves as the input for the task scheduling model in the next scheduling cycle, S210. To improve the quality of the reconstructed mapping, the system introduces a scoring smoothing mechanism. This compares the range of score fluctuations from the previous round with the weight adjustment trend, preventing dramatic fluctuations in scores caused by transient anomalies that could impact system scheduling stability.

[0217] Finally, the system stores the updated task resource mapping in the central task configuration database and writes it to the mapping version control module to ensure that the task-node mapping version is consistent with the model version and communication configuration structure version. This mapping version number will be used as the retrieval basis for subsequent S210 task scheduling operations, ensuring that the system task scheduling chain has complete configuration traceability capabilities.

[0218] Through the implementation of this step, the system realizes closed-loop control from task execution feedback to configuration structure adjustment, providing support for high robustness and dynamic control capabilities for the global multi-node task scheduling mechanism.

[0219] The present invention achieves the following technical performance optimization and system function closed loop in multi-level forest resource monitoring tasks:

[0220] Establish a standardized extraction mechanism for returned data structures, support cross-node data consistency assessment and multi-dimensional indicator fusion processing, and improve system perception accuracy;

[0221] Build a dual feedback system of task weight strategy and node resource scoring, support dynamic adjustment of task priority and classification of node capabilities, and effectively improve resource scheduling efficiency;

[0222] Implement version-level closed-loop updates of the task-resource mapping structure to ensure that the task-resource relationship structure has the ability to dynamically and adaptively evolve with the operation results, thereby enhancing the overall stability of the system.

[0223] Example 2: Figure 2 FIG. 1 shows a structural block diagram of a multi-level forest resource monitoring system according to an embodiment of the present invention. Figure 2 As shown, the structure may include:

[0224] The task resource construction module 10 is used to obtain edge node status information and forest resource monitoring task sets and construct task resource mapping.

[0225] Obtain computing power, storage, and communication status information of edge nodes;

[0226] Normalize the above information (including unit unification and sliding window standardization);

[0227] Obtain a set of forest resource monitoring tasks and perform structured processing based on parameter labels such as granularity, priority, and complexity;

[0228] Establish a schedulable relationship scoring table between tasks and nodes, and output the task resource mapping structure.

[0229] The task scheduling module 20 is used to perform scheduling optimization based on task resource mapping and generate a matching allocation structure.

[0230] Construct a multi-constraint bipartite graph structure (node-task);

[0231] Introducing a three-dimensional adaptation scoring weight mechanism for computing power, storage, and communication;

[0232] Perform preliminary matching and functional optimization to remove unstable nodes;

[0233] The output contains a matching allocation structure with deployment priority scores, which serves as the basis for compression and deployment of subsequent modules.

[0234] The model compression and deployment module 30 is used to generate a lightweight model according to the matching allocation structure and deploy it to the edge node.

[0235] Perform model compression based on knowledge distillation to control the model size to not exceed the set resource threshold;

[0236] Generate multi-instance lightweight models based on node resource configuration and complete local format adaptation;

[0237] Generate a deployment structure (including model identifier, node number, interface path, and version information);

[0238] Complete registration binding, model index construction and version control table registration.

[0239] The communication configuration module 40 is used to obtain the model deployment structure and communication status information, build the communication configuration structure and complete the channel configuration.

[0240] Select the mode corresponding to high, medium, or low communication level (such as Wi-Fi, TDMA, or LoRa).

[0241] Build a communication configuration structure, configure channel number, time slot strategy, and retransmission mechanism;

[0242] Execute edge node synchronization of communication configuration and write communication configuration table;

[0243] Establish a three-dimensional linkage relationship between task, model and communication parameters to support dynamic channel reconstruction.

[0244] The local reasoning and return module 50 is used to perform local reasoning tasks and return the reasoning results to generate a return data structure.

[0245] Load the corresponding model to perform local inference tasks;

[0246] Standardize, quantify and package monitoring results;

[0247] Call communication configuration for asynchronous data transmission and automatic retransmission;

[0248] Generate a return data structure containing fields such as inference label, confidence, and communication parameters for use by the feedback module.

[0249] The data feedback module 60 is used to generate a task weight strategy and node resource score based on the returned data structure, and update the task resource mapping.

[0250] Perform feature extraction such as reasoning confidence, multi-node consistency, and communication success rate;

[0251] Aggregate task labels and model response indicators to identify abnormal nodes;

[0252] Output task priority adjustment strategy and node capability level label;

[0253] Write the updated content into the task resource mapping structure to drive the next round of task-resource rescheduling.

[0254] Through the integrated design of the above system structure and functional modules, the present invention achieves the following significant beneficial effects in practical applications:

[0255] Full-process closed-loop control: The system builds a full-lifecycle closed-loop structure of "task modeling → optimized scheduling → model deployment → communication configuration → local inference → feedback → mapping update", ensuring that each round of task execution can adaptively evolve based on feedback, improving system operation stability and resource utilization efficiency.

[0256] Optimization of coordinated scheduling of computing power and communication: The S200 module introduces an improved Hungarian algorithm and a stability partial derivative mechanism to achieve global optimal matching of tasks and nodes under multi-dimensional constraints. The S400 module supports heterogeneous communication modes such as TDMA and LoRa, significantly reducing bandwidth usage and supporting large-scale multi-node coordinated data backhaul.

[0257] Edge intelligent distributed reasoning: Through the S300 module, GNN model compression and S500 module edge deployment are implemented to achieve local processing and intelligent reasoning of forest resource data, minimize center load and latency, and meet the timeliness requirements of data analysis in complex forest terrain.

[0258] Adaptive feedback and dynamic reconstruction mechanism: The S600 module has built a scoring feedback mechanism based on tasks and nodes. Combined with the mapping version update mechanism, it ensures that task priorities and resource scheduling can be dynamically adjusted according to the environment, thereby improving system robustness.

[0259] Obviously, the embodiments described above are only some of the embodiments of the present application, rather than all of the embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions described in the aforementioned specific embodiments, or to make equivalent replacements for some of the technical features therein. Any equivalent structure made using the contents of the present application specification and the accompanying drawings, directly or indirectly used in other related technical fields, is also within the scope of patent protection of the present application.

Claims

1. A multi-level forest resource monitoring method, characterized in that: include: Generate edge node status information by normalizing the computing power status information, storage status information, and communication status information of the edge node; The computing power status information includes dynamic sampling of the node's available processing resources, combined with historical task response time and peak scheduling load to construct the computing power index value for the current cycle; the storage status information includes the node's current remaining storage capacity, data cache size, and write rate; the communication status information includes the average signal-to-noise ratio, packet loss rate, average bandwidth occupancy, and the number of recent communication failures with the communication base station; Obtain a set of forest resource monitoring tasks, generate task set information based on task granularity, task priority, and computational complexity, and construct a task resource mapping; the forest resource monitoring task set includes inspection tasks regularly issued by the central forestry management platform, high-priority response tasks triggered by sudden abnormality alarms, and temporary tasks generated based on remote sensing satellite or UAV telemetry analysis; the task granularity is assessed based on the task space scope and sampling density; the task priority is assessed and assigned based on the task source type and response time limit; the computational complexity is assessed and modeled based on the inference time, collected data volume, and processing path structure of similar task models in the past; Based on the task resource mapping, an enhanced scheduling optimization model is constructed and scheduling matching is performed: Comprehensive adaptation score of computing tasks and edge nodes: in, For the task With node Comprehensive fit score; Score the adaptability in the computing power dimension; Score the adaptation in the storage dimension; Score the fit in the communication dimension; is the adaptation weight coefficient; Compute node scheduling deviation risk indicators: in, For nodes Current scheduling deviation risk indicator; For nodes The state vector includes three dimensions: computing power, storage, and communication; is the sensitive partial derivative function of the adaptation score to the node state; is the vector norm operation; Generate a matching assignment structure containing a model deployment priority score, where the expression for the deployment priority score is: ; in, Score the deployment priority of task-node pairs; The matching allocation structure includes task number, node number, final score, deployment priority score and state vector information; Perform knowledge distillation compression on the central GNN model according to the matching allocation structure, generate a lightweight model and deploy it to the edge node, and build a model deployment structure including a model identification code, a node unique identifier, and a compression ratio; Generate a communication configuration structure based on the model deployment structure and complete edge node synchronization; the communication configuration structure process includes: if the communication mode is a TDMA structure, generate a time slot allocation table based on the model period and the predicted return window; if the communication mode is a LoRa structure, perform dynamic adjustment calculations on the spreading factor, coding rate, and transmit power; Execute local inference tasks, generate a return data structure, perform feature extraction and data aggregation on the return data structure, generate task weight strategies and node resource scores, feed back to the task resource mapping and update the configuration strategy.

2. The multi-level forest resource monitoring method according to claim 1, characterized in that: Building an enhanced scheduling optimization model includes: Obtain task resource mapping and build a scheduling optimization model based on edge node status information and task set information; Execute task scheduling matching operations and generate scheduling intermediate results; Perform iterative optimization and constraint rearrangement based on the intermediate scheduling results to update the scheduling results.

3. The multi-level forest resource monitoring method according to claim 1, characterized in that: Generating a matching allocation structure involves: Extract the optimal allocation relationship based on the scheduling results after iterative optimization and build a matching allocation structure; Generate edge node task execution list based on matching allocation structure.

4. The multi-level forest resource monitoring method according to claim 1, characterized in that: Building a model deployment structure includes: Obtain the matching allocation structure, perform compression processing on the central graph neural network model, and generate a lightweight model; Deploy the lightweight model to the edge nodes according to the matching distribution structure and build the model deployment structure; Register the model deployment structure and bind the version information.

5. The multi-level forest resource monitoring method according to claim 1, characterized in that: The process of communicating the configuration structure also includes: Obtain model deployment structure and communication status information, and select a communication mode based on the communication quality level; Build communication configuration structure and allocate communication channel resources; Complete the edge node synchronization processing of the communication configuration structure and store the configuration results in the communication configuration table.

6. The multi-level forest resource monitoring method according to claim 1, characterized in that: Executing local inference tasks includes: Obtain the model deployment structure deployed on the edge node and call the lightweight model to perform local inference tasks; Obtain inference result data and encode the data according to the communication configuration structure; Asynchronous transmission of inference results is performed through the communication configuration structure to generate the transmission data structure.

7. The multi-level forest resource monitoring method according to claim 1, characterized in that: Update configuration strategies include: Obtain the returned data structure, perform global feature extraction and multi-node data aggregation; Generate task weight strategy and node resource score; Feedback the task weight strategy and node resource score to the task resource mapping to complete the strategy update.

8. A multi-level forest resource monitoring system, applied to the multi-level forest resource monitoring method according to any one of claims 1 to 7, characterized in that: include: The task resource construction module is used to obtain edge node status information and forest resource monitoring task sets and build task resource mapping; The task scheduling module is used to perform scheduling optimization based on task resource mapping and generate matching allocation structure; Model compression and deployment module, used to generate lightweight models based on the matching allocation structure and deploy them to edge nodes; The communication configuration module is used to obtain the model deployment structure and communication status information, build the communication configuration structure and complete the channel configuration; The local reasoning and return module is used to perform local reasoning tasks, complete the return of reasoning results, and generate the return data structure; The data feedback module is used to generate task weight strategies and node resource scores based on the returned data structure, and update the task resource mapping.

Images