A multi-dimensional migrant bird activity intelligent monitoring and construction influence early warning method

Abstract

This invention relates to the field of ecological protection and intelligent construction management technology, specifically a multi-dimensional intelligent monitoring and construction impact early warning method for migratory bird activities. The method includes the following steps: multi-source data acquisition and synchronization; data preprocessing and edge computing; spatiotemporal feature fusion and behavioral analysis; construction-migratory bird causal analysis and impact assessment; tiered early warning and construction adjustment; continuous learning and federated optimization. The beneficial effects are: through the coordinated deployment of multiple types of sensors such as high-definition cameras, array microphones, meteorological sensors, and construction progress monitoring equipment, simultaneous acquisition of multi-dimensional data on images, acoustics, environment, and construction is achieved. This overcomes the limitations of existing single-detection methods, comprehensively captures the correlation information between migratory bird activities and the construction environment, and improves the comprehensiveness and accuracy of monitoring data.

Description

Technical Field

[0001] This invention relates to the field of ecological protection and intelligent construction management technology, specifically a multi-dimensional intelligent monitoring method for migratory bird activities and early warning of construction impacts. Background Technology

[0002] In ecologically sensitive areas, construction noise, vibration, and changes in light can significantly disrupt the habitat and migration of migratory birds, thereby damaging the regional ecological balance. Therefore, monitoring migratory bird activities and early warning of construction impacts are particularly important during the construction period.

[0003] Current technologies for migratory bird monitoring primarily rely on manual patrols and single image detection methods, resulting in low monitoring efficiency, poor timeliness, and limited monitoring dimensions. This fails to achieve comprehensive and real-time perception of migratory bird activity. Furthermore, in assessing the impact of construction on migratory birds, existing technologies lack the ability to integrate and analyze multi-dimensional data, only capable of simple correlation judgments. They cannot accurately quantify the degree of impact of construction factors on migratory bird behavior, nor can they provide scientific and specific decision support for construction adjustments.

[0004] In recent years, multimodal sensor technology, deep learning algorithms, edge computing, and causal analysis methods have been widely applied in the field of ecological monitoring. However, most related systems have failed to effectively combine construction influencing factors with migratory bird behavior analysis in real time and achieve dynamic adjustments to construction. Existing technologies often cannot fully identify the complex interaction between construction activities and the ecological environment, and lack intelligent mechanisms that can automatically trigger early warnings and adjust construction plans based on real-time monitoring data, making it difficult to balance project construction progress with ecological protection requirements.

[0005] To address the problems of existing migratory bird monitoring and construction impact early warning technologies, such as single monitoring dimensions, poor real-time performance, inaccurate impact assessment, and lack of dynamic construction adjustment mechanisms, the present invention aims to provide a multi-dimensional intelligent monitoring and construction impact early warning method for migratory bird activities. Summary of the Invention

[0006] The purpose of this invention is to provide a multi-dimensional intelligent monitoring and construction impact early warning method for migratory bird activities, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a multi-dimensional intelligent monitoring method for migratory bird activity and early warning of construction impact, comprising the following steps: Step S1: Multi-source data acquisition and synchronization; Step S2: Data preprocessing and edge computing; The multi-source raw dataset output from step S1 is input into the edge computing node for rapid preprocessing and feature extraction. Image, acoustic, environmental and construction data are processed in a targeted manner and finally fused into a structured migratory bird event stream. Step S3: Spatiotemporal feature fusion and behavior analysis; Using the structured migratory bird event stream output from Step S2 as input, multidimensional feature fusion and behavior analysis are performed using the spatiotemporal Transformer model, and the migratory bird migration trend is predicted by combining the Markov chain model. Step S4: Causal analysis and impact assessment of construction-migratory bird activities; using the spatiotemporal characteristics output in Step S3 as input, a systematic analysis of the causal relationship between construction activities and migratory bird activities is conducted to quantify the degree of impact of construction activities on migratory bird behavior. Step S5: Tiered early warning and construction adjustment; Using the causal analysis and quantitative results from Step S4 as input, establish a multi-level ecological early warning mechanism and automatically generate construction adjustment suggestions based on the early warning level; Step S6: Continuous learning and federated optimization; edge nodes regularly collect actual execution results of construction adjustments and monitoring feedback data of migratory bird activities to build a closed loop for continuous learning and optimization of the system.

[0008] Preferably, the specific steps of multi-source data acquisition and synchronization include: Multiple types of sensors are deployed at the construction site and surrounding ecologically sensitive areas, including high-definition cameras, array microphones, noise meters, weather monitoring instruments, and construction progress data acquisition terminals, and are equipped with GPS and time synchronization modules. High-definition cameras are used to collect images and videos of migratory bird activities, microphones are used to collect bird calls and ambient sounds, noise meters record construction noise levels, weather instruments monitor temperature, humidity, light intensity, and wind speed, and construction progress terminals collect information on equipment start-up and shutdown, work intensity, and work type. All sensors achieve unified clock synchronization via NTP / GPS, and various data are divided into time windows. Grouping and generating multi-source synchronized datasets ; It is an image frame sequence. It is an audio signal. For environmental meteorological and noise data, This provides a unified raw data foundation for construction activities and subsequent processing.

[0009] Preferably, the data preprocessing and edge computing process includes: Image data preprocessing and migratory bird detection: The video stream undergoes brightness normalization and histogram equalization for image enhancement, followed by bilateral filtering for noise reduction. The formula is as follows: , Bird detection using a lightweight detection model, outputting a set of detection boxes. , For the detection frame, For confidence level, Classification; multi-target tracking is performed using Kalman filtering + Hungarian algorithm. Output migratory bird trajectory and direction and speed of movement; Acoustic signal preprocessing and bird call recognition: Pre-emphasis and short-time Fourier transform of acoustic data. ; Calculate the Mel spectrum and MFCC, Birdsong events are identified using a lightweight CNN model, and the probability of each birdong event is output. Acoustic Activity Index ; Environmental and construction signal processing: Weighted moving average smoothing is applied to noise and meteorological data. , The smoothing factor is used; the construction intensity is calculated by weighted summation. , For device weights, Let be the workload of the i-th device at time t; Structured event construction: When the detection meets the threshold condition At that time, structured migratory bird events are generated. , For the number of migratory birds, Characteristics of migratory bird activities, For equivalent sound level, : This is the smoothed environmental meteorological data. For construction strength, Acoustic activity index; calculate correlation score ,like The events are marked as "requiring in-depth analysis" and uploaded to the cloud; the final output is a structured migratory bird event stream with timestamps. .

[0010] Preferably, the spatiotemporal feature fusion and behavior analysis includes the following steps: Spatiotemporal feature construction: Migratory bird activities, acoustic signals, environmental features, and construction features are uniformly mapped onto the same time axis to construct a multi-dimensional spatiotemporal feature input sequence. ; Spatiotemporal Transformer modeling: learns the temporal dependencies of features across various dimensions through an attention mechanism. This is the attention weight matrix. / / To query the key / value matrix, For the embedding dimension, output fused features To depict the spatiotemporal dependence of migratory bird behavior; Behavioral pattern recognition: fusion of features Using K-means clustering , For the set of cluster centers, For the first The cluster centers categorize migratory bird behavior into different patterns, such as normal migration, short stay, and disturbed flight. Migration trend prediction: Construct a Markov state transition matrix based on historical migration data, and calculate the state transition probability. ,pass Predicting the future The study outputs the migratory status changes and activity area trends of migratory birds at each time step, as well as their behavioral categories, state transition probabilities, and response trends to construction factors.

[0011] Preferably, the construction-migratory bird causal analysis and impact assessment includes the following steps: Statistical causality test: using migratory bird activity indicators Construction intensity is the dependent variable. ,noise ,meteorological Construct a Granger causality test model with [variable name] as the independent variable: , These are the autoregressive coefficients. : represents the coefficients of external variables. For the random error term; if the coefficient If the result is significant, then the corresponding construction factor is determined to have a causal effect on migratory bird behavior; Nonlinear deep causal modeling: Introducing deep time series networks , build model By calculating Shapley values ​​or attention weights to generate causal intensity maps, the causal weights of each construction factor are quantified, an impact index is generated, and the quantitative impact results of construction behavior on migratory bird ecological activities are output, thus identifying the main sources of impact.

[0012] Preferably, the tiered early warning and construction adjustment includes the following steps: The four-level early warning system is divided into four levels based on the magnitude of the construction impact index: green: normal, no intervention required; yellow: slight abnormality, mild disturbance; orange: significant disturbance, significant impact; and red: severe impact, ecological risk level four. Automated early warning triggering: The system monitors the impact index in real time. When an orange or red warning is detected, it automatically triggers a higher-level warning and identifies the type of impact source. Construction adjustment suggestions are generated: When the system detects an orange or red alert, it generates suggestions based on the type of impact source, current construction status, and ecological environment parameters, using a function. Automatically generate targeted construction adjustment suggestions, including reducing machinery power, adjusting high-noise operation periods, setting up sound barriers, suppressing construction lighting, and suspending some construction procedures; the early warning information and adjustment suggestions are directly sent to the construction scheduling terminal through the on-site control system, and the results are used as feedback input to step S6 for model optimization, and finally output the early warning level, the type of impact source and the construction adjustment plan; Rapid Response Execution: Construction adjustment suggestions are directly sent to the construction dispatch terminal through the on-site control system to achieve automated or semi-automated response. The final output includes the warning level, the type of impact source, and the construction adjustment plan. This result is used as feedback input for subsequent steps in model optimization.

[0013] Preferably, the continuous learning and federated optimization includes the following steps: Local model training: Each edge computing node uses locally collected feedback data to calculate the gradient of model parameters. The local model is updated and optimized, and only error information and local model update data are uploaded to the central server to ensure data privacy. Gradient upload for global model aggregation: Edge nodes upload the gradient differences and local model update information of their local models to the central server. The server performs a weighted average based on the sample size ni of each node to generate new global model parameters. The update formula is as follows: , The total amount of data across all nodes. For the first Local model parameters for each node; Model distribution and iteration: The central server redistributes the updated global model parameters to each edge computing node. The edge nodes continue to train on local data based on the new global model, continuously improving the model's accuracy and adaptability. Finally, the optimized model parameter set and new threshold strategy are output and applied to steps S2 and S3, enabling the system to form a complete closed loop of data collection, analysis, feedback and relearning, thereby improving the generalization ability for different migratory bird species, geographical environments and construction types.

[0014] Compared with the prior art, the beneficial effects of the present invention are: The multi-dimensional intelligent monitoring and construction impact early warning method proposed in this invention has comprehensive monitoring dimensions and high data accuracy: by coordinating the deployment of multiple types of sensors such as high-definition cameras, array microphones, meteorological sensors, and construction progress monitoring equipment, it realizes the synchronous acquisition of multi-dimensional data of images, acoustics, environment, and construction, breaking through the limitations of existing single detection methods, comprehensively capturing the correlation information between migratory bird activities and the construction environment, and improving the comprehensiveness and accuracy of monitoring data.

[0015] By utilizing edge computing nodes to preprocess and extract features from multi-source data, there is no need to upload large amounts of raw data to the cloud, significantly reducing data transmission latency and enabling real-time monitoring, analysis, and early warning of migratory bird behavior, thus solving the problem of poor timeliness in existing monitoring technologies.

[0016] By integrating statistical causal tests with nonlinear deep causal models, this study achieves a quantitative analysis of the impact of construction activities on migratory bird behavior, clarifies the causal weights and core sources of influence of each construction factor, and provides a more accurate and scientific assessment result compared to the simple correlation judgments of existing technologies, thus providing a reliable basis for decision-making on construction adjustments.

[0017] Establish a standardized four-level ecological early warning mechanism, automatically trigger different levels of early warning based on the construction impact index, and generate targeted construction adjustment suggestions in combination with the type of impact source. This enables automated and semi-automated response of early warning and construction scheduling, and can dynamically coordinate the relationship between construction activities and ecological protection, taking into account both construction progress and ecological requirements.

[0018] By introducing a federated learning mechanism, the system model can be continuously learned and iteratively optimized while ensuring the data privacy of each monitoring station. This enables the system to adapt to the needs of different regions, seasons, migratory bird species, and construction types, significantly improving the model's generalization ability and environmental adaptability.

[0019] The system adopts a six-step closed-loop technical architecture, which integrates data collection, analysis, early warning, decision-making and model optimization into an organic whole. Data flows in a closed loop within the system, and each link feeds back to the others and iterates continuously, ensuring the stability and intelligence of the system and achieving the technical goals of "real-time monitoring, scientific early warning, dynamic adjustment and continuous optimization".

[0020] This invention can effectively reduce the interference of construction activities on the habitat and migration behavior of migratory birds in ecologically sensitive areas. While ensuring the smooth progress of the project, it maximizes the protection of the regional ecological balance and provides important technical support for ecological protection construction. It has both good engineering application value and ecological benefits. Attached Figure Description

[0021] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the present invention clear and complete, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only some, not all, embodiments of the present invention, and are merely illustrative of the embodiments of the present invention. They are not intended to limit the embodiments of the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Please see Figure 1 This invention provides a technical solution: a multi-dimensional intelligent monitoring and construction impact early warning method for migratory bird activities, comprising the following steps: Step S1: Multi-source data acquisition and synchronization Multiple types of sensors are deployed at the construction site and surrounding ecologically sensitive areas, including high-definition cameras (visible light / infrared), array microphones, noise meters, weather monitors, and construction progress data acquisition terminals, along with GPS and time synchronization modules. The cameras are used to collect images and videos of migratory bird activities, the microphones are used to collect bird calls and ambient sounds, the noise meters record construction noise levels, the weather monitors temperature, humidity, light intensity, and wind speed, and the construction progress terminal collects information on equipment start / stop, work intensity, and work type.

[0024] All sensors achieve unified clock synchronization via NTP / GPS, and various data are divided into time windows. Grouping and generating multi-source synchronized datasets ( It is an image frame sequence. It is an audio signal. For environmental meteorological and noise data, (This provides construction activity data) and a unified raw data foundation for subsequent processing.

[0025] Step S2: Data Preprocessing and Edge Computing The multi-source raw dataset output by S1 is input into the edge computing node for rapid preprocessing and feature extraction. Targeted processing is applied to image, acoustic, environmental, and construction data respectively, ultimately fusing them into a structured migratory bird event stream. The specific processing steps are as follows: 1. Image Data Preprocessing and Migratory Bird Detection: The video stream undergoes brightness normalization and histogram equalization for image enhancement, followed by bilateral filtering for noise reduction. The formula is as follows: Migratory bird detection is performed using lightweight detection models (Tiny-YOLO or ViT-lite), and the output detection box set is generated. ( For the detection frame, For confidence level, (Classification); multi-target tracking is performed using Kalman filtering + Hungarian algorithm. Output migratory bird trajectory And the direction and speed of movement.

[0026] 2. Acoustic signal preprocessing and bird call recognition: Pre-emphasis and short-time Fourier transform (STFT) are performed on the acoustic data. ; Calculate the Mel spectrum and MFCC, Birdsong events are identified using a lightweight CNN model, and the probability of each birdong is output. Acoustic Activity Index .

[0027] 3. Environmental and Construction Signal Processing: Weighted moving average smoothing is applied to noise and meteorological data. ( (as a smoothing factor); the construction intensity is calculated by weighted summation. ( For device weights, (where t is the workload of the i-th device).

[0028] 4. Structured event construction: When the detection meets the threshold condition At that time, structured migratory bird events are generated. ( For the number of migratory birds, Characteristics of migratory bird activities, For equivalent sound level, : This is the smoothed environmental meteorological data. For construction strength, (Acoustic activity index); calculate correlation score. ,like The events are marked as "requiring in-depth analysis" and uploaded to the cloud; the final output is a structured migratory bird event stream with timestamps. .

[0029] Step S3: Spatiotemporal Feature Fusion and Behavior Analysis Using the structured migratory bird event stream output by S2 as input, a spatiotemporal Transformer model is used for multidimensional feature fusion and behavioral analysis, and a Markov chain model is combined to predict migratory bird trends. Specific steps include: 1. Spatiotemporal Feature Construction: Migratory bird activities (location, quantity, speed), acoustic signals (MFCC, AI index), environmental features (temperature, humidity, noise), and construction features (construction intensity) are uniformly mapped onto the same time axis to construct a multi-dimensional spatiotemporal feature input sequence. .

[0030] 2. Spatiotemporal Transformer Modeling: Learning the temporal dependencies of features across various dimensions through an attention mechanism. ( This is the attention weight matrix. / / To query the key / value matrix, (for the embedding dimension), output fused features This depicts the spatiotemporal dependence of migratory bird behavior.

[0031] 3. Behavioral pattern recognition: fusion of features Using K-means clustering, ( For the set of cluster centers, For the first (Each cluster center) divides migratory bird behavior into different patterns such as normal migration, short stay, and disturbed flight.

[0032] 4. Migration Trend Prediction: Construct a Markov state transition matrix based on historical migration data, and calculate the state transition probability. ,pass Predicting the future The study outputs the migratory status changes and activity area trends of migratory birds at each time step, as well as their behavioral categories, state transition probabilities, and response trends to construction factors.

[0033] Step S4: Construction-Migratory Bird Causal Analysis and Impact Assessment Using the spatiotemporal characteristics output by S3 as input, a systematic analysis of the causal relationship between construction activities and migratory bird activities is conducted to quantify the degree of impact of construction activities on migratory bird behavior, specifically including: 1. Statistical causality test: using migratory bird activity indicators Construction intensity is the dependent variable. ,noise ,meteorological Construct a Granger causality test model with [variable name] as the independent variable: ( These are the autoregressive coefficients. : represents the coefficients of external variables. (for random error terms); if the coefficient If the result is significant, then the corresponding construction factor is determined to have a causal effect on migratory bird behavior.

[0034] 2. Nonlinear Deep Causal Modeling: Introducing Deep Time Series Networks , build model By calculating Shapley values ​​or attention weights to generate causal intensity maps, the causal weights of each construction factor are quantified, an impact index is generated, and the quantitative impact results of construction behavior on migratory bird ecological activities are output, thus identifying the main sources of impact.

[0035] Step S5: Tiered Early Warning and Construction Adjustment Using the causal analysis and quantitative results of S4 as input, a multi-level ecological early warning mechanism is established, and construction adjustment suggestions are automatically generated based on the early warning level. Specifically, the following is implemented: 1. Four-level warning classification: Based on the magnitude of the construction impact index, the warning level is divided into four levels: green (normal, no intervention required), yellow (minor abnormality, slight disturbance), orange (significant disturbance, significant impact), and red (severe impact, ecological risk); 2. Automated early warning triggering: The system monitors the impact index in real time. When an orange or red warning is detected, it automatically triggers a higher-level warning and identifies the type of impact source. 3. Construction Adjustment Suggestion Generation: When the system detects an orange or red alert, it generates construction adjustment suggestions based on the type of impact source, current construction status, and ecological environment parameters, using a function. The system automatically generates targeted construction adjustment suggestions, including reducing machinery power, adjusting high-noise operation periods, setting up sound barriers, suppressing construction lighting, and suspending some construction procedures. The early warning information and adjustment suggestions are directly sent to the construction dispatch terminal through the on-site control system. At the same time, the results are used as feedback input to S6 for model optimization. Finally, the system outputs the early warning level, the type of impact source, and the construction adjustment plan.

[0036] 4. Rapid Response Execution: Construction adjustment suggestions are directly issued to the construction dispatch terminal through the on-site control system to achieve automated or semi-automated response. The final output includes the warning level, the type of impact source, and the construction adjustment plan. This result is used as feedback input for subsequent steps in model optimization.

[0037] Step S6: Continuous Learning and Federated Optimization Edge nodes periodically collect actual implementation results of construction adjustments and monitoring feedback data on migratory bird activities to build a continuous learning and optimization loop for the system. The specific process is as follows: 1. Local Model Training: Each edge computing node uses locally collected feedback data to calculate the gradient of model parameters. The local model is updated and optimized, and only error information and local model update data are uploaded to the central server to ensure data privacy. 2. Gradient Upload and Global Model Aggregation: Edge nodes upload the gradient differences and local model update information of their local models to the central server. The server performs a weighted average based on the sample size ni of each node to generate new global model parameters. The update formula is as follows: ( The total amount of data across all nodes. For the first Local model parameters for each node.

[0038] 3. Model Distribution and Iteration: The central server redistributes the updated global model parameters to each edge computing node. The edge nodes continue to train on local data based on the new global model, continuously improving the model's accuracy and adaptability. Finally, the optimized model parameter set and new threshold strategy are output and applied to steps S2 and S3, enabling the system to form a complete closed loop of data collection, analysis, feedback, and relearning, thereby improving its generalization ability for different migratory bird species, geographical environments, and construction types.

[0039] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-dimensional migratory activity intelligent monitoring and construction impact early warning method, characterized in that: Includes the following steps: Step S1: Multi-source data acquisition and synchronization; Step S2: Data preprocessing and edge computing; The multi-source raw dataset output from step S1 is input into the edge computing node for rapid preprocessing and feature extraction. Image, acoustic, environmental and construction data are processed in a targeted manner and finally fused into a structured migratory bird event stream. Step S3: Spatiotemporal feature fusion and behavior analysis; Using the structured migratory bird event stream output from Step S2 as input, multidimensional feature fusion and behavior analysis are performed using the spatiotemporal Transformer model, and the migratory bird migration trend is predicted by combining the Markov chain model. Step S4: Causal analysis and impact assessment of construction-migratory bird activities; using the spatiotemporal characteristics output in Step S3 as input, a systematic analysis of the causal relationship between construction activities and migratory bird activities is conducted to quantify the degree of impact of construction activities on migratory bird behavior. Step S5: Tiered early warning and construction adjustment; Using the causal analysis and quantitative results from Step S4 as input, establish a multi-level ecological early warning mechanism and automatically generate construction adjustment suggestions based on the early warning level; Step S6: Continuous learning and federated optimization; edge nodes regularly collect actual execution results of construction adjustments and monitoring feedback data of migratory bird activities to build a closed loop for continuous learning and optimization of the system.

2. The method for intelligent monitoring of multi-dimensional migratory bird activities and early warning of construction impact according to claim 1, characterized in that: The specific steps for multi-source data acquisition and synchronization include: Multiple types of sensors are deployed at the construction site and surrounding ecologically sensitive areas, including high-definition cameras, array microphones, noise meters, weather monitoring instruments, and construction progress data acquisition terminals, and are equipped with GPS and time synchronization modules. High-definition cameras are used to collect images and videos of migratory bird activities, microphones are used to collect bird calls and ambient sounds, noise meters record construction noise levels, weather instruments monitor temperature, humidity, light intensity, and wind speed, and construction progress terminals collect information on equipment start-up and shutdown, work intensity, and work type. All sensors achieve unified clock synchronization via NTP / GPS, and various data are divided into time windows. Grouping and generating multi-source synchronized datasets ; It is an image frame sequence. It is an audio signal. For environmental meteorological and noise data, This provides a unified raw data foundation for construction activities and subsequent processing.

3. The method for intelligent monitoring of multi-dimensional migratory bird activities and early warning of construction impact according to claim 2, characterized in that: The data preprocessing and edge computing process includes: Image data preprocessing and migratory bird detection: The video stream undergoes brightness normalization and histogram equalization for image enhancement, followed by bilateral filtering for noise reduction. The formula is as follows: ， Bird detection using a lightweight detection model, outputting a set of detection boxes. , For the detection frame, For confidence level, Classification; multi-target tracking is performed using Kalman filtering + Hungarian algorithm. Output migratory bird trajectory and direction and speed of movement; Acoustic signal preprocessing and bird call recognition: Pre-emphasis and short-time Fourier transform of acoustic data. ; Calculate the Mel spectrum and MFCC, Birdsong events are identified using a lightweight CNN model, and the probability of each birdong is output. Acoustic Activity Index ; Environmental and construction signal processing: Weighted moving average smoothing is applied to noise and meteorological data. , The smoothing factor is used; the construction intensity is calculated by weighted summation. , For device weights, Let be the workload of the i-th device at time t; Structured event construction: When the detection meets the threshold condition At that time, structured migratory bird events are generated. , For the number of migratory birds, Characteristics of migratory bird activities, For equivalent sound level, : This is the smoothed environmental meteorological data. For construction strength, Acoustic activity index; calculate correlation score ,like The events are marked as "requiring in-depth analysis" and uploaded to the cloud; the final output is a structured migratory bird event stream with timestamps. .

4. The method for intelligent monitoring of multi-dimensional migratory bird activities and early warning of construction impact according to claim 1, characterized in that: The spatiotemporal feature fusion and behavior analysis Includes the following steps: Spatiotemporal feature construction: Migratory bird activities, acoustic signals, environmental features, and construction features are uniformly mapped onto the same time axis to construct a multi-dimensional spatiotemporal feature input sequence. ; Spatiotemporal Transformer modeling: learning the temporal dependencies of features across various dimensions through an attention mechanism. Here is the attention weight matrix. / / To query the key / value matrix, For the embedding dimension, output fused features To depict the spatiotemporal dependence of migratory bird behavior; Behavioral pattern recognition: fusion of features Using K-means clustering , For the set of cluster centers, For the first The cluster centers categorize migratory bird behavior into different patterns, such as normal migration, short stay, and disturbed flight. Migration trend prediction: Construct a Markov state transition matrix based on historical migration data, and calculate the state transition probability. ,pass Predicting the future The study outputs the migratory status changes and activity area trends of migratory birds at each time step, as well as their behavioral categories, state transition probabilities, and response trends to construction factors.

5. The method for intelligent monitoring of multi-dimensional migratory bird activities and early warning of construction impact according to claim 1, characterized in that: The construction-migratory bird causal analysis and impact assessment includes the following steps: Statistical causality test: using migratory bird activity indicators Construction intensity is the dependent variable. ,noise ,meteorological Construct a Granger causality test model with [variable name] as the independent variable: , These are the autoregressive coefficients. : represents the coefficients of external variables. For the random error term; if the coefficient If the result is significant, then the corresponding construction factor is determined to have a causal effect on migratory bird behavior; Nonlinear deep causal modeling: Introducing deep time series networks , build model By calculating Shapley values ​​or attention weights to generate causal intensity maps, the causal weights of each construction factor are quantified, an impact index is generated, and the quantitative impact results of construction behavior on migratory bird ecological activities are output, thus identifying the main sources of impact.

6. The method for intelligent monitoring of multi-dimensional migratory bird activities and early warning of construction impact according to claim 1, characterized in that: The tiered early warning and construction adjustment include the following steps: The four-level early warning system is divided into four levels based on the magnitude of the construction impact index: green: normal, no intervention required; yellow: slight abnormality, mild disturbance; orange: significant disturbance, significant impact; and red: severe impact, ecological risk level four. Automated early warning triggering: The system monitors the impact index in real time. When an orange or red warning is detected, it automatically triggers a higher-level warning and identifies the type of impact source. Construction adjustment suggestions are generated: When the system detects an orange or red alert, it generates suggestions based on the type of impact source, current construction status, and ecological environment parameters, using a function. Automatically generate targeted construction adjustment suggestions, including reducing machinery power, adjusting high-noise operation periods, setting up sound barriers, suppressing construction lighting, and suspending some construction procedures; the early warning information and adjustment suggestions are directly sent to the construction scheduling terminal through the on-site control system, and the results are used as feedback input to step S6 for model optimization, and finally output the early warning level, the type of impact source and the construction adjustment plan; Rapid Response Execution: Construction adjustment suggestions are directly issued to the construction dispatch terminal through the on-site control system to achieve automated or semi-automated response. The final output includes the warning level, the type of impact source, and the construction adjustment plan. This result is used as feedback input for subsequent steps in model optimization.

7. The method for intelligent monitoring of multi-dimensional migratory bird activities and early warning of construction impact according to claim 1, characterized in that: The continuous learning and federated optimization includes the following steps: Local model training: Each edge computing node uses locally collected feedback data to calculate the gradient of model parameters. The local model is updated and optimized, and only error information and local model update data are uploaded to the central server to ensure data privacy. Gradient upload for global model aggregation: Edge nodes upload the gradient differences and local model update information of their local models to the central server. The server performs a weighted average based on the sample size ni of each node to generate new global model parameters. The update formula is as follows: , The total amount of data across all nodes. For the first Local model parameters for each node; Model distribution and iteration: The central server redistributes the updated global model parameters to each edge computing node. The edge nodes continue to train on local data based on the new global model, continuously improving the model's accuracy and adaptability. Finally, the optimized model parameter set and new threshold strategy are output and applied to steps S2 and S3, enabling the system to form a complete closed loop of data collection, analysis, feedback and relearning, thereby improving the generalization ability for different migratory bird species, geographical environments and construction types.

Images