Internet of Things monitoring and early warning method, system and equipment for water conservancy facilities and medium

Abstract

The invention relates to an Internet of Things monitoring and early warning method, system and device for a water conservancy facility and a medium, and the method comprises the steps: obtaining data collected by various types of sensors disposed on the water conservancy facility, and carrying out the preliminary classification processing, and obtaining the preliminary monitoring data and image information; performing target detection and recognition on the image information through a preset E-YOLO algorithm to obtain an image recognition result; performing comprehensive processing on the preliminary monitoring data and the image information to obtain comprehensive monitoring data; performing fusion analysis on the image recognition result and the comprehensive monitoring data to obtain a comprehensive state analysis result of the water conservancy facility; and performing early warning analysis on the comprehensive state analysis result according to a preset early warning threshold to obtain initial early warning information. The system can adapt to different types of water conservancy facility characteristics and actual demand changes.

Description

Technical Field

[0001] The present invention relates to the technical field of water conservancy facilities, and in particular to an Internet of Things monitoring and early warning method, system, equipment and medium for water conservancy facilities. Background Art

[0002] The safe and stable operation of water conservancy facilities is of great significance for flood control, flood prevention and water supply. With the development of Internet of Things technology, the use of sensor networks for real-time monitoring and early warning of water conservancy facilities has become an advanced solution. However, traditional methods usually rely on only a single type of sensor data, ignoring the importance of multi-source data fusion, resulting in insufficient comprehensiveness and accuracy of monitoring data. Secondly, in terms of image monitoring, existing methods lack efficient target detection and recognition technology, and cannot fully utilize image information for accurate status judgment. Finally, existing early warning mechanisms are often based on fixed thresholds, lack the ability to dynamically adjust, and cannot adapt to complex and changing actual situations, which may lead to lags or false alarms in early warning information. Summary of the invention

[0003] The main purpose of the present invention is to provide an Internet of Things monitoring and early warning method, system, equipment and medium for water conservancy facilities, which can flexibly adjust the generation strategy of early warning information according to the characteristics of different types of water conservancy facilities and actual demand changes.

[0004] To achieve the above object, the present invention provides an Internet of Things monitoring and early warning method for water conservancy facilities, comprising: Acquire data collected by various types of sensors arranged on water conservancy facilities, and perform preliminary classification and processing to obtain preliminary monitoring data and image information; Performing target detection and recognition on the image information through a preset E-YOLO algorithm to obtain an image recognition result; Comprehensively processing the preliminary monitoring data and image information to obtain comprehensive monitoring data; The image recognition result is integrated with the comprehensive monitoring data for analysis to obtain a comprehensive status analysis result of the water conservancy facilities; Perform a warning analysis on the comprehensive status analysis result according to a preset warning threshold to obtain initial warning information.

[0005] Furthermore, the data collected by various types of sensors arranged on water conservancy facilities are obtained, and preliminary classification and processing are performed to obtain preliminary monitoring data and image information, including: Preliminarily screening and processing the data collected by the multiple types of sensors arranged at each target position of the water conservancy facility to obtain effective monitoring data; The effective monitoring data are sorted by type to obtain environmental monitoring data, structural stress monitoring data, flow monitoring data and temperature monitoring data; Acquire original image data collected by an image sensor arranged on the water conservancy facility, and perform pre-standard processing to obtain standard image data; The environmental monitoring data, the structural stress monitoring data, the flow monitoring data and the temperature monitoring data are subjected to data fusion processing, and the data are aligned according to timestamps and spatial positions to obtain preliminary monitoring data; Extracting feature data based on the preliminary monitoring data to obtain a target feature data set; The target feature data set is subjected to image feature fusion processing with the standard image data to obtain the image information.

[0006] Furthermore, the target detection and recognition is performed on the image information by using a preset E-YOLO algorithm to obtain an image recognition result, including: Extracting image features from the image information to obtain image feature data; The image feature data is input into the E-YOLO algorithm for target detection to obtain the detected target frame, including: Dividing the image feature data into an S×S grid to obtain a plurality of grid units; Performing multiple bounding box predictions on each of the grid cells to obtain a three-dimensional bounding box including position coordinates, width, and height; Calculating a confidence score of the existence of an object for each of the bounding boxes according to the three-dimensional bounding box to obtain a bounding box confidence score; Calculate the conditional probabilities of multiple categories for each grid unit to obtain the category probability; Calculating a confidence score for each of the bounding boxes according to the bounding box confidence and the category probability to obtain a final confidence score; According to a preset non-maximum suppression rule, the final confidence level is sorted and screened for overlapping frames to obtain a target detection frame; Classifying the target detection frame and performing target recognition to obtain a corresponding target recognition result; Annotating the image information according to the target recognition result to obtain annotated image information; The target recognition result is fused with the annotated image information to obtain the image recognition result.

[0007] Furthermore, the conditional probability calculation of multiple categories is performed on each grid unit to obtain the category probability, including: Assigning predefined category labels to the grid units to obtain a target category set; Performing vectorized encoding processing on the target category set to obtain a category feature mapping matrix; Performing function calculation on the category feature mapping matrix to obtain an initial category probability value; Performing cross entropy loss calculation on the category feature mapping matrix according to the initial category probability value to obtain a category loss value; Performing back propagation optimization processing on the category loss value to obtain optimized category features; Performing correction calculation on the initial category probability value according to the optimized category feature to obtain a corrected probability value; Performing threshold screening processing on the corrected probability value to obtain a candidate category probability set; The maximum probability category is selected according to the candidate category probability set to obtain the category probability.

[0008] Furthermore, the preliminary monitoring data and image information are comprehensively processed to obtain comprehensive monitoring data, including: Performing periodic abnormal change analysis on the preliminary monitoring data to obtain periodic sequence analysis results; Extracting correlation features from the preliminary monitoring data according to the periodic sequence analysis results to obtain a monitoring correlation feature set; Fusing the monitoring-related feature set with the image information to obtain preliminary comprehensive data; Performing anomaly identification on the preliminary comprehensive data according to a preset anomaly detection rule to obtain anomaly detection data; The preliminary comprehensive data is corrected and adjusted according to the abnormal detection data to obtain the comprehensive monitoring data.

[0009] Furthermore, the image recognition result is integrated and analyzed with the comprehensive monitoring data to obtain a comprehensive status analysis result of the water conservancy facilities, including: Performing feature extraction processing on the image recognition result to obtain image feature data; Performing multi-dimensional feature extraction processing on the comprehensive monitoring data to obtain comprehensive monitoring feature data; Performing weighted fusion processing on the image feature data and the comprehensive monitoring feature data according to a preset weight factor to obtain fused feature data; Performing correlation analysis on the image recognition result and the comprehensive monitoring data according to the fusion feature data to obtain a correlation analysis result; A comprehensive evaluation of the operating status of the water conservancy facilities is performed based on the correlation analysis results to obtain a comprehensive status analysis result.

[0010] Furthermore, the early warning analysis is performed on the comprehensive status analysis result according to the preset early warning threshold to obtain initial early warning information, including: Performing sub-state segmentation on the comprehensive state analysis result to obtain a plurality of sub-state data; Performing threshold judgment processing on each of the sub-state data according to the early warning threshold to obtain a threshold judgment result; Aggregating the threshold judgment results to obtain an aggregate judgment result; Performing risk analysis on the aggregated judgment results according to preset early warning risk rules to obtain initial risk data; Performing risk level classification processing on the comprehensive status analysis result according to the aggregate judgment result to obtain initial risk level data; Performing early warning information analysis on the initial risk data and the initial risk level data to obtain the initial early warning information; Acquire the historical status analysis result of the water conservancy facility from a preset historical database, and perform data extraction on the historical status analysis result to obtain historical status data; Performing indicator difference calculation on the comprehensive status analysis result according to preset monitoring indicators to obtain preliminary difference data; Performing deviation analysis on the comprehensive state analysis result according to the historical state data to obtain deviation characteristic data; Performing correction analysis on the preliminary difference data and the deviation characteristic data to obtain correction warning information; The initial warning information is dynamically adjusted according to the revised warning information to obtain final warning information.

[0011] The present invention also provides an Internet of Things monitoring and early warning system for water conservancy facilities, characterized in that the Internet of Things monitoring and early warning method for water conservancy facilities applied to any one of the above items comprises: A collection module, which is used to obtain data collected by various types of sensors arranged on water conservancy facilities, and perform preliminary classification processing to obtain preliminary monitoring data and image information; An analysis module, wherein the analysis module is used to perform target detection and recognition on the image information through a preset E-YOLO algorithm to obtain an image recognition result; A correlation module, wherein the correlation module is used to comprehensively process the preliminary monitoring data and the image information to obtain comprehensive monitoring data; A processing module, the processing module is used to fuse and analyze the image recognition result with the comprehensive monitoring data to obtain a comprehensive status analysis result of the water conservancy facility; A control module is used to perform early warning analysis on the comprehensive status analysis result according to a preset early warning threshold to obtain initial early warning information.

[0012] The present invention also provides an Internet of Things monitoring and early warning device for water conservancy facilities, comprising: Memory, used to store programs; The processor is used to execute the program to implement each step of the Internet of Things monitoring and early warning method for water conservancy facilities described in any one of the above.

[0013] The present invention also provides a storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute any of the above methods.

[0014] The present invention provides an Internet of Things monitoring and early warning method, system, device and medium for water conservancy facilities, which has the following beneficial effects: By preliminarily classifying and processing the various sensor data of water conservancy facilities, and combining the E-YOLO algorithm to detect and identify the target of the image information, we can more accurately obtain the various status information of water conservancy facilities. This method improves the comprehensiveness and accuracy of monitoring data and provides a more reliable basis for the safety management of water conservancy facilities. By comprehensively processing sensor data and image information, and performing deviation analysis on the comprehensive state based on the preset warning threshold and historical data, dynamic monitoring of the status of water conservancy facilities is achieved, which helps to timely discover potential safety hazards and avoid misjudgment or omission due to single data. Based on the fusion analysis of image recognition results and sensor comprehensive monitoring data, it can ensure that the system can operate efficiently under various environmental conditions, reduce false alarms and omissions of warning information, and improve the accuracy and timeliness of warnings. By dynamically adjusting the warning information, a more reasonable warning strategy is formulated, and the optimized operation of the entire monitoring system is achieved through a comprehensive adjustment strategy, thereby effectively improving the safety protection capabilities of water conservancy facilities and reducing the risks caused by environmental changes or facility aging. By combining historical status analysis results with deviation analysis, the early warning information generation strategy can be flexibly adjusted according to the characteristics of different types of water conservancy facilities and actual demand changes, making the monitoring and early warning system more adaptable to diverse application scenarios, realizing refined management of water conservancy facilities, and effectively improving the comprehensive operation efficiency and safety assurance level of water conservancy facilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a flow chart of an Internet of Things monitoring and early warning method for water conservancy facilities provided by the present invention; Figure 2 This is a structural diagram of an Internet of Things monitoring and early warning system for water conservancy facilities provided by the present invention; Figure 3This is a structural diagram of an Internet of Things monitoring and early warning device for water conservancy facilities provided by the present invention.

[0016] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] The present invention is further described below in conjunction with the accompanying drawings and specific implementation methods.

[0019] Reference Figure 1 As shown, the present invention provides an Internet of Things monitoring and early warning method for water conservancy facilities, comprising: Step S1: acquiring data collected by various types of sensors arranged on water conservancy facilities, and performing preliminary classification processing to obtain preliminary monitoring data and image information; Step S2: Detect and identify the target on the image information through the preset E-YOLO algorithm to obtain the image recognition result; Step S3: Comprehensively process the preliminary monitoring data and image information to obtain comprehensive monitoring data; Step S4: fusing and analyzing the image recognition results with the comprehensive monitoring data to obtain comprehensive status analysis results of the water conservancy facilities; Step S5: Perform early warning analysis on the comprehensive status analysis result according to the preset early warning threshold to obtain initial early warning information.

[0020] Based on the above steps, the detailed process is as follows: Step S1: Various types of sensors are arranged at key target locations of water conservancy facilities, such as pressure sensors, temperature sensors, humidity sensors, vibration sensors, etc. These sensors are responsible for real-time monitoring of the operating status and environmental conditions of water conservancy facilities. The sensors collect and transmit data in real time.

[0021] The collected data are preliminarily classified according to their source and type, such as pressure data, temperature data, vibration data, etc., to form preliminary monitoring data.

[0022] The preliminary monitoring data is structured data after classification and processing, including the readings of each sensor and its preliminary analysis results. The image information comes from the cameras installed around the water conservancy facilities, capturing the appearance of the facilities and the surrounding environment.

[0023] Step S2: E-YOLO is an improved target detection algorithm that combines the real-time and high efficiency of the YOLO algorithm and is suitable for target recognition in complex backgrounds.

[0024] The E-YOLO algorithm can quickly identify multiple targets in an image, classify them, and locate them. The preprocessed image information is extracted through the E-YOLO algorithm, and the target is located on the feature map. The E-YOLO algorithm can identify multiple targets in an image, such as cracks, obstructions, mechanical parts of water conservancy facilities, etc., and mark their locations and categories.

[0025] The recognition result includes the target’s category, location (bounding box coordinates), and confidence score. The confidence score is used to measure the reliability of the recognition result. The generated recognition result is output in the form of structured data, which is convenient for fusion analysis with other monitoring data.

[0026] Step S3: Preliminary monitoring data and image information provide status information of water conservancy facilities in different dimensions. In order to comprehensively evaluate the operating status of the facilities, it is necessary to process these two types of data in a comprehensive manner. Data fusion helps to make up for the shortcomings of a single data source and improve the overall accuracy and reliability of monitoring information. During the data fusion process, data from different sources need to be aligned in time and space to ensure the timeliness and consistency of the fusion results.

[0027] Comprehensive monitoring data contains multi-dimensional information of water conservancy facilities, such as structural integrity, environmental impact factors, equipment operating status, etc.

[0028] Step S4: The main goal of fusing the image recognition results with the comprehensive monitoring data is to obtain comprehensive status information of water conservancy facilities. By analyzing the association between the image recognition results and the comprehensive monitoring data, the potential relationship between the data is discovered. For example, there may be a correlation between the image of the crack and the abnormal fluctuation of the vibration data. Using multi-dimensional data fusion technology, different types of data are integrated according to the time and space dimensions to form a unified analysis data set.

[0029] The comprehensive status analysis results are a comprehensive assessment of the current operating status of water conservancy facilities, including the health status of the facilities, potential risks and identification of abnormal events. The results are presented in the form of reports or charts, including key indicators of the facilities, abnormal alarms, risk assessments, etc., to facilitate decision-making and management by relevant personnel.

[0030] Step S5: Set warning thresholds based on the operating characteristics and historical data of the water conservancy facilities. These thresholds can be the critical value of a single indicator or a combination of multiple indicators. For example, pressure exceeding a certain value or abnormal temperature fluctuations may trigger a warning.

[0031] Compare the comprehensive status analysis results with the preset warning thresholds to identify abnormal situations that exceed the thresholds. Warning analysis can be performed using rule engines or model-based methods. For example, rule engines can make quick judgments based on pre-defined rules, while model-based methods can use historical data for trend prediction and anomaly detection.

[0032] The initial warning information includes the type, severity, time of occurrence and possible causes of the warning. For example, the warning type can be structural damage, equipment failure, environmental change, etc.; the severity can be divided into high, medium, low and other different levels.

[0033] The present invention provides an Internet of Things monitoring and early warning method for water conservancy facilities. By preliminarily classifying and processing a variety of sensor data of water conservancy facilities, and combining the E-YOLO algorithm to detect and identify the target of image information, we can more accurately obtain various status information of water conservancy facilities. This method improves the comprehensiveness and accuracy of monitoring data and provides a more reliable basis for the safety management of water conservancy facilities. By comprehensively processing sensor data and image information, and performing deviation analysis on the comprehensive state according to the preset early warning threshold and historical data, dynamic monitoring of the state of water conservancy facilities is achieved, which helps to timely discover potential safety hazards and avoid misjudgment or missed judgment due to single data. Based on the fusion analysis of image recognition results and sensor comprehensive monitoring data, it can ensure that the system can operate efficiently under various environmental conditions, reduce false alarms and missed alarms of early warning information, and improve the accuracy and timeliness of early warning. By dynamically adjusting the early warning information, a more reasonable early warning strategy is formulated, and the optimized operation of the entire monitoring system is achieved through a comprehensive adjustment strategy, thereby effectively improving the safety protection capability of water conservancy facilities and reducing the risks caused by environmental changes or aging of facilities. By combining historical status analysis results with deviation analysis, the early warning information generation strategy can be flexibly adjusted according to the characteristics of different types of water conservancy facilities and actual demand changes, making the monitoring and early warning system more adaptable to diverse application scenarios, realizing refined management of water conservancy facilities, and effectively improving the comprehensive operation efficiency and safety assurance level of water conservancy facilities.

[0034] In one embodiment, data collected by various types of sensors arranged on water conservancy facilities are obtained, and preliminary classification processing is performed to obtain preliminary monitoring data and image information, including: Various types of sensors are placed at various target locations of water conservancy facilities, including environmental monitoring sensors, structural stress monitoring sensors, flow monitoring sensors and temperature monitoring sensors. The data collected by these sensors are initially screened and processed to remove invalid or abnormal data and retain valid monitoring data.

[0035] The effective monitoring data were sorted by type to obtain environmental monitoring data, structural stress monitoring data, flow monitoring data and temperature monitoring data.

[0036] Image sensors placed on water conservancy facilities collect raw image data, which are processed in advance to remove noise, adjust brightness and contrast, and form standard image data.

[0037] The environmental monitoring data, structural stress monitoring data, flow monitoring data and temperature monitoring data are fused and processed, and the data are aligned through timestamps and spatial positions to ensure that the monitoring data at different times and locations can be analyzed uniformly. The data alignment process is achieved by matching the timestamps and sensor positions of the data, and multiple monitoring data at the same moment and location are comprehensively processed to obtain preliminary monitoring data.

[0038] Based on the preliminary monitoring data, feature data extraction is performed to extract key feature data reflecting the status of water conservancy facilities. During the feature data extraction process, based on preset rules and conditions, such as data change trends, exceeding warning thresholds, etc., possible abnormal conditions or potential risks are identified to form a target feature data set.

[0039] The target feature data set is combined with the standard image data for image feature fusion processing. Image feature fusion processing comprehensively utilizes the key features in the monitoring data and the visual features in the image data, and generates more comprehensive image information through image recognition and feature matching methods.

[0040] This embodiment uses various types of sensors arranged at various target locations of water conservancy facilities to comprehensively collect multi-dimensional data such as environment, structural stress, flow and temperature, thereby realizing all-round monitoring of the status of water conservancy facilities. Secondly, by pre-standard processing and data alignment of preliminary monitoring data and image information, the rationality and consistency of data fusion are ensured, and the impact of data errors and noise on monitoring results is effectively reduced. In addition, based on preset rules and conditions during feature data extraction, potential abnormal conditions or risks of water conservancy facilities can be identified in a timely manner to ensure the timeliness and accuracy of early warning. By performing image feature fusion processing on the target feature data set and the standard image data, more intuitive and accurate image information can be generated, thereby improving the reliability and intuitiveness of monitoring and early warning.

[0041] In one embodiment, the image information is subjected to target detection and recognition by a preset E-YOLO algorithm to obtain an image recognition result, including: After obtaining the image information of the water conservancy facilities, image features are extracted from the image information. The extracted image feature data contains various details in the image, such as color, texture, edge, etc.

[0042] The extracted image feature data is input into the preset E-YOLO algorithm for target detection. The E-YOLO algorithm divides the image feature data into S×S grids, and each grid unit represents a part of the image.

[0043] Multiple bounding box predictions are performed for each grid cell to obtain the three-dimensional information of the bounding box including the position coordinates (x, y), width (w) and height (h). The bounding box prediction is based on the output of the neural network, combined with the image feature data, to predict the geometric properties of each bounding box.

[0044] After obtaining the 3D information of the bounding box, the confidence score of the object in each bounding box is further calculated. The confidence score indicates the possibility of the object in the bounding box, and is calculated by multiplying the confidence value output by the network by the conditional probability of the object in the bounding box. The conditional probabilities of multiple categories in each grid cell also need to be calculated, and the conditional probability reflects the possibility of different categories of objects appearing in the grid cell.

[0045] The final confidence score of each bounding box is calculated based on the confidence and class probability of the bounding box. The final confidence score is the product of the bounding box confidence and the class probability, reflecting the possibility of a specific class of objects appearing in a specific bounding box.

[0046] The non-maximum suppression rule is used to sort and filter the overlapping boxes with the final confidence. Non-maximum suppression is to suppress the bounding boxes with overlaps higher than the threshold by setting a threshold, retain the bounding box with the highest confidence, and finally obtain the target detection box.

[0047] The target detection frame is classified and processed, and the target recognition is performed. Target recognition is achieved through a preset classification algorithm, which classifies the detected target into a specific category and obtains the corresponding target recognition result.

[0048] According to the target recognition results, the original image information is annotated. The annotation process is to annotate the detected target box and its category information on the image to generate annotated image information. The annotated image information contains the location information and category information of each target in the image.

[0049] The target recognition results are fused with the annotated image information to obtain the final image recognition results. Information fusion is to combine the target recognition results with the image feature data to further improve the accuracy and completeness of recognition. The image recognition results contain all the detection targets in the water conservancy facilities and their specific information, which are used for subsequent monitoring and early warning analysis.

[0050] This embodiment can accurately obtain the position information and geometric properties of the target object in the image through image feature extraction and bounding box prediction, thereby improving the accuracy of detection. Secondly, the bounding box confidence and category probability are calculated, and the overlapping boxes are screened by the non-maximum suppression rule, which ensures the high confidence and low false detection rate of the target detection box and improves the reliability of the detection result. Furthermore, by classifying and identifying the target detection box, the category information of each target can be accurately determined to ensure the accuracy of the recognition result. The labeling of image information and information fusion steps further improves the integrity and accuracy of the image recognition results, making the final image recognition results more practical. In addition, this method does not require hardware replacement, and can adapt to the needs of different scenarios through algorithm optimization, which greatly improves the flexibility and adaptability of the monitoring system and provides strong technical support for the intelligent monitoring and early warning of water conservancy facilities.

[0051] In one embodiment, conditional probabilities of multiple categories are calculated for each grid unit to obtain category probabilities, including: Assign predefined category labels to each grid unit. Predefined category labels are set according to the monitoring requirements and characteristics of water conservancy facilities, such as water level, flow, rainfall and other category labels, and assign these labels to the corresponding grid units to form a target category set. The establishment of the target category set is to facilitate the subsequent vectorized coding processing.

[0052] Perform vectorized encoding on the target category set. Vectorized encoding is a processing method that converts category labels into numerical form. In this way, the category labels are converted into numerical forms that can be recognized and processed by computers to form a category feature mapping matrix. The category feature mapping matrix is ​​a matrix containing the numerical features of each category label, which is used to express the relationship and characteristics between category labels.

[0053] Perform function calculation on the category feature mapping matrix to obtain the initial category probability value. According to the numerical features in the category feature mapping matrix, calculate the initial probability value corresponding to each category label. These initial category probability values ​​reflect the initial distribution of each category label in the current grid unit.

[0054] According to the initial category probability value, the cross entropy loss is calculated for the category feature mapping matrix to obtain the category loss value. Cross entropy loss is a method to measure the difference between the predicted probability distribution and the actual distribution. By calculating the category loss value, the accuracy and reliability of the initial category probability value are evaluated. The smaller the category loss value, the closer the initial category probability value is to the actual situation.

[0055] The class loss value is back-propagated and optimized to obtain the optimized class features. In the back-propagation optimization process, the class loss value is gradually reduced by adjusting the parameters in the class feature mapping matrix, thereby optimizing the class features.

[0056] According to the optimized category features, the initial category probability value is corrected and calculated to obtain the corrected probability value. The correction calculation is to adjust and correct the initial category probability value according to the optimized category features so that it can more accurately reflect the actual distribution of the category label. The corrected probability value is the optimized and corrected category probability value with higher accuracy.

[0057] The corrected probability values ​​are subjected to threshold screening to obtain a candidate category probability set. Threshold screening is to screen the corrected probability values ​​according to the preset threshold standard, retain the category probability values ​​that meet the threshold standard, and form a candidate category probability set. The threshold standard is set according to specific monitoring needs and conditions to screen out category probability values ​​with higher credibility.

[0058] The maximum probability class selection is performed based on the candidate class probability set to obtain the class probability. In the candidate class probability set, the class label with the maximum probability value is selected as the final class probability. The maximum probability class selection is to determine the most likely class label in each grid unit and provide accurate class information for monitoring and early warning of water conservancy facilities.

[0059] This embodiment ensures the structuring and digitization of category label data and improves data processing efficiency by pre-defining category labels and vectorized encoding processing. Secondly, the calculation of the initial category probability value and the evaluation of the cross entropy loss provide an effective mechanism to measure and optimize the accuracy of the category probability value. Through back-propagation optimization processing, the category loss value is further reduced, making the optimized category features more accurate. The calculation of the corrected probability value and the threshold screening process ensure that the screened candidate category probability set has a high degree of credibility. Finally, through the maximum probability category selection, the most likely category label in each grid unit is determined, which improves the accuracy and reliability of monitoring and early warning. Overall, this method provides a scientific basis for the monitoring and early warning of water conservancy facilities through refined probability calculation and optimization processing, and effectively improves the accuracy and practicality of the early warning system.

[0060] In one embodiment, the preliminary monitoring data and the image information are processed comprehensively to obtain comprehensive monitoring data, including: After collecting the initial monitoring data, we will conduct periodic abnormal change analysis based on the changes in historical data and current data. Through this analysis, we can identify the periodic fluctuation characteristics of the monitoring data over time, such as water level changes, flow fluctuations and other time-related abnormal change trends. The results of this periodic sequence analysis will serve as an important basis for subsequent data processing.

[0061] After obtaining the results of the periodic sequence analysis, the preliminary monitoring data is further subjected to correlation feature extraction. The purpose of correlation feature extraction is to find out the data features with mutual correlation from the preliminary monitoring data. For example, when the water level changes abnormally, whether other data such as flow rate and pressure show synchronous or asynchronous change trends. This process forms a monitoring correlation feature set, which contains the potential correlation relationships between the monitoring items. These relationships play a key role in early warning judgment.

[0062] The monitoring-related feature set is fused with the image information to generate preliminary comprehensive data. Image information is mainly used to identify changes in the external environment around water conservancy facilities, such as potential risk factors such as river blockage and landslides in the surrounding area. During the fusion process, the data in the monitoring-related feature set is compared and associated with the image information to ensure that the comprehensive data is not only based on numerical values, but also includes the dynamic changes in the environment around the facility. This comprehensive processing makes the monitoring data more three-dimensional and comprehensive.

[0063] The preliminary comprehensive data is identified for anomalies according to the preset anomaly detection rules. The anomaly detection rules may include limit judgments on periodic features, identification of danger signs based on image information, etc. In this process, based on the abnormal features in the comprehensive data, such as sudden abnormal water level changes or unstable factors in the image, a preliminary judgment is made on the safety status of the facility to form anomaly detection data for further correction and analysis.

[0064] The preliminary comprehensive data is corrected and adjusted based on the abnormal detection data to obtain the final comprehensive monitoring data. The correction and adjustment include eliminating occasional invalid data, smoothing periodic changes, or optimizing false alarms caused by image noise.

[0065] This embodiment can effectively identify the periodic fluctuation characteristics in the monitoring data of water conservancy facilities through periodic abnormal change analysis, so as to better understand the changing rules of data and avoid false alarms. The associated feature extraction and image information fusion processing make the monitoring data not only based on numerical analysis, but also combined with actual environmental changes, further improving the comprehensiveness and reliability of the data. Furthermore, according to the preset anomaly detection rules, the preliminary comprehensive data is anomaly identified and corrected, which can quickly identify and correct anomalies in the data and ensure the high accuracy of the monitoring data. Finally, the generated comprehensive monitoring data can reflect the safety status of water conservancy facilities in real time, provide reliable data support for management and maintenance, and significantly improve the safety management level of water conservancy facilities. This method not only enhances the accuracy and real-time nature of monitoring and early warning, but also reduces false alarms and missed reports caused by data errors, greatly improving the overall efficiency and effectiveness of the water conservancy facility monitoring system.

[0066] In one embodiment, the image recognition results are integrated with the comprehensive monitoring data for analysis to obtain comprehensive status analysis results of water conservancy facilities, including: The images of water conservancy facilities are collected and processed to obtain image recognition results. In the image recognition process, deep learning algorithms are used to extract feature information such as cracks and deformations, and these feature information are converted into quantitative data, namely image feature data. These data include the length, width, location of the cracks, the amplitude of deformation, etc.

[0067] Multi-dimensional feature extraction and processing of comprehensive monitoring data can be performed, and data mining and statistical analysis methods can be used to extract characteristic data related to the safety status of water conservancy facilities, such as water level change trends, flow mutations, pressure fluctuations, etc., and convert them into comprehensive monitoring characteristic data.

[0068] The image feature data and the comprehensive monitoring feature data are weighted and fused by weight factors. The preset weight factors are determined based on historical data analysis, and different weight factors reflect the degree of influence of different features on the status of water conservancy facilities. For example, the length and width of the crack may have a higher weight, while the changes in temperature and humidity may have a lower weight. Through weighted fusion processing, fused feature data are obtained, which more comprehensively reflects the comprehensive status of water conservancy facilities.

[0069] Based on the fusion feature data, the image recognition results and the comprehensive monitoring data are analyzed for correlation. The correlation analysis uses big data analysis and machine learning technology to analyze the relationship between image feature data and comprehensive monitoring feature data, such as the correlation between crack expansion and water level changes, and the correlation between vibration frequency and pressure fluctuations. Through this correlation analysis, the correlation analysis results are obtained to identify potential risk factors and abnormal conditions.

[0070] The operation status of water conservancy facilities is comprehensively evaluated based on the results of correlation analysis. During the comprehensive evaluation process, different correlation analysis results are comprehensively considered to generate comprehensive status analysis results of water conservancy facilities. This result includes the current operation status evaluation of water conservancy facilities, potential risk warnings, and maintenance measures that need to be taken.

[0071] In this embodiment, the IoT monitoring and early warning method of water conservancy facilities can achieve a more comprehensive and accurate status assessment by fusing and analyzing the image recognition results with the comprehensive monitoring data. Image recognition combined with multi-dimensional feature extraction processing makes the monitoring of water conservancy facilities not limited to a single data source, but comprehensively considers multiple influencing factors, thereby improving the ability to identify potential risks and abnormal conditions. By using weighted fusion processing, flexible adjustments can be made according to the importance of different features to make the assessment results more targeted and accurate. The results of the association analysis can reveal the inherent connection between different monitoring data, help to discover possible safety hazards in advance, and provide timely early warning information. The comprehensive status assessment can not only reflect the health status of water conservancy facilities in real time, but also recommend necessary maintenance measures, thereby extending the service life of the facilities, reducing maintenance costs, and improving overall operating efficiency. This method improves the comprehensiveness and accuracy of monitoring through multi-source data fusion and intelligent analysis, ensuring the safe and stable operation of water conservancy facilities in various complex environments.

[0072] In one embodiment, a warning analysis is performed on the comprehensive status analysis result according to a preset warning threshold to obtain initial warning information, including: The comprehensive state analysis results are divided into sub-states to obtain multiple sub-state data. These sub-state data are detailed results of different parts and functions of water conservancy facilities, such as the structural state of the dam, the operation of the flood discharge equipment, water quality monitoring, etc.

[0073] The threshold judgment result is obtained by performing threshold judgment processing on each sub-state data according to the preset warning threshold. The warning threshold is a key indicator set by long-term monitoring data and expert experience. Exceeding these indicators may mean that the facility has potential risks. For example, the water level exceeds a certain height or the concentration of a specific pollutant in the water quality exceeds the standard.

[0074] Aggregate the threshold judgment results of all sub-status data to obtain the aggregate judgment result. The aggregate judgment result reflects the current status of the overall water conservancy facilities and is the comprehensive result of the analysis of multiple sub-status data. This step involves weighting and comprehensive calculation of each sub-status data in order to obtain a comprehensive status result.

[0075] The aggregated judgment results are analyzed according to the preset early warning risk rules to obtain the initial risk data. The early warning risk rules are set according to the historical operation data of the facility, environmental conditions and potential risk factors. For example, in the high rainfall season, some thresholds may need to be adjusted.

[0076] The aggregated judgment results are processed for risk level classification to obtain initial risk level data. Risk levels are usually divided into several levels, such as low, medium, high, and very high, in order to clearly indicate the current risk level.

[0077] After obtaining the initial risk data and initial risk level data, the early warning information is analyzed to form the initial early warning information. The initial early warning information includes a detailed description of the current facility status, its possible risk level and recommended measures.

[0078] The historical status analysis results of water conservancy facilities are obtained from the preset historical database, and data is extracted from these historical status analysis results to obtain historical status data. Historical status data is a summary and refinement of past monitoring data, which can reflect the operation status and various status changes of facilities in different time periods.

[0079] The indicator difference calculation is performed on the comprehensive status analysis results according to the preset monitoring indicators to obtain preliminary difference data. The monitoring indicators include water level, flow, water quality, etc. The changes in these indicators can directly reflect the operation of the facilities. The preliminary difference data is the quantitative result of the difference between the current state and the normal state.

[0080] Deviation analysis is performed on the comprehensive state analysis results based on historical state data to obtain deviation feature data. Deviation feature data is a description of the deviation of the current state from the historical state, which can help identify the causes and trends of abnormal changes.

[0081] The preliminary difference data and deviation feature data are corrected and analyzed to obtain the corrected warning information. The purpose of this step is to correct the preliminary warning information and improve the accuracy of the warning through the comparison and deviation analysis of historical data.

[0082] The initial warning information is dynamically adjusted based on the revised warning information to form the final warning information.

[0083] This embodiment adopts the method of sub-state segmentation and threshold judgment of the comprehensive state, which can finely identify the state changes of each functional part of the facility, making the early warning information more accurate. The introduction of aggregate judgment and risk analysis steps ensures a comprehensive assessment of the overall facility status, thereby enhancing the comprehensiveness and scientificity of risk identification. By using the historical database for deviation analysis and correction, the early warning information can be dynamically adjusted to ensure its high consistency with the actual situation and reduce the possibility of false alarms and omissions. In addition, by combining real-time data with historical data, the timeliness and accuracy of the early warning information are further improved, providing a strong guarantee for the safe operation of water conservancy facilities. This method not only improves the monitoring efficiency of water conservancy facilities, but also provides managers with a more valuable reference basis for decision-making.

[0084] Reference Figure 2 As shown, the present invention provides an Internet of Things monitoring and early warning system for water conservancy facilities, which is applied to any one of the above-mentioned Internet of Things monitoring and early warning methods for water conservancy facilities, including: The acquisition module is used to obtain data collected by various types of sensors arranged on water conservancy facilities, and perform preliminary classification processing to obtain preliminary monitoring data and image information; An analysis module, which is used to perform target detection and recognition on image information through a preset E-YOLO algorithm to obtain an image recognition result; A correlation module, which is used to comprehensively process the preliminary monitoring data and image information to obtain comprehensive monitoring data; A processing module is used to integrate and analyze the image recognition results with the comprehensive monitoring data to obtain the comprehensive status analysis results of the water conservancy facilities; The control module is used to perform early warning analysis on the comprehensive status analysis results according to the preset early warning threshold value to obtain initial early warning information.

[0085] The present invention provides an Internet of Things monitoring and early warning method, system, device and medium for water conservancy facilities, which has the following beneficial effects: By preliminarily classifying and processing the various sensor data of water conservancy facilities, and combining the E-YOLO algorithm to detect and identify the target of the image information, we can more accurately obtain the various status information of water conservancy facilities. This method improves the comprehensiveness and accuracy of monitoring data and provides a more reliable basis for the safety management of water conservancy facilities. By comprehensively processing sensor data and image information, and performing deviation analysis on the comprehensive state based on the preset warning threshold and historical data, dynamic monitoring of the status of water conservancy facilities is achieved, which helps to timely discover potential safety hazards and avoid misjudgment or omission due to single data. Based on the fusion analysis of image recognition results and sensor comprehensive monitoring data, it can ensure that the system can operate efficiently under various environmental conditions, reduce false alarms and omissions of warning information, and improve the accuracy and timeliness of warnings. By dynamically adjusting the warning information, a more reasonable warning strategy is formulated, and the optimized operation of the entire monitoring system is achieved through a comprehensive adjustment strategy, thereby effectively improving the safety protection capabilities of water conservancy facilities and reducing the risks caused by environmental changes or facility aging. By combining historical status analysis results with deviation analysis, the early warning information generation strategy can be flexibly adjusted according to the characteristics of different types of water conservancy facilities and actual demand changes, making the monitoring and early warning system more adaptable to diverse application scenarios, realizing refined management of water conservancy facilities, and effectively improving the comprehensive operation efficiency and safety assurance level of water conservancy facilities.

[0086] Reference Figure 3 As shown, the present invention also provides an Internet of Things monitoring and early warning device for water conservancy facilities, including: Memory, used to store programs; The processor is used to execute the program to implement each step of any one of the above-mentioned Internet of Things monitoring and early warning methods for water conservancy facilities.

[0087] In this embodiment, the processor and the memory may be connected via a bus or other means. The memory may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a read-only memory, a flash memory, a hard disk, or a solid-state drive. The processor may be a general-purpose processor, such as a central processing unit, a digital signal processor, an application-specific integrated circuit, or one or more integrated circuits configured to implement the embodiments of the present invention.

[0088] The present invention also provides a storage medium storing computer instructions, wherein the computer instructions are used to enable a computer to execute any of the above methods.

[0089] It should be noted that technicians in the relevant technical field can clearly understand that for the convenience and conciseness of description, the specific working process of the system and each module described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0090] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An Internet of Things monitoring and early warning method for water conservancy facilities, characterized in that: include: Acquire data collected by various types of sensors arranged on water conservancy facilities, and perform preliminary classification and processing to obtain preliminary monitoring data and image information; Performing target detection and recognition on the image information through a preset E-YOLO algorithm to obtain an image recognition result; Comprehensively processing the preliminary monitoring data and image information to obtain comprehensive monitoring data; The image recognition result is integrated with the comprehensive monitoring data for analysis to obtain a comprehensive status analysis result of the water conservancy facilities; Perform a warning analysis on the comprehensive status analysis result according to a preset warning threshold to obtain initial warning information.

2. The Internet of Things monitoring and early warning method for water conservancy facilities according to claim 1 is characterized in that: The data collected by various types of sensors arranged on water conservancy facilities are obtained, and preliminary classification and processing are performed to obtain preliminary monitoring data and image information, including: Preliminarily screening and processing the data collected by the multiple types of sensors arranged at each target position of the water conservancy facility to obtain effective monitoring data; The effective monitoring data are sorted by type to obtain environmental monitoring data, structural stress monitoring data, flow monitoring data and temperature monitoring data; Acquire original image data collected by an image sensor arranged on the water conservancy facility, and perform pre-standard processing to obtain standard image data; The environmental monitoring data, the structural stress monitoring data, the flow monitoring data and the temperature monitoring data are subjected to data fusion processing, and the data are aligned according to timestamps and spatial positions to obtain preliminary monitoring data; Extracting feature data based on the preliminary monitoring data to obtain a target feature data set; The target feature data set is subjected to image feature fusion processing with the standard image data to obtain the image information.

3. The Internet of Things monitoring and early warning method for water conservancy facilities according to claim 1 is characterized in that: The target detection and recognition is performed on the image information by using a preset E-YOLO algorithm to obtain an image recognition result, including: Extracting image features from the image information to obtain image feature data; The image feature data is input into the E-YOLO algorithm for target detection to obtain the detected target frame, including: Dividing the image feature data into an S×S grid to obtain a plurality of grid units; Performing multiple bounding box predictions on each of the grid cells to obtain a three-dimensional bounding box including position coordinates, width, and height; Calculating a confidence score of the existence of an object for each of the bounding boxes according to the three-dimensional bounding box to obtain a bounding box confidence score; Calculate the conditional probabilities of multiple categories for each grid unit to obtain the category probability; Calculating a confidence score for each of the bounding boxes according to the bounding box confidence and the category probability to obtain a final confidence score; According to a preset non-maximum suppression rule, the final confidence level is sorted and screened for overlapping frames to obtain a target detection frame; Classifying the target detection frame and performing target recognition to obtain a corresponding target recognition result; Annotating the image information according to the target recognition result to obtain annotated image information; The target recognition result is fused with the annotated image information to obtain the image recognition result.

4. The Internet of Things monitoring and early warning method for water conservancy facilities according to claim 3 is characterized in that: The conditional probability calculation of multiple categories is performed on each grid unit to obtain the category probability, including: Assigning predefined category labels to the grid units to obtain a target category set; Performing vectorized encoding processing on the target category set to obtain a category feature mapping matrix; Performing function calculation on the category feature mapping matrix to obtain an initial category probability value; Performing cross entropy loss calculation on the category feature mapping matrix according to the initial category probability value to obtain a category loss value; Performing back propagation optimization processing on the category loss value to obtain optimized category features; Performing correction calculation on the initial category probability value according to the optimized category feature to obtain a corrected probability value; Performing threshold screening processing on the corrected probability value to obtain a candidate category probability set; The maximum probability category is selected according to the candidate category probability set to obtain the category probability.

5. The Internet of Things monitoring and early warning method for water conservancy facilities according to claim 1 is characterized in that: The comprehensive processing of the preliminary monitoring data and the image information to obtain comprehensive monitoring data includes: Performing periodic abnormal change analysis on the preliminary monitoring data to obtain periodic sequence analysis results; Extracting correlation features from the preliminary monitoring data according to the periodic sequence analysis results to obtain a monitoring correlation feature set; Fusing the monitoring-related feature set with the image information to obtain preliminary comprehensive data; Performing anomaly identification on the preliminary comprehensive data according to a preset anomaly detection rule to obtain anomaly detection data; The preliminary comprehensive data is corrected and adjusted according to the abnormal detection data to obtain the comprehensive monitoring data.

6. The Internet of Things monitoring and early warning method for water conservancy facilities according to claim 1 is characterized in that: The image recognition result is integrated with the comprehensive monitoring data for analysis to obtain a comprehensive status analysis result of the water conservancy facilities, including: Performing feature extraction processing on the image recognition result to obtain image feature data; Performing multi-dimensional feature extraction processing on the comprehensive monitoring data to obtain comprehensive monitoring feature data; Performing weighted fusion processing on the image feature data and the comprehensive monitoring feature data according to a preset weight factor to obtain fused feature data; Performing correlation analysis on the image recognition result and the comprehensive monitoring data according to the fusion feature data to obtain a correlation analysis result; A comprehensive evaluation of the operating status of the water conservancy facilities is performed based on the correlation analysis results to obtain a comprehensive status analysis result.

7. The Internet of Things monitoring and early warning method for water conservancy facilities according to claim 1 is characterized in that: The step of performing a warning analysis on the comprehensive status analysis result according to a preset warning threshold to obtain initial warning information includes: Performing sub-state segmentation on the comprehensive state analysis result to obtain a plurality of sub-state data; Performing threshold judgment processing on each of the sub-state data according to the early warning threshold to obtain a threshold judgment result; Aggregating the threshold judgment results to obtain an aggregate judgment result; Performing risk analysis on the aggregated judgment results according to preset early warning risk rules to obtain initial risk data; Performing risk level classification processing on the comprehensive status analysis result according to the aggregate judgment result to obtain initial risk level data; Performing early warning information analysis on the initial risk data and the initial risk level data to obtain the initial early warning information; Acquire the historical status analysis result of the water conservancy facility from a preset historical database, and perform data extraction on the historical status analysis result to obtain historical status data; Performing indicator difference calculation on the comprehensive status analysis result according to preset monitoring indicators to obtain preliminary difference data; Performing deviation analysis on the comprehensive state analysis result according to the historical state data to obtain deviation characteristic data; Performing correction analysis on the preliminary difference data and the deviation characteristic data to obtain correction warning information; The initial warning information is dynamically adjusted according to the revised warning information to obtain final warning information.

8. An Internet of Things monitoring and early warning system for water conservancy facilities, characterized in that: The Internet of Things monitoring and early warning method for water conservancy facilities applied to any one of claims 1 to 7 above comprises: A collection module, which is used to obtain data collected by various types of sensors arranged on water conservancy facilities, and perform preliminary classification processing to obtain preliminary monitoring data and image information; An analysis module, wherein the analysis module is used to perform target detection and recognition on the image information through a preset E-YOLO algorithm to obtain an image recognition result; A correlation module, wherein the correlation module is used to comprehensively process the preliminary monitoring data and the image information to obtain comprehensive monitoring data; A processing module, the processing module is used to fuse and analyze the image recognition result with the comprehensive monitoring data to obtain a comprehensive status analysis result of the water conservancy facility; A control module is used to perform early warning analysis on the comprehensive status analysis result according to a preset early warning threshold to obtain initial early warning information.

9. An Internet of Things monitoring and early warning device for water conservancy facilities, characterized in that: include: Memory, used to store programs; The processor is used to execute the program to implement the various steps of the Internet of Things monitoring and early warning method for water conservancy facilities as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: Computer instructions are stored, and the computer instructions are used to make a computer execute the method according to any one of claims 1 to 7.

Images