Subway construction safety monitoring system based on construction data analysis
By combining data processing and weight adjustment with physical sensors and image analysis modules, the problems of time-consuming and error-prone manual processing in traditional subway construction safety monitoring systems have been solved, achieving efficient and accurate on-site safety monitoring.
Patent Information
- Application Number
- CN202311735562.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-18
- Publication Date
- 2026-04-10
AI Technical Summary
Traditional subway construction safety monitoring systems rely on manual data processing, which consumes a lot of time and human resources and is prone to human error, making it impossible to dynamically adjust the accuracy of monitoring results.
The monitoring data is processed by a physical sensor analysis module and an image analysis module respectively. Initial weights are calculated using a linear regression model and a decision tree model. The results are then fused and analyzed by a result fusion module to perform weighted averaging and automated weight adjustment, ensuring the accuracy of the monitoring results.
It enables real-time and automated data processing and monitoring, reduces human error, and improves the accuracy and flexibility of safety monitoring at construction sites.
Smart Images
Figure CN121834637A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of subway safety monitoring technology, specifically a subway construction safety monitoring system based on construction data analysis. Background Technology
[0002] Traditional subway construction safety monitoring systems use sensors to collect monitoring data at the construction site in order to capture changes and anomalies during the construction process. The data collected by the sensors is collected and recorded regularly. This process usually requires human intervention. For example, engineers or monitoring personnel will regularly inspect the construction site, use the corresponding instruments and equipment to collect sensor data, and record it.
[0003] Next, traditional systems process and analyze the collected data. This process typically requires specialized technicians to use specific software or computing tools to filter, clean, and organize the data, as well as apply statistical analysis and models to obtain relevant indicators and parameters about the construction site.
[0004] Therefore, traditional systems typically rely on limited physical sensor data for monitoring and analysis, requiring manual data processing and analysis, which can consume a lot of time and human resources and is prone to human error. In addition, traditional systems often treat sensor data equally and cannot make dynamic adjustments based on the accuracy of the monitoring results. Summary of the Invention
[0005] The purpose of this invention is to provide a subway construction safety monitoring system based on construction data analysis to solve the problems mentioned in the background art.
[0006] To achieve the above objectives, the present invention provides the following technical solution: a subway construction safety monitoring system based on construction data analysis, which includes a physical sensor analysis module, an image analysis module, and a result fusion analysis module, wherein:
[0007] The physical sensor analysis module uses different physical sensors to collect monitoring data, processes the monitoring data to obtain physical sensor monitoring results, and calculates the initial weights of the physical sensors using a linear regression model based on the physical sensor monitoring results. The physical sensor monitoring results include physical sensor vibration values, physical sensor temperature values, and physical sensor settlement values.
[0008] The image analysis module uses a camera to collect image data, uses image processing technology to obtain image monitoring results, and uses a decision tree model to calculate the initial weight of the image based on the image monitoring results. The image monitoring results include image vibration value, image temperature value, and image settlement value.
[0009] The result fusion analysis module performs a weighted average calculation based on the monitoring results and initial weights from the physical sensor analysis module and the image analysis module. The calculated weighted average result is compared with the monitoring results from the physical sensor analysis module and the image analysis module. When the difference between the weighted average result and the monitoring result exceeds a set threshold, the weights in the physical sensor analysis module and the image analysis module are adjusted, and the weighted average calculation is performed again until all differences are less than the set threshold.
[0010] As a further improvement to this technical solution, the physical sensor analysis module includes a sensor data acquisition unit and a sensor data processing unit. The sensor data acquisition unit collects monitoring data using different physical sensors. The sensor data processing unit processes the monitoring data to obtain physical sensor monitoring results, calculates the initial weights of the physical sensors using a linear regression model based on the physical sensor monitoring results, and sends the physical sensor monitoring results and the initial weights of the physical sensors to the result fusion analysis module.
[0011] As a further improvement to this technical solution, the image analysis module includes an image collection unit, an image processing unit, and an image analysis unit. The image collection unit collects image data using a camera; the image processing unit extracts image features based on the image data using computer vision technology; the image analysis unit analyzes the image monitoring results based on the image features using image processing technology, calculates the initial weight of the image using a decision tree model based on the image monitoring results, and sends the image monitoring results and the initial weight of the image to the result fusion analysis module.
[0012] As a further improvement to this technical solution, the result fusion analysis module includes a result fusion feedback unit and a method optimization unit. The result fusion feedback unit receives monitoring results and initial weights sent by the sensor data processing unit and the image analysis unit. The method optimization unit performs weighted average calculations based on the monitoring results and initial weights in the physical sensor analysis module and the image analysis module, respectively, and compares the calculated weighted average result with the monitoring results in the result fusion feedback unit. When the difference between the weighted average result and the monitoring result exceeds a set threshold, the weights in the sensor data processing unit and the image analysis unit are adjusted.
[0013] As a further improvement to this technical solution, the sensor data processing unit calculates the initial weights of the physical sensors using a linear regression model based on the monitoring results of the physical sensors, specifically including:
[0014] Collect a large amount of monitoring data, including the monitoring values of various physical sensors and the corresponding initial weights of the physical sensors; preprocess the collected data, including data cleaning and outlier removal;
[0015] Identify the characteristic variables affecting the monitoring results, including the monitoring values of physical sensors; select a linear regression model as the basic model for weight calculation, the linear regression model is expressed as Y=β0+β1X1+β2X2+…+βn*Xn, where Y is the initial weight of the physical sensors, X1,X2,…,Xn are the monitoring values of the physical sensors, and β0,β1,β2,…,βn are the coefficients of the model;
[0016] The collected data is used to train the linear regression model, and the optimal coefficient values β0, β1, β2, ..., βn are fitted based on the initial weights of the physical sensors and the monitoring values of the physical sensors. Finally, the initial weights of the physical sensors are calculated based on the linear regression model.
[0017] As a further improvement to this technical solution, the image analysis unit calculates the initial weights of the image based on the image monitoring results using a decision tree model, specifically including:
[0018] Collect image monitoring data, including various image feature data and initial image weights; select the feature with the smallest Gini coefficient in the data as the root node. The Gini coefficient is used to measure the purity and uncertainty of the feature.
[0019] For each child node under the root node, further subdivide according to the size of the Gini coefficient, and keep iterating until the nodes belong to the same category, then stop growing and the model training is complete;
[0020] The trained model is used to calculate the initial weights of the image monitoring results. Starting from the root node, the feature values of the samples are compared with the values of the partitioning features corresponding to that node. Based on the different values, the samples are moved along the corresponding child nodes. The previous step is repeated, and the samples are moved to the next child node based on the partitioning features and values of the current node, until the leaf node is reached. The leaf node corresponds to the value of the initial weight of the image.
[0021] As a further improvement to this technical solution, the image analysis unit utilizes image processing technology to analyze image features and derive image monitoring results, specifically including:
[0022] Vibration: Through image processing algorithms, signs of vibration amplitude were detected in the working area, and the vibration amplitude was inferred based on the degree of image change;
[0023] Temperature: Temperature changes are inferred by using images from cameras to observe color variations, material expansion, and physical changes in objects in the environment.
[0024] Settlement: Settlement is determined by detecting changes in the relative position of a specific marked object in an image.
[0025] As a further improvement to this technical solution, the method optimization unit performs weighted average calculations based on the monitoring results and initial weights in the physical sensor analysis module and the image analysis module. The weighted average of each monitoring parameter is calculated by using weights. For each parameter, the sum of the two sets of data multiplied by their corresponding weights is then divided by the sum of the weights.
[0026] As a further improvement to this technical solution, the method optimization unit compares the calculated weighted average result with the monitoring result in the result fusion feedback unit. When the difference between the weighted average result and the monitoring result exceeds a set threshold, the weights in the sensor data processing unit and the image analysis unit are adjusted, specifically including:
[0027] A threshold is set. When the difference between the weighted fusion result and the monitoring result of the sensor data processing unit or image analysis unit exceeds the threshold value, the weight needs to be adjusted. Based on the difference assessment, the weight is adjusted in increments of 0.1 or decreases until all differences are eventually less than the set threshold.
[0028] Compared with the prior art, the beneficial effects of the present invention are:
[0029] 1. This subway construction safety monitoring system based on construction data analysis first uses different physical sensors to collect monitoring data, processes the data to obtain monitoring results, and calculates the initial weights of the physical sensors using a linear regression model based on the monitoring results. It then uses cameras to collect image data, employs image processing technology to obtain monitoring results, and calculates the initial weights of the images using a decision tree model based on the image monitoring results. Finally, it performs weighted average calculations based on different monitoring results and initial weights to obtain the final monitoring results. Through the acquisition and fusion analysis of multi-source data, it can provide more comprehensive and accurate monitoring results, helping to determine the safety of the construction site. Furthermore, it enables real-time monitoring and automated data processing and analysis, improving work efficiency and reducing human error.
[0030] 2. This subway construction safety monitoring system based on construction data analysis compares the calculated weighted average result with the monitoring results in the physical sensor analysis module and the image analysis module. When the difference between the weighted average result and the monitoring result exceeds the set threshold, the weights in the physical sensor analysis module and the image analysis module are adjusted, and the weighted average is calculated again until all differences are less than the set threshold. This automated weight adjustment makes the system more flexible and refined, effectively improving the accuracy of analysis and prediction. Attached Figure Description
[0031] Figure 1 This is a schematic diagram of the overall modules of the present invention;
[0032] Figure 2 This is a schematic diagram of the physical sensor analysis module unit of the present invention;
[0033] Figure 3 This is a schematic diagram of the image analysis module unit of the present invention;
[0034] Figure 4 This is a schematic diagram of the result fusion analysis module unit of the present invention.
[0035] In the diagram: 100, Physical sensor analysis module; 101, Sensor data acquisition unit; 102, Sensor data processing unit; 200, Image analysis module; 201, Image collection unit; 202, Image processing unit; 203, Image analysis unit; 300, Result fusion analysis module; 301, Result fusion feedback unit; 302, Method optimization unit. Detailed Implementation
[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on 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.
[0037] Please see Figure 1-4 The present invention provides a technical solution: a subway construction safety monitoring system based on construction data analysis, including a physical sensor analysis module 100, an image analysis module 200, and a result fusion analysis module 300.
[0038] The sensor data acquisition unit 101 in the physical sensor analysis module 100 deploys different physical sensors on the walls and structures surrounding the subway tunnel and collects monitoring data using these sensors, which include vibration sensors, temperature sensors, and settlement sensors. Specifically, the physical sensors include:
[0039] Vibration sensors are used to measure vibration conditions at subway construction sites, including vibration amplitude. The sensors acquire vibration data by measuring the acceleration and displacement of objects. For example, a vibration amplitude of 3.2 mm / s is recorded. Vibration sensors measure seismic vibrations around subway tunnels and mechanical vibrations on construction sites. Vibration sensors have different measurement ranges and sensitivities to adapt to different vibration scenarios. For example, a vibration sensor has a measurement range of 0-200 mm / s and an accuracy of 0.1 mm / s.
[0040] Temperature sensors: Temperature sensors are used to monitor changes in ambient temperature at subway construction sites, including site temperature. The sensors directly measure the temperature of the surrounding air and also measure temperature by contacting the surface of objects. For example, a site temperature of 27°C may be recorded. Ambient temperature sensors measure temperature changes inside subway tunnels and at construction sites. The measurement range and accuracy of temperature sensors vary depending on the type of sensor. For example, a temperature sensor may have a measurement range of -40°C to +80°C and an accuracy of ±0.5°C.
[0041] Settlement sensors: Settlement sensors are used to detect changes in ground settlement, including the amount of settlement. The sensors acquire settlement data by measuring changes in the height of the ground and structures. For example, the settlement amount is recorded as 2 mm. In subway tunnel construction, settlement sensors are installed on the ground to measure changes in ground settlement. Settlement sensors have specific measurement ranges and accuracies. For example, the measurement range of a settlement sensor can be 0-20 mm, and the accuracy can reach 0.01 mm.
[0042] To obtain more accurate and reliable data to support subsequent data analysis, evaluation, and decision-making, the sensor data acquisition unit 101 sends the acquired data to the sensor data processing unit 102. The sensor data processing unit 102 processes the vibration data, temperature, and settlement data, specifically including:
[0043] Filtering and noise reduction: Since sensors may be subject to unavoidable noise interference during data acquisition, filtering is used to smooth the data and remove these noisy signals to provide more reliable and useful data. For example, a low-pass filter allows low-frequency signals to pass through but attenuates high-frequency signals, and is used to remove high-frequency noise while retaining low-frequency components.
[0044] Calibration and verification: Sensor data needs to be calibrated and verified to ensure its accuracy and reliability. Calibration is to adjust the sensor's measurement deviation by comparing it with known standard values. Verification is to check the consistency and reasonableness of the sensor data. For example, anomaly detection checks for outliers, such as extreme values that are outside the reasonable range or abnormal fluctuations that deviate from the normal trend.
[0045] The sensor data processing unit 102 will calculate the initial weights of the physical sensors based on the monitoring results of the physical sensors using a linear regression model, specifically including:
[0046] Collect and organize a large amount of monitoring data, including the monitoring values of each physical sensor and the corresponding initial weights of the physical sensors;
[0047] The collected data is preprocessed, including data cleaning, outlier removal, and data normalization or standardization, to ensure data quality and reliability.
[0048] Identify the characteristic variables that affect the monitoring results, namely the monitoring values of physical sensors. The characteristic variables are measurable, highly correlated, and have a significant impact on the monitoring results.
[0049] The linear regression model is chosen as the basis for weight calculation. The linear regression model is expressed as Y=β0+β1X1+β2X2+…+βn*Xn, where Y is the initial weight of the physical sensor, X1,X2,…,Xn are the monitored values of the physical sensor, and β0,β1,β2,…,βn are the coefficients of the model.
[0050] The collected data is used to train the linear regression model, that is, to fit the optimal coefficient values β0, β1, β2, ..., βn based on the initial weights of the physical sensors and the monitoring values of the physical sensors.
[0051] Finally, Y (the initial weights of the physical sensors) is calculated based on the linear regression model.
[0052] The sensor data processing unit 102 sends the initial weights of the physical sensors and the processed monitoring results of the physical sensors (physical sensor vibration value, physical sensor temperature value and physical sensor settlement value) to the result fusion feedback unit 301 in the result fusion and analysis module.
[0053] The image collection unit 201 in the image analysis module 200 first places the camera in a location that covers the construction area, and fixes the camera on a wall, bracket or facility to obtain a stable shooting angle; then the acquisition frequency is set to one frame every few seconds to periodically collect images of the subway construction site.
[0054] The image collection unit 201 sends the collected image data to the image processing unit 202, which extracts features from the collected images for subsequent monitoring result generation. Specifically, this includes:
[0055] Image processing algorithms: The acquired image data will be input into image processing algorithms. Image processing algorithms are a set of mathematical and computer vision techniques used to analyze and process images, including edge detection, object detection, and image segmentation.
[0056] Edge detection: Edge detection can detect the boundaries or contours in an image. By detecting the edges in an image, the shape and contour information of the object can be obtained.
[0057] Object detection: Object detection algorithms are used to detect and locate specific targets and objects in an image, identify objects of a specific category, and obtain the location information of targets in the image through object detection;
[0058] Image segmentation: Image segmentation is the process of dividing an image into multiple regions, each with similar attributes. Through image segmentation, features of different regions in an image can be extracted, such as the state of a construction area, for more detailed analysis and monitoring.
[0059] Feature extraction: After processing by image processing algorithms, useful features are extracted from the image. These features include the location of the equipment and the state of the construction area. Feature extraction is achieved through image recognition and target tracking methods to obtain key image features.
[0060] Image processing unit 202 sends the extracted image features to image analysis unit 203, which performs data analysis based on the image features, specifically including:
[0061] Vibration: Through image processing algorithms, signs of vibration amplitude were detected in the working area. Based on the inference of the degree of image change, the vibration amplitude was estimated to be approximately 3.5 mm / s. That is, vibration was detected by observing a continuous sequence of images captured by a surveillance camera. If an object has a slight positional shift between consecutive image frames, the magnitude of the vibration is estimated by the amplitude and frequency of these shifts. The method detects and quantifies pixel-level movement or change, thereby estimating the vibration velocity of the object.
[0062] Temperature: The image shows localized overheating in the construction area, with the temperature expected to exceed the normal value, possibly reaching 30°C. The camera infers temperature changes by using indirect factors such as color changes, material expansion, and physical changes of objects in the environment. For example, some materials may exhibit specific color changes or shape changes at high temperatures. These changes, combined with pre-obtained baseline data and standards, are used to estimate temperature changes.
[0063] Settlement: Image analysis showed that the position of a specific marked object had changed, and the estimated settlement was about 1.8 mm. That is, settlement was determined by detecting the relative position change of a specific marked object in the image. If the image analysis showed that the position of the marked object relative to other reference points had shifted downward, the settlement was estimated based on the amount of these position changes. The image analysis software measured the position of the marked object in each image and compared the changes of these positions over time to infer the settlement situation.
[0064] Image analysis unit 203 will calculate the initial weights of the images using a decision tree model based on the monitoring results, specifically including:
[0065] Collect and organize a large amount of monitoring data, including various image feature data and initial image weights;
[0066] The feature with the smallest Gini coefficient in the data is selected as the root node. The formula for calculating the Gini coefficient is: Gini = 1 - Σ(pi)^2, where Gini is the Gini coefficient, which measures the purity and uncertainty of a feature; pi is the probability of a certain class, representing the probability that the sample belongs to that class; Σ is the summation symbol, representing the summation of the probabilities of all classes; (pi)^2 represents the square of the probability of a certain class, that is, the probability of that class multiplied by itself.
[0067] For each child node under the root node, further subdivide according to the size of the Gini coefficient, and keep iterating until the node reaches a pure state, that is, the samples in the node belong to the same category, then stop growing and the model training is completed.
[0068] The trained model is used to calculate the initial weights of the image monitoring results. Starting from the root node, the feature values of the sample are compared with the values of the partitioning features corresponding to that node. Based on the different values, the sample is moved along the corresponding child node and the data of the moved node is recorded. The previous step is repeated. Based on the partitioning features and values of the current node, the sample is moved to the next child node until the leaf node is reached. The leaf node corresponds to the value of the initial weight of the image.
[0069] The image analysis unit 203 sends the initial image weights and the analyzed image monitoring results to the result fusion feedback unit 301.
[0070] The result fusion feedback unit 301 receives the initial weights and monitoring results (physical sensor vibration value, physical sensor temperature value, and physical sensor settlement value) of the physical sensor sent by the sensor data processing unit 102, and the initial weights and monitoring results (image vibration value, image temperature value, and image settlement value) of the image sent by the image analysis unit 203. The method optimization unit 302 analyzes and processes these two results, specifically including:
[0071] Calculate the weighted average: Use weights to calculate the weighted average of each monitoring parameter. For each parameter, multiply the two sets of data by their corresponding weights and then divide by the sum of the weights. For example, the weighted average of vibration = (initial weight of physical sensor * physical sensor vibration value + initial weight of image * image vibration value) / (initial weight of physical sensor + initial weight of image).
[0072] Set a comparison threshold: Set a threshold value. When the difference between the weighted fusion result and the monitoring result of the sensor data processing unit 102 or the image analysis unit 203 exceeds this value, the weight needs to be adjusted. The threshold value is 0.1.
[0073] Assess whether weight adjustment is needed: If the difference of any monitored parameter exceeds the threshold, it means that the weighted average does not adequately reflect the data of the two methods. Assess the need to adjust the weight based on the difference. If the difference exceeds the threshold, it indicates that a higher weight should be given to the sensor data processing unit 102 or the image analysis unit 203.
[0074] Adjusting weights: Adjusting weights is an iterative process that adjusts weights in increments of 0.1 or by decrementing them until all differences are eventually less than a set threshold.
[0075] Determine the final weighted average result: Once the difference between the weighted fusion result and the monitoring results in the sensor data processing unit 102 and the image analysis unit 203 remains within the threshold range, the final weight allocation is confirmed, and the final fusion monitoring result is obtained using these weights.
[0076] The following uses specific data to illustrate this, for example:
[0077] Initial weights for physical sensors: WA = 0.5; Initial weights for images: WB = 0.5;
[0078] The sensor data processing unit 102 measured a vibration value of 3.2 mm / s, a temperature value of 27℃, and a settlement value of 2 mm.
[0079] The vibration value measured by the image analysis unit 203 was 3.5 mm / s; the temperature value was 30℃; and the settlement value was 1.8 mm.
[0080] Vibration data fusion results:
[0081] (WA*3.2+WB*3.5) / (WA+WB)=(0.5*3.2+0.5*3.5) / (0.5+0.5)=3.35mm / s;
[0082] Temperature data fusion results:
[0083] (WA*27+WB*30) / (WA+WB)=(0.5*27+0.5*30) / (0.5+0.5)=28.5℃;
[0084] Settlement data fusion results:
[0085] (WA*2+WB*1.8) / (WA+WB)=(0.5*2+0.5*1.8) / (0.5+0.5)=1.9mm;
[0086] Set a threshold of 0.1. If the difference between the fusion result and any other method is greater than this threshold, the weights need to be adjusted.
[0087] Comparison difference:
[0088] Vibration data comparison: Difference in sensor data processing unit 102 = |3.35 - 3.2| = 0.15; Difference in image analysis unit 203 = |3.5 - 3.35| = 0.15.
[0089] Temperature data comparison: Difference in sensor data processing unit 102 = |28.5-27| = 1.5; Difference in image analysis unit 203 = |30-28.5| = 1.5.
[0090] Settlement data comparison: Difference of sensor data processing unit 102 = |1.9-2| = 0.1, Difference of image analysis unit 203 = |1.8-1.9| = 0.1;
[0091] Since the temperature difference is the greatest, the weight of temperature is adjusted by increasing WA by 0.1 and decreasing WB by 0.1, i.e., setting WA to 0.6 and WB to 0.4. Then the weighted average is recalculated until the difference between the results of the sensor data processing unit 102 and the image analysis unit 203 is less than the threshold we set, so as to ensure the reliability and accuracy of the fusion result.
[0092] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.
Claims
1. A subway construction safety monitoring system based on construction data analysis, characterized in that: It includes a physical sensor analysis module (100), an image analysis module (200), and a result fusion analysis module (300), wherein: The physical sensor analysis module (100) uses different physical sensors to collect monitoring data, processes the monitoring data to obtain physical sensor monitoring results, and calculates the initial weights of the physical sensors using a linear regression model based on the physical sensor monitoring results. The physical sensor monitoring results include physical sensor vibration values, physical sensor temperature values, and physical sensor settlement values. The image analysis module (200) uses a camera to collect image data, uses image processing technology to obtain image monitoring results, and uses a decision tree model to calculate the initial weight of the image based on the image monitoring results. The image monitoring results include image vibration value, image temperature value and image settlement value. The result fusion analysis module (300) performs weighted average calculation based on the monitoring results and initial weights in the physical sensor analysis module (100) and the image analysis module (200). The calculated weighted average result is compared with the monitoring results in the physical sensor analysis module (100) and the image analysis module (200). When the difference between the weighted average result and the monitoring result exceeds a set threshold, the weights in the physical sensor analysis module (100) and the image analysis module (200) are adjusted, and the weighted average calculation is performed again until all differences are less than the set threshold.
2. The subway construction safety monitoring system based on construction data analysis according to claim 1, characterized in that: The physical sensor analysis module (100) includes a sensor data acquisition unit (101) and a sensor data processing unit (102). The sensor data acquisition unit (101) collects monitoring data using different physical sensors. The sensor data processing unit (102) processes the monitoring data to obtain physical sensor monitoring results, calculates the initial weights of the physical sensors using a linear regression model based on the physical sensor monitoring results, and sends the physical sensor monitoring results and the initial weights of the physical sensors to the result fusion analysis module (300).
3. The subway construction safety monitoring system based on construction data analysis according to claim 2, characterized in that: The image analysis module (200) includes an image collection unit (201), an image processing unit (202), and an image analysis unit (203). The image collection unit (201) collects image data using a camera; the image processing unit (202) extracts image features based on the image data using computer vision technology. The image analysis unit (203) uses image processing technology to analyze the image monitoring results based on the image features, and calculates the initial weight of the image using a decision tree model based on the image monitoring results. The image monitoring results and the initial weight of the image are then sent to the result fusion analysis module (300).
4. The subway construction safety monitoring system based on construction data analysis according to claim 3, characterized in that: The result fusion analysis module (300) includes a result fusion feedback unit (301) and a method optimization unit (302). The result fusion feedback unit (301) receives the monitoring results and initial weights sent by the sensor data processing unit (102) and the image analysis unit (203). The method optimization unit (302) performs weighted average calculations based on the monitoring results and initial weights in the physical sensor analysis module (100) and the image analysis module (200), and compares the calculated weighted average result with the monitoring results in the result fusion feedback unit (301). When the difference between the weighted average result and the monitoring result exceeds a set threshold, the weights in the sensor data processing unit (102) and the image analysis unit (203) are adjusted.
5. The subway construction safety monitoring system based on construction data analysis according to claim 2, characterized in that: The sensor data processing unit (102) calculates the initial weights of the physical sensors using a linear regression model based on the monitoring results of the physical sensors, specifically including: Collect a large amount of monitoring data, including the monitoring values of various physical sensors and the corresponding initial weights of the physical sensors; preprocess the collected data, including data cleaning and outlier removal; Identify the characteristic variables affecting the monitoring results, including the monitoring values of physical sensors; select a linear regression model as the basic model for weight calculation, the linear regression model is expressed as Y=β0+β1X1+β2X2+…+βn*Xn, where Y is the initial weight of the physical sensors, X1,X2,…,Xn are the monitoring values of the physical sensors, and β0,β1,β2,…,βn are the coefficients of the model; The collected data is used to train the linear regression model, and the optimal coefficient values β0, β1, β2, ..., βn are fitted based on the initial weights of the physical sensors and the monitoring values of the physical sensors. Finally, the initial weights of the physical sensors are calculated based on the linear regression model.
6. The subway construction safety monitoring system based on construction data analysis according to claim 3, characterized in that: The image analysis unit (203) calculates the initial weights of the image based on the image monitoring results using a decision tree model, specifically including: Collect image monitoring data, including various image feature data and initial image weights; select the feature with the smallest Gini coefficient in the data as the root node. The Gini coefficient is used to measure the purity and uncertainty of the feature. For each child node under the root node, further subdivide according to the size of the Gini coefficient, and keep iterating until the nodes belong to the same category, then stop growing and the model training is complete; The trained model is used to calculate the initial weights of the image monitoring results. Starting from the root node, the feature values of the samples are compared with the values of the partitioning features corresponding to that node. Based on the different values, the samples are moved along the corresponding child nodes. The previous step is repeated, and the samples are moved to the next child node based on the partitioning features and values of the current node, until the leaf node is reached. The leaf node corresponds to the value of the initial weight of the image.
7. The subway construction safety monitoring system based on construction data analysis according to claim 3, characterized in that: The image analysis unit (203) uses image processing technology to analyze image features and obtain image monitoring results, specifically including: Vibration: Through image processing algorithms, signs of vibration amplitude were detected in the working area, and the vibration amplitude was inferred based on the degree of image change; Temperature: Temperature changes are inferred by using images from cameras to observe color variations, material expansion, and physical changes in objects in the environment. Settlement: Settlement is determined by detecting changes in the relative position of a specific marked object in an image.
8. The subway construction safety monitoring system based on construction data analysis according to claim 4, characterized in that: The method optimization unit (302) performs weighted average calculations based on the monitoring results and initial weights in the physical sensor analysis module (100) and the image analysis module (200). The weighted average of each monitoring parameter is calculated by using weights. For each parameter, the sum of the two sets of data multiplied by their corresponding weights is divided by the sum of the weights.
9. The subway construction safety monitoring system based on construction data analysis according to claim 4, characterized in that: The method optimization unit (302) compares the calculated weighted average result with the monitoring result in the result fusion feedback unit (301). When the difference between the weighted average result and the monitoring result exceeds a set threshold, the weights in the sensor data processing unit (102) and the image analysis unit (203) are adjusted, specifically including: A threshold is set. When the difference between the weighted fusion result and the monitoring result of the sensor data processing unit (102) or the image analysis unit (203) exceeds the threshold value, the weight needs to be adjusted. Based on the difference assessment, the weight is adjusted in increments of 0.1 or decreases until all the differences are eventually less than the set threshold.