Field monitoring method based on engineering informatization

Abstract

The invention discloses a field monitoring method based on engineering informatization, and the method comprises the steps: collecting the tightness data of a rope in real time through a tension sensor integrated on a protection rope, and transmitting the data to a field monitoring system; the method comprises the following steps: acquiring a worker operation picture shot by video monitoring equipment deployed in a construction site, and extracting posture features such as a body inclination angle and arm extension length of a worker in the picture; determining a real-time position coordinate of the worker according to the information of the positioning tag worn by the worker, and judging whether the coordinate is in a preset dangerous area electronic fence or not; according to the distance relation between the connection points of the worker and the protection rope, the swing track of the protection rope and the movement track calculated by the position coordinate change of the worker are analyzed, and whether the worker breaks away from the protection rope illegally or not is judged; if it is judged that the worker is separated from the protection rope, the position of a connection point of the protection rope suspected to fall off is marked in the monitoring picture according to the material characteristic and the color characteristic of the protection rope.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a field monitoring method based on engineering informatization. Background Art

[0002] In the engineering safety monitoring system, there is a key technical problem: how to accurately identify whether workers have detached from the protective rope in a complex construction environment. At present, the system collects rope data through tension sensors and analyzes it in combination with video monitoring and positioning tag information. However, due to factors such as noise interference, light changes and obstructions at the construction site, it is difficult to make accurate judgments based solely on these data. Especially when workers move quickly or work in a small space, the data collected by the sensor may fluctuate instantaneously, resulting in false alarms or missed alarms. At the same time, changes in the worker's posture will also affect the rope tension, making it difficult for the system to distinguish between normal operation and detachment from protection. Relying solely on the vibration frequency and motion trajectory of the rope, it is difficult to fully reflect the actual working status of the workers and the potential dangers of the surrounding environment. How to establish a dynamic and accurate protective rope detachment model is a technical challenge that needs to be solved urgently. Summary of the invention

[0003] The purpose of the present invention is to solve the problems in the prior art and to propose a field monitoring method based on engineering informatization.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions:

[0005] A field monitoring method based on engineering informationization, S1, collects rope tightness data in real time through a tension sensor integrated on the protection rope, and transmits the data to a field monitoring system;

[0006] S2, obtaining the worker working picture taken by the video surveillance equipment deployed at the construction site, and extracting the posture features of the worker's body tilt angle and arm extension length in the picture;

[0007] S3, determining the real-time location coordinates of the worker according to the positioning tag information worn by the worker, and judging whether the coordinates are within the preset dangerous area electronic fence;

[0008] S4, if the worker is in the danger zone, the body tilt angle, arm extension length posture characteristics and the tightness data of the safety rope are correlated and analyzed to obtain the distance relationship between the worker and the connection point of the safety rope;

[0009] S5, according to the distance relationship between the worker and the connection point of the protection rope, analyzing the swing trajectory of the protection rope and the motion trajectory calculated by the change of the worker's position coordinates, to determine whether the worker has violated the rules by leaving the protection rope;

[0010] S6, if it is determined that the worker has detached from the safety rope, the location of the connection point of the safety rope suspected to have detached is marked in the monitoring image according to the material characteristics and color characteristics of the safety rope;

[0011] S7, based on the worker's body swing frequency and the vibration frequency of the safety rope, combined with the location of the safety rope connection point, evaluates whether the worker's working movement range exceeds the protection range of the safety rope, calculates the worker's center of gravity offset distance, determines whether he is out of the rope, and issues a warning if so.

[0012] Compared with the prior art, the present invention has the following advantages:

[0013] The present invention firstly integrates tension sensors on the protective rope, analyzes the workers' postures in the video surveillance screen and locates the workers' positions, so as to monitor in real time whether the workers have detached from the protective rope. It comprehensively utilizes multi-source information such as tension sensor data, video surveillance data, and positioning tag data to improve the accuracy and comprehensiveness of the judgment. At the same time, the tightness of the protective rope, the posture characteristics of the workers and the position information are correlated and analyzed to judge the distance relationship between the worker and the connection point of the protective rope. In combination with the swing trajectory of the protective rope and the movement trajectory of the workers, it is evaluated whether the working action exceeds the protection range, and when it is found that the workers have detached from the protective rope, a warning can be issued in time, thereby realizing intelligent analysis and judgment of complex situations, effectively preventing high-altitude falling accidents, and improving working safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A flowchart of a field monitoring method based on engineering informationization proposed by the present invention;

[0015] Figure 2 A schematic diagram of a field monitoring method based on engineering informatization proposed by the present invention;

[0016] Figure 3 This is another schematic diagram of an on-site monitoring method based on engineering informatization proposed by the present invention. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0018] Reference Figure 1-3 The present embodiment provides a method for on-site monitoring based on engineering informationization, which specifically includes the following steps:

[0019] S1. The tension sensor integrated on the protective rope collects the rope tightness data in real time and transmits the data to the on-site monitoring system.

[0020] More specifically, the original data of rope tightness collected by the tension sensor integrated on the protective rope is obtained; according to the on-site ambient temperature value measured by the temperature sensor, the original data of rope tightness is corrected using a preset temperature compensation curve to obtain the corrected tension data; the corrected tension data is processed by median filtering, and the median filtering with a window size of 5 is used to remove outliers to obtain the tension data without outliers, and the ZigBee protocol is used for data transmission according to the set transmission priority, and the tension data without outliers is transmitted to the on-site monitoring system. The on-site monitoring system receives and analyzes the transmitted data through special software, and displays the tension data of each protective rope in real time.

[0021] Exemplarily, a tension sensor of model XL-100 is installed on the protective rope, and the acquisition frequency is set to 10Hz to collect rope tightness data in real time. At the same time, a PT100 temperature sensor is used to measure the on-site ambient temperature, and the preset temperature compensation curve y=ax^2+bx+c is used, where a, b, and c are constants determined according to the rope material, to correct the original tension data. The corrected tension data is filtered by a median filter algorithm with a window size of 5 to remove abnormal values, such as removing 150 from [100,102,150,103,105]. The ZigBee protocol is used for data transmission to the on-site monitoring system. The on-site monitoring system receives and parses the transmitted data through dedicated software, and displays the tightness data of each protective rope in real time on the graphical interface.

[0022] S2. Obtain the worker working picture taken by the video surveillance equipment deployed at the construction site, and extract the posture features of the worker's body tilt angle and arm extension length in the picture.

[0023] The original video frames collected by the video surveillance equipment at the construction site are obtained, and the original video frames are preprocessed. Gaussian filtering is used for image denoising, and brightness adjustment and contrast enhancement are performed through histogram equalization to obtain the preprocessed image. Human body detection is performed based on the preprocessed image. The YOLOv5s target detection algorithm is used to locate the position of the workers in the picture. The confidence threshold and non-maximum suppression threshold are set, such as the confidence threshold is 0.6 and the non-maximum suppression threshold is 0.5. The bounding box coordinates of each detected worker are output. For multi-person scenes, all detected human bounding boxes are retained. A single worker image is cropped from the preprocessed image according to the bounding box coordinates, and the BODY_25 human posture estimation algorithm is used to extract the worker's skeletal key points, including the two-dimensional coordinates of at least 25 key points such as the head, torso, and limbs. The key point detection confidence threshold is set to 0.5. If some key point detection fails, the linear interpolation method is used to estimate the position of the missing key points. Based on the coordinates of the skeleton key points, the worker's body inclination angle and arm extension length are calculated. The body inclination angle is calculated through the three key points of the top of the head, neck, and ankle, considering the three-dimensional space projection. The arm extension length is calculated through the spatial position relationship of the three key points of the shoulder, elbow, and wrist. For occlusion, the timing information is used to predict the key point position. For lighting changes, the adaptive threshold is used to adjust the detection parameters.

[0024] For example, the high-definition camera deployed at the construction site captures 1920x1080 resolution video images at a frame rate of 30fps. Each frame of the image is filtered and denoised by a 3x3 Gaussian kernel, and then the pixel intensity distribution is expanded from the original range of 100-150 to the full range of 0-255 through histogram equalization to enhance the image contrast. The preprocessed image is input into the YOLOv5s model, which is pre-trained on the COCO dataset and fine-tuned on the construction scene dataset. The detection confidence threshold is set to 0.6, the non-maximum suppression threshold is 0.5, and the bounding box coordinates of each worker are output. For multi-person scenes, if 5 workers are detected, 5 sets of bounding box coordinates are retained. The single worker image is cropped out according to the bounding box and sent to the BODY_25 model, which is based on the OpenPose architecture and can detect 25 key points of the human body, including 5 facial points, 6 torso points, and 14 limb points. The confidence threshold of key point detection is set to 0.5. If the confidence of a point is lower than the threshold, the position of the point in two adjacent frames is linearly interpolated and estimated. After obtaining the coordinates of the key points, the body inclination angle is calculated using the top of the head, neck, left and right ankles. Considering that the camera is installed at a height of 5 meters and the pitch angle of the viewing angle is about 30 degrees, the two-dimensional coordinates are converted into three-dimensional space coordinates for calculation. The arm extension length is obtained by summing the Euclidean distances of the shoulder, elbow, and wrist. For the case of being blocked by other objects, the Kalman filter is used to predict the trajectory of the key points in the three frames before the blockage and estimate the position of the current frame. When the lighting changes, the detection threshold of YOLOv5s is dynamically adjusted by calculating the image brightness mean in real time, reducing it to 0.5 in the shadow area and increasing it to 0.7 in the bright light area to adapt to different lighting conditions.

[0025] S3. Determine the real-time position coordinates of the worker according to the positioning tag information worn by the worker, and judge whether the coordinates are within the preset dangerous area electronic fence.

[0026] The wireless signal emitted by the positioning tag worn by the worker is obtained every second. The wireless signal is received by the signal receivers distributed at the construction site; the worker ID and initial position data are obtained based on the wireless signal analysis; the three receivers with the strongest signals are selected, and the real-time two-dimensional coordinates of the worker are calculated using the known position of the receiver and the time difference of the signal arrival; the height value is estimated by triangulation based on the signal strength and known height information of multiple receivers to obtain the three-dimensional spatial coordinates of the worker; the three-dimensional spatial coordinates are processed by Kalman filtering, and the Kalman filtering processing includes setting the state vector to include position and velocity, the observation vector to be position, and the process noise covariance matrix and the observation noise covariance matrix to be determined according to the actual measurement error; the coordinate jitter caused by signal interference and multipath effect is eliminated by Kalman filtering processing to obtain a smoothed sequence of worker position coordinates; the latest coordinate point in the smoothed sequence of worker position coordinates is converted to a local coordinate system with a fixed point at the construction site as the origin through affine transformation; the ray method is used to determine whether the latest coordinate point is within the preset dangerous area electronic fence polygon, and the ray method includes making rays from the judgment point to any direction, and calculating the number of intersections with the polygon boundary to determine whether the point is within the polygon.

[0027] For example, 20 signal receivers are deployed at the construction site, and the positioning tags worn by the workers send a 2.4GHz wireless signal containing the worker ID and initial position once a second. After receiving the signal, the signal receiver selects the three receivers with the highest signal strength, assuming that their coordinates are (0,0,0), (10,0,0), (0,10,0), and calculate the two-dimensional coordinates of the worker through the signal arrival time difference, such as Δt1 = 0.00000005 seconds and Δt2 = 0.00000008 seconds. Combined with the signal strength of all receivers such as -60dBm, -62dBm, -65dBm, etc. and the known height information, the height value is estimated using triangulation, and the three-dimensional spatial coordinates of the worker are finally obtained (5.2, 4.8, 1.7). Kalman filtering is applied to these raw coordinate data, with the state vector set to [x, y, z, vx, vy, vz], the observation vector to [x, y, z], the process noise covariance matrix Q and the observation noise covariance matrix R set according to the actual measurement error, such as Q = diag (0.1, 0.1, 0.1, 0.01, 0.01, 0.01), R = diag (0.2, 0.2, 0.2), and a smoothed position coordinate sequence is obtained after filtering. The latest smoothed coordinates (5.3, 4.9, 1.8) are transformed into a local coordinate system with the southwest corner of the construction site as the origin through an affine transformation matrix, and (105.3, 94.9, 1.8) are obtained. The preset danger zone electronic fence is a polygon with vertex coordinates [(100,90,0),(110,90,0),(110,100,0),(100,100,0)]. The ray method is used to determine whether the point is within the polygon. Draw a ray from the point (105.3,94.9,1.8) in the positive direction of the x-axis. The number of intersections with the polygon boundary is an odd number 1, and the point is determined to be within the polygon.

[0028] S4. If the worker is in the danger zone, the body inclination angle, arm extension length posture characteristics and the tightness data of the safety rope are correlated and analyzed to obtain the distance relationship between the worker and the connection point of the safety rope.

[0029] The data of workers located within the danger polygon are obtained from the real-time monitoring database, and the worker data includes identity identification, location coordinates, body inclination angle, and arm extension length, wherein the body inclination angle includes the torso inclination angle, the arm extension angle, the angle of the legs standing or squatting, and the head orientation; the position of the worker's center of gravity is calculated based on the worker data, including using a simplified human skeleton model to determine the position of the joint points, and combining the body inclination angle and the arm extension length to obtain the center of gravity offset; the tightness data of the protective rope is obtained, a mapping function is established based on the tightness data, and the actual length and shape of the protective rope are determined through the mapping function;

[0030] For example, at a construction site, the real-time monitoring database uses R-tree spatial indexing technology to define the danger zone as a polygon [(0,0), (10,0), (10,10), (0,10)], and quickly finds worker A in the area, whose identity is worker_001, location coordinates (5,5,2), body tilt angle of 30 degrees, and arm extension length of 0.8 meters. At the same time, the tightness data of the worker's protective rope is obtained from the protective rope monitoring subsystem, and the tension value is 200N.

[0031] S401, analyzes the worker's body posture, including the torso inclination angle, the length and angle of the arms, the angle of the legs when standing or squatting, and the head direction, analyzes the state of the safety rope, calculates the bending degree and slackness of the rope through the tension sensor data, and estimates the relative position of the worker and the connection point of the safety rope.

[0032] The worker's body posture characteristic parameters are obtained from the real-time monitoring database. The real-time data of multiple tension sensors on the protective rope are obtained, and the tension sensor data points are fitted using the cubic spline interpolation method to obtain a continuous tension distribution curve. According to the worker's body posture characteristic parameters and the tension distribution curve, a three-dimensional spatial relationship model of the worker-protection rope is established to determine the relative position coordinates of the connection point between the protective rope and the worker.

[0033] Exemplarily, real-time data of multiple tension sensors on the protective rope are obtained, and the data points of the tension sensors are fitted using the cubic spline interpolation method to obtain a continuous tension distribution curve, and the bending degree of the rope is calculated. Combined with the stress-strain characteristic curve of the rope material pre-measured in the laboratory, the actual length and slack of the rope are estimated. According to the characteristic parameters of the worker's posture and the state data of the protective rope, a three-dimensional spatial relationship model of the worker-protection rope is established, and the three-dimensional curve of the protective rope is fitted using the least squares method. The nearest point is calculated as the connection point in combination with the worker's position to obtain the relative position coordinates of the connection point between the protective rope and the worker. A tension sensor is installed every 0.5m on the protective rope, for a total of 10, and tension data [100N, 105N, 110N, 108N, 106N, 104N, 102N, 101N, 100N, 99N] is collected in real time. These data points are fitted using the cubic spline interpolation method to obtain a continuous tension distribution curve f(x)=ax^3+bx^2+cx+d, where a, b, c, d are fitting coefficients. According to the pre-measured stress-strain curve ε=0.001σ+0.0001σ^2, where ε is strain and σ is stress, the actual length of the rope is estimated to be 5.2m and the slack is 4%. The three-dimensional curve of the protective rope g(t)=(x(t), y(t), z(t)) is fitted using the least squares method, and the connection point coordinates (2.1, 3.2, 1.6) are calculated in combination with the worker position (2, 3, 1.5).

[0034] S402, establish an association model between the worker's posture and the state of the protective rope, match the spatial coordinates of various parts of the worker's body with the coordinates of the protective rope connection point, calculate the offset distance of the worker's center of gravity position relative to the protective rope connection point, and obtain the distance between the worker and the protective rope connection point.

[0035] The three-dimensional spatial coordinates of each part of the worker's body collected by the depth camera and the real-time coordinates of the protective rope connection point obtained by the inertial measurement unit are obtained. The rigid body transformation matrix is ​​used to convert the inertial measurement unit coordinates to the depth camera coordinate system, and the two sets of coordinate data are unified into the same coordinate system. According to the spatial coordinates of each part of the worker's body, a 15-node simplified human body model is used to calculate the center of gravity of the worker according to the mass ratio of each part. At the same time, based on the tension sensor data of the protective rope, the finite element method is used to solve the elastic catenary model, and the spatial form of the protective rope is estimated according to the tension sensor data of the protective rope. A model for associating the worker's posture with the state of the protective rope is established, and a weight function based on distance and angle is constructed to quantify the influence of the worker's posture on the state of the protective rope. The KD tree structure is used to optimize the nearest neighbor search process, and the spatial coordinates of each part of the worker's body are paired with the discrete points on the protective rope to form a spatial correspondence between the worker and the protective rope. The three-dimensional offset vector of the worker's center of gravity relative to the connection point of the protective rope is calculated, and the absolute value of the offset distance is calculated using the Euclidean distance formula. Combined with the offset direction, the distance relationship between the worker and the connection point of the protective rope is obtained. The Kalman filter is used to predict the worker's movement trajectory and update the distance relationship in real time. The Bayesian estimation method is used to fuse multi-source data from depth cameras, inertial measurement units, and tension sensors to improve the accuracy of distance relationship calculation.

[0036] For example, an Intel Real Sense D455 depth camera is deployed at the construction site to collect 1280x720 resolution depth images at a frequency of 30 frames per second, and an Xsens MTi-300 AHRS inertial measurement unit is installed at the connection point of the protective rope with a sampling rate of 100 Hz. A 15-node simplified human body model is used, in which the head accounts for 7%, the trunk accounts for 43%, the upper arms account for 3%, the forearms account for 2%, the thighs account for 12%, the calves account for 5%, and the feet account for 1.5%. The center of gravity of the worker is calculated to be (2.1, 1.5, 0.9) meters. A tension sensor is installed every 0.5 meters on the protective rope, and the measured tension data is [120N, 115N, 110N, 105N, 100N]. The fourth-order Runge-Kutta method is used to solve the elastic catenary equation y=acosh(x / a), and the protective rope spatial morphology curve y=2.5cosh(x / 2.5) is obtained. The distance weight function w_d=1 / (1+d^2) and the angle weight function w_θ=cos(θ) are constructed, where d is the Euclidean distance and θ is the angle. The weight of the worker's posture on the state of the protective rope is obtained comprehensively, w=0.6w_d+0.4w_θ. The KD tree is used to perform nearest neighbor matching between the 15 nodes of the worker and the 50 discrete points of the protective rope, and the average search time is 0.5 milliseconds. The offset vector of the worker's center of gravity (2.1, 1.5, 0.9) relative to the safety rope connection point (0, 0, 3) is calculated to be (2.1, 1.5, -2.1), and the Euclidean distance is 3.4 meters. Using the Kalman filter, the state vector is [x, y, z, vx, vy, vz], the observation vector is [x, y, z], and the predicted position of the worker 0.1 seconds later is (2.15, 1.52, 0.91). Finally, through the Bayesian estimation method, the depth image, inertial data and tension data are integrated to obtain the posterior probability distribution of the distance between the worker and the safety rope connection point, with a mean of 3.38 meters and a standard deviation of 0.05 meters.

[0037] S5. According to the distance relationship between the worker and the connection point of the safety rope, analyze the swing trajectory of the safety rope and the motion trajectory calculated by the change of the worker's position coordinates to determine whether the worker has violated the regulations by leaving the safety rope.

[0038] The acceleration and angular velocity data of the protection rope connection point collected in real time by the inertial measurement unit are used. The fourth-order Runge-Kutta method is used to perform numerical integration based on the acceleration and angular velocity data to obtain the protection rope swing trajectory, and the Kalman filter is used to correct the accumulated error. At the same time, the depth camera is used to obtain the three-dimensional position coordinate sequence of the worker, and the position difference at adjacent moments is calculated based on the three-dimensional position coordinate sequence to obtain the worker's motion trajectory; the protection rope swing trajectory and the worker's motion trajectory are aligned in time series, and the dynamic time warping algorithm is used to calculate the similarity score of the protection rope swing trajectory and the worker's motion trajectory. The window size is set to 50 frames, the step size is 10 frames, the Euclidean distance is used as the cost function, and the similarity threshold is set to determine whether the worker is Whether to follow the swing of the safety rope, calculate the rate of change of the Euclidean distance between the worker's position and the connection point of the safety rope over time, and detect the mutation point of the change rate through the sliding window method. The window size is set to 1 second, such as 30 frames, and the step size is 0.1 second, such as 3 frames. If the change rate exceeds the preset threshold, it is marked as a potential separation event; according to the similarity score and the change rate, the support vector machine classifier is used to classify the worker's behavior, and the SVM model is trained through historical annotation data. The RBF kernel function is used, and the hyperparameters C and gamma are optimized through grid search to output the judgment result of whether the violation of separation from the safety rope occurs. When the distance between the worker and the connection point of the safety rope exceeds 2 meters and the duration exceeds 3 seconds, it is judged as a violation of separation from the safety rope.

[0039] For example, an XsensMTi-G-710 inertial measurement unit is deployed at the construction site, and the sampling rate is set to 100Hz to collect acceleration and angular velocity data of the connection point of the protective rope. The fourth-order Runge-Kutta method is used for numerical integration, and the step size is set to 0.01 seconds to calculate the swing trajectory of the protective rope. The Kalman filter is applied, the state vector contains position and velocity, the process noise covariance matrix Q = diag(0.01, 0.01, 0.01, 0.1, 0.1, 0.1), and the observation noise covariance matrix R = diag(0.1, 0.1, 0.1) is used to correct the cumulative error. At the same time, the Intel Real Sense D455 depth camera is used with a frame rate of 30fps and a resolution of 1280x720 to obtain the three-dimensional coordinate sequence of the workers. The two trajectories are dynamically time warped with a window size of 50 frames and a step size of 10 frames. The Euclidean distance is used as the cost function, and the similarity score is calculated to be 0.85, with a full score of 1. The similarity threshold is set to 0.7 to determine whether the worker follows the swing of the safety rope. A sliding window of 1 second in size and a step size of 0.1 second are used to calculate the rate of change of the distance between the worker and the connection point, and the preset threshold is 0.5 m / s. The support vector machine classifier is trained using the LibSVM library, and the RBF kernel function is used. The optimal parameters C=10 and gamma=0.1 are obtained through five-fold cross validation and grid search optimization. The input feature vector contains the trajectory similarity score and the distance change rate, and the output violation probability is 0.92. According to the preset rules, when the distance exceeds 2 meters and lasts for more than 3 seconds, the system determines that the worker has violated the rules by leaving the safety rope. When a signal abnormality is detected, such as no valid data within 3 seconds, the linear interpolation method is used to estimate the missing data to ensure continuous operation of the system.

[0040] S5 also includes calculating the real-time deviation between the worker's movement trajectory and the swing trajectory of the protective rope based on the distance relationship between the worker and the connection point of the protective rope, setting a safety distance threshold between the worker and the connection point of the protective rope, and continuously monitoring whether the actual distance exceeds this threshold.

[0041] The three-dimensional coordinate data of the connection point of the protective rope is collected by the inertial measurement unit, and the real-time three-dimensional coordinates of the worker are obtained by the depth camera. The coordinates of the inertial measurement unit are converted to the camera coordinate system using the rigid body transformation matrix. The Euclidean distance between the worker and the connection point of the protective rope is calculated every 100 milliseconds according to the two sets of coordinate data, and the coordinates of the connection point of the protective rope and the worker in the unified coordinate system are obtained; the coordinates of the connection point of the protective rope and the coordinates of the worker are predicted and smoothed by the Kalman filter. The state vector of the Kalman filter contains position and velocity. The process noise covariance matrix and the observation noise covariance matrix are set according to the actual measurement error to reduce the influence of measurement noise and obtain more accurate motion trajectory and swing trajectory; the real-time deviation between the worker's motion trajectory and the swing trajectory of the protective rope is calculated by the dynamic time warping algorithm. The window size of the dynamic time warping algorithm is set to the preset time length and the step size is set to the preset value; if the real-time deviation exceeds the preset safety distance threshold, the safety warning mechanism is triggered, and the safety distance threshold is dynamically adjusted by the fuzzy logic controller according to the working environment and task type. For example, the window size is set to 2 seconds and the step size is set to 0.5 seconds. The deviation value is compared with the preset safety distance threshold to determine whether there is potential danger. According to the working environment and task type, a fuzzy logic controller is used to dynamically adjust the safety distance threshold. The initial threshold is set to 2 meters, and a sliding time window is set to continuously monitor the relationship between the actual distance and the threshold. When the distances of five consecutive sampling points exceed the threshold, the safety warning mechanism is triggered.

[0042] For example, an XsensMTi-G-710 inertial measurement unit is deployed at the construction site to collect data on the connection points of the protective rope, with a sampling rate of 100Hz. At the same time, an IntelRealSenseD455 depth camera is used with a frame rate of 30fps and a resolution of 1280x720 to obtain the real-time three-dimensional coordinates of the workers. The Euclidean distance is calculated every 100ms, d = sqrt((x1-x2)^2+(y1-y2)^2+(z1-z2)^2). A Kalman filter is applied, with the state vector X = [x, y, z, vx, vy, vz], the observation vector Z = [x, y, z], the process noise covariance Q = diag(0.01, 0.01, 0.01, 0.1, 0.1), and the observation noise covariance R = diag(0.1, 0.1, 0.1), to smooth the coordinates. The dynamic time warping algorithm is used with a window size of 60 frames and 2 seconds and a step size of 15 frames and 0.5 seconds to calculate the trajectory deviation. The Mamdani fuzzy logic controller is used to dynamically adjust the safety threshold. The input variables are the work difficulty 1-10 and the environmental complexity 1-10. The output is the threshold adjustment amount -0.5m to +0.5m, and the initial threshold is 2m. A sliding window of 0.5 seconds and 5 sampling points is set. If the distance of 5 consecutive points exceeds the current threshold, such as [2.1, 2.2, 2.15, 2.3, 2.25]m, an early warning is triggered.

[0043] S5 also includes analyzing the acceleration and direction changes of the worker's movement trajectory, identifying abnormal movements that lead to sudden detachment from the safety rope, evaluating the correlation between the change in the safety rope tension and the change in the worker's position, and detecting whether there is an abnormal situation of sudden tension reduction.

[0044] The worker movement data is obtained, and the worker movement data is collected by an inertial measurement unit with a sampling frequency of 100 Hz; the three-dimensional acceleration and angular velocity are calculated based on the worker movement data; the three-dimensional acceleration and angular velocity are smoothed by the Kalman filter algorithm, the state vector contains position, velocity and acceleration, and the observation vector is position and acceleration, and the acceleration and direction change sequence of the worker's movement trajectory is obtained; for the acceleration and direction change sequence, the sliding window method is used to segment it, and the window size is set to 1 second, 100 samples, and the step size is 0.1 second, 10 samples. The acceleration peak, direction change rate and posture angle are selected from the segmented sequence as features; the support vector machine algorithm is used to train the abnormal action recognition model, and the abnormal action recognition model is used to detect whether the worker has made an abnormal action in real time. The worker identifies an abnormal action that may lead to a sudden detachment from the safety rope; obtains the safety rope tension data, which is collected by a tension sensor with a sampling frequency of 50 Hz; calculates the tension change rate and the position change rate based on the safety rope tension data and the worker's position coordinates; uses the Pearson correlation coefficient to evaluate the correlation between the tension change rate and the position change rate; if the tension change rate exceeds the preset threshold and the correlation is lower than the preset correlation coefficient threshold, it is judged that an abnormal situation exists, such as setting the tension change threshold to 3 times the standard deviation of the mean and the correlation coefficient threshold to 0.5 based on the abnormal action recognition results and correlation analysis, using statistical analysis of historical data; when the tension suddenly decreases and the correlation with the position change is significantly reduced, it is judged that an abnormal situation exists, triggering the alarm mechanism and recording the abnormal event.

[0045] For example, at the construction site, workers wear XsensMTi-G-710 inertial measurement units, with a sampling frequency of 100Hz, to collect triaxial acceleration and angular velocity data. A Kalman filter is used for data smoothing, with state vector X = [x, y, z, vx, vy, vz, ax, ay, az], observation vector Z = [x, y, z, ax, ay, az], process noise covariance Q = diag(0.01, 0.01, 0.01, 0.1, 0.1, 1, 1, 1), and observation noise covariance R = diag(0.1, 0.1, 0.1, 0.5, 0.5, 0.5). The sliding window size is set to 100 samples, 1 second, with a step size of 10 samples, 0.1 second, to extract the characteristic acceleration peak, direction change rate, and attitude angle. The support vector machine model was trained using LibSVM, with RBF as the kernel function, parameters C=10, gamma=0.1, and optimized by 5-fold cross validation. An S-type tension sensor HBMS9M with a sampling frequency of 50Hz and a range of 0-5kN was installed on the protective rope. The tension change rate ΔT / Δt and position change rate Δs / Δt of 50 samples within 1 second were calculated, and the Pearson correlation coefficient r=cov(ΔT / Δt,Δs / Δt) / (σ(ΔT / Δt)σ(Δs / Δt)) was used to evaluate the correlation. Based on historical data, the tension change threshold was set to μ+3σ=500N / s, and the correlation coefficient threshold was 0.5. When an abnormal action was detected (SVM output was 1), and the tension change rate was >500N / s, and the correlation coefficient was <0.5, the system determined that it was a sudden detachment, and sent an alarm to the monitoring center through the MQTT protocol, including the worker ID, timestamp, and abnormality type.

[0046] S6. If it is determined that the worker has detached from the safety rope, the location of the connection point of the safety rope that is suspected to have detached is marked in the monitoring image based on the material characteristics and color characteristics of the safety rope.

[0047] After determining that the worker has detached from the safety rope, obtain the real-time image of the high-definition surveillance camera, preprocess the real-time image to obtain the preprocessed image, including using Gaussian filtering for denoising, using adaptive histogram equalization for illumination equalization and contrast enhancement, so as to improve the recognizability of the safety rope in the picture; establish a texture descriptor based on the grayscale co-occurrence matrix according to the preprocessed image, convert the preprocessed image from the RGB color space to the HSV color space, and obtain the HSV image; use the threshold segmentation algorithm to process the HSV image, extract the area that meets the characteristics of the safety rope, and obtain the preliminary contour image of the safety rope; use morphological operations such as opening operation to remove noise and closing operation to fill small holes, and use the Canny edge detection algorithm to refine and optimize the preliminary contour image, extract the precise edge and endpoint of the safety rope, and obtain the edge endpoint image as the candidate area of ​​the suspected detached connection point; use the extracted edge and endpoint information as input, use the YOLOv5 target detection algorithm trained based on the self-built safety rope connection point data set, identify and locate the suspected detached safety rope connection point in the candidate area, and mark it with a specific color and shape mark in the monitoring screen to obtain the marked monitoring screen.

[0048] For example, a 4K resolution HD camera with a frame rate of 30fps is deployed at the construction site. When the system determines that the worker is out of the protective rope, the image processing process is immediately triggered. First, a 5x5 Gaussian filter kernel is applied to the captured image for denoising, and then an 8x8 block adaptive histogram equalization is used to improve the image contrast. Next, the grayscale co-occurrence matrix of the image is calculated, and four features of contrast, energy, homogeneity and correlation are extracted as texture descriptors of the protective rope material. The image is converted from RGB to HSV color space, and the suspected protective rope area is extracted using threshold segmentation according to the preset protective rope color range, for example, hue H20-40, saturation S0.6-1.0, and brightness V0.4-0.8. The 3x3 structural element is applied to the segmentation result for opening and closing operations, removing noise points less than 50 pixels and filling holes less than 30 pixels. Then the Canny edge detection algorithm is used to extract the contour of the protective rope with a low threshold of 50 and a high threshold of 150. The processed edge image is fed into the pre-trained YOLOv5s model, which is trained based on 1,000 labeled images of rope connection points, with an mAP@0.5 of 0.92, to detect possible connection point locations. Finally, the detected connection points are annotated on the original image with a red rectangular frame with a line width of 5 pixels, and the confidence score is annotated in the upper left corner of the frame.

[0049] S7. Based on the worker's body swing frequency and the vibration frequency of the safety rope, combined with the location of the safety rope connection point, evaluate whether the worker's working movement range exceeds the protection range of the safety rope, calculate the worker's center of gravity offset distance, and determine whether he is out of the range. If so, issue a warning.

[0050] Acceleration data is collected through an inertial measurement unit worn on the worker, and a 1024-point fast Fourier transform algorithm is used to perform spectrum analysis on the data of a 2-second sliding window based on the acceleration data to obtain the main frequency components of the worker's body swing; the vibration data of the tension sensor on the protective rope is obtained, and the vibration frequency of the protective rope is calculated based on the vibration data to obtain the ratio of the worker's body swing frequency to the protective rope vibration frequency, and set a safe frequency ratio range; based on the worker's motion trajectory captured by the camera and the pre-calibrated position of the protective rope connection point, a cubic spline function is used to fit the worker's motion trajectory curve, and the maximum amplitude and average amplitude of the operating action are calculated based on the motion trajectory curve, and compared with the preset protective range of the protective rope, a spherical space with the connection point as the center and a radius of 80% of the length of the protective rope, to determine whether the amplitude of the operating action exceeds the preset The protection range of the protective rope is obtained by simplifying the human biomechanical model and inertial measurement unit data with 15 nodes, setting the mass proportion of each part according to the human anatomical data, estimating the mass distribution and spatial position of each part of the worker's body, calculating the center of gravity position, and using the Kalman filter algorithm to smooth the center of gravity trajectory. The state variables are the three-dimensional coordinates and speed of the center of gravity, and the observed variable is the estimated center of gravity position. The offset distance of the worker's center of gravity relative to the connection point of the protective rope is obtained; the calculated frequency ratio, operating motion amplitude, and center of gravity offset distance are compared with the dynamic safety threshold set based on historical data statistical analysis and expert experience. If the threshold is exceeded, it is determined to be out of state, triggering the warning mechanism, and sending a warning message to the smart device worn by the worker through the wireless communication module. At the same time, the position and out of state of the worker are marked on the display screen of the monitoring center.

[0051] For example, at the construction site, workers wear XsensMTi-G-710 inertial measurement units with a sampling rate of 100Hz to collect three-axis acceleration data. Use 1024-point FFT to perform spectrum analysis on the data of 200 samples in a 2-second sliding window, and extract the frequency with the largest amplitude in the range of 0.5-5Hz as the worker's body swing frequency. At the same time, an HBMS9M tension sensor is installed on the protective rope with a sampling rate of 50Hz, and the vibration frequency of the protective rope is calculated by the same FFT method. Calculate the frequency ratio and set the safety range to 0.8-1.2. Use IntelRealSenseD455 depth camera to collect the worker's motion trajectory at 30fps, use scipy.interpolate.splrep function to fit the cubic spline curve, and calculate the maximum amplitude. Assuming that the protective rope is 5 meters long, the protection range is a spherical space with a radius of 4 meters. Using a 15-node human body model, the center of gravity position is estimated with a mass ratio of 7% for the head, 43% for the trunk, 3% for each upper arm, 2% for each forearm, 12% for each thigh, 5% for each calf, and 1.5% for each foot. Applying a Kalman filter, the state vector X = [x, y, z, vx, vy, vz], the observation vector Z = [x, y, z], the process noise covariance Q = diag (0.01, 0.01, 0.01, 0.1, 0.1, 0.1), and the observation noise covariance R = diag (0.1, 0.1, 0.1), the center of gravity offset is calculated. Based on 1000 historical data, the dynamic threshold frequency ratio is set to exceed ±20%, the movement amplitude exceeds 3.5 meters, or the center of gravity offset exceeds 3 meters to determine the separation. If triggered, a warning is sent to the worker's smart bracelet through the LoRaWAN network, and the worker's position is marked with a red icon on the 55-inch display screen of the monitoring center.

[0052] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A field monitoring method based on engineering informationization, characterized in that: The following steps are involved: S1, collects rope tightness data in real time through the tension sensor integrated on the protection rope, and transmits the data to the on-site monitoring system; S2, obtaining the worker working picture taken by the video surveillance equipment deployed at the construction site, and extracting the posture features of the worker's body tilt angle and arm extension length in the picture; S3, determining the real-time location coordinates of the worker according to the positioning tag information worn by the worker, and judging whether the coordinates are within the preset dangerous area electronic fence; S4, if the worker is in the danger zone, the body tilt angle, arm extension length posture characteristics and the tightness data of the safety rope are correlated and analyzed to obtain the distance relationship between the worker and the connection point of the safety rope; S5, according to the distance relationship between the worker and the connection point of the protection rope, analyzing the swing trajectory of the protection rope and the motion trajectory calculated by the change of the worker's position coordinates, to determine whether the worker has violated the rules by leaving the protection rope; S6, if it is determined that the worker has detached from the safety rope, the location of the connection point of the safety rope suspected to have detached is marked in the monitoring image according to the material characteristics and color characteristics of the safety rope; S7, based on the worker's body swing frequency and the vibration frequency of the safety rope, combined with the location of the safety rope connection point, evaluates whether the worker's working movement range exceeds the protection range of the safety rope, calculates the worker's center of gravity offset distance, determines whether he is out of the rope, and issues a warning if so.

2. The on-site monitoring method based on engineering informationization according to claim 1 is characterized in that: The S1 includes: Obtaining raw data on the tightness of the rope collected by a tension sensor integrated on the protective rope; According to the on-site ambient temperature value measured by the temperature sensor, the original data of the tightness of the rope is corrected using a preset temperature compensation curve to obtain corrected tension data; The corrected tension data is subjected to median filtering, and a filter window size is preset to obtain tension data with outliers removed.

3. The on-site monitoring method based on engineering informationization according to claim 1 is characterized in that: The S2 includes: Obtaining original video frames captured by video surveillance equipment at the construction site, preprocessing the original video frames, performing image denoising using Gaussian filtering, and performing brightness adjustment and contrast enhancement through histogram equalization to obtain a preprocessed image; Performing human body detection according to the preprocessed image, locating the position of the worker in the picture, setting a confidence threshold and a non-maximum suppression threshold, and outputting the bounding box coordinates of each detected worker; cropping a single worker image from the preprocessed image according to the bounding box coordinates, and extracting the worker's skeletal key points; If some key points fail to be detected, the positions of the missing key points are estimated; Based on the coordinates of the skeleton key points, the worker's body inclination angle and arm extension length are calculated.

4. The on-site monitoring method based on engineering informationization according to claim 1 is characterized in that: The S3 includes: Acquire a wireless signal emitted every second by the positioning tag worn by the worker, wherein the wireless signal is received by a signal receiver distributed at the construction site; Obtaining worker ID and initial location data according to the wireless signal analysis; Selecting three receivers with the strongest signals and calculating the real-time two-dimensional coordinates of the workers using the known positions of the receivers and the time difference of arrival of the signals; Combining the signal strength of multiple receivers with known height information, the height value is estimated to obtain the three-dimensional spatial coordinates of the worker; Processing the three-dimensional space coordinates to eliminate coordinate jitter caused by signal interference and multipath effects, and obtaining a smoothed worker position coordinate sequence; Converting the latest coordinate point in the smoothed worker position coordinate sequence into a local coordinate system with a fixed point at the construction site as the origin through affine transformation; Determine whether the latest coordinate point is within the preset dangerous area electronic fence polygon.

5. The on-site monitoring method based on engineering informationization according to claim 1 is characterized in that: The S4 includes: firstly, obtaining worker data located in the danger polygon from the real-time monitoring database, wherein the worker data includes identity identification, location coordinates, body inclination angle and arm extension length, wherein the body inclination angle includes trunk inclination angle, arm extension angle and angle of standing or squatting legs and head orientation; Calculating the center of gravity of the worker according to the worker data, including using a simplified human skeleton model to determine the location of joint points, and combining the body inclination angle and the arm extension length to obtain the center of gravity offset; The tightness data of the protective rope is obtained, a mapping function is established according to the tightness data, and the actual length and shape of the protective rope are determined by the mapping function.

6. The on-site monitoring method based on engineering informationization according to claim 5 is characterized in that: The S4 specifically comprises the following steps: S401, analyzing the body posture of the worker, including the inclination angle of the torso, the length and angle of the arms, the angle of the legs when standing or squatting, and the head orientation, analyzing the state of the safety rope, calculating the bending degree and slackness of the rope through the tension sensor data, and estimating the relative position of the connection point between the worker and the safety rope; S402, establish an association model between the worker's posture and the state of the protective rope, match the spatial coordinates of various parts of the worker's body with the coordinates of the protective rope connection point, calculate the offset distance of the worker's center of gravity position relative to the protective rope connection point, and obtain the distance between the worker and the protective rope connection point.

7. The on-site monitoring method based on engineering informationization according to claim 6 is characterized in that: The S401 specifically includes: Obtaining the worker's body posture characteristic parameters from the real-time monitoring database; And obtain real-time data of multiple tension sensors on the protection rope, fit the data points of the tension sensors, and obtain a continuous tension distribution curve; According to the body posture characteristic parameters of the worker and the tension distribution curve, a three-dimensional spatial relationship model of the worker and the protection rope is established to determine the relative position coordinates of the connection point between the protection rope and the worker; The S402 specifically includes: Obtain the three-dimensional spatial coordinates of various parts of the worker's body collected by the depth camera and the real-time coordinates of the connection points of the protective rope obtained by the inertial measurement unit, and use the rigid body transformation matrix to transform the inertial measurement unit coordinates into the depth camera coordinate system; According to the spatial coordinates of each part of the worker's body, a 15-node simplified human body model is used to calculate the center of gravity of the worker, and the center of gravity is determined by the mass ratio of each part; estimating the spatial shape of the protective rope according to the protective rope tension sensor data; Establishing a model for associating worker posture with the state of the protective rope, constructing a weight function based on distance and angle, and quantifying the influence of the worker posture on the state of the protective rope; Calculate the three-dimensional offset vector of the worker's center of gravity relative to the connection point of the protective rope, calculate the absolute value of the offset distance, and obtain the distance relationship between the worker and the connection point of the protective rope; The worker's movement trajectory is predicted, and the distance relationship is updated in real time.

8. The on-site monitoring method based on engineering informationization according to claim 1 is characterized in that: The S5 includes: Acquire the acceleration and angular velocity data of the connection point of the protection rope collected in real time by the inertial measurement unit, and obtain the swing trajectory of the protection rope by numerical integration according to the acceleration and angular velocity data; The worker's three-dimensional position coordinate sequence is obtained through a depth camera, and the position difference at adjacent moments is calculated according to the three-dimensional position coordinate sequence to obtain the worker's movement trajectory; Performing time series alignment on the swing trajectory of the protection rope and the movement trajectory of the worker, and calculating a similarity score between the swing trajectory of the protection rope and the movement trajectory of the worker; Calculate the rate of change of the Euclidean distance between the worker's position and the connection point of the protective rope over time, and detect the mutation point of the change rate; According to the similarity score and the change rate, a support vector machine classifier is used to classify the worker's behavior, and if the worker's behavior is classified as being out of the safety rope, it is determined to be a violation; It also includes: calculating the real-time deviation between the worker's movement trajectory and the protection rope's swing trajectory based on the distance relationship between the worker and the protection rope's connection point, setting a safety distance threshold between the worker and the protection rope's connection point, and continuously monitoring whether the actual distance exceeds this threshold; Analyze the acceleration and direction changes of the worker's movement trajectory, identify abnormal movements that lead to sudden detachment from the safety rope, evaluate the correlation between the change in the safety rope tension and the change in the worker's position, and detect whether there is an abnormal situation of sudden tension reduction; The method calculates the real-time deviation between the worker's motion trajectory and the swing trajectory of the protective rope according to the distance relationship between the worker and the connection point of the protective rope, sets a safety distance threshold between the worker and the connection point of the protective rope, and continuously monitors whether the actual distance exceeds the threshold, specifically including: Obtaining the three-dimensional coordinate data of the protective rope connection point and the real-time three-dimensional coordinates of the worker, wherein the three-dimensional coordinate data is collected by an inertial measurement unit and the real-time three-dimensional coordinates are obtained by a depth camera; The inertial measurement unit coordinates are converted to the camera coordinate system according to the rigid body transformation matrix to obtain the protective rope connection point coordinates and the worker coordinates in the unified coordinate system; A Kalman filter is used to predict and smooth the coordinates of the protective rope connection point and the worker coordinates, wherein the state vector of the Kalman filter includes position and velocity, and the process noise covariance matrix and the observation noise covariance matrix are set according to actual measurement errors; Calculating the real-time deviation between the worker's motion trajectory and the protection rope's swing trajectory; If the real-time deviation exceeds the preset safety distance threshold, a safety warning mechanism is triggered; The analysis of the acceleration and direction changes of the worker's motion trajectory, identification of abnormal movements that lead to sudden separation from the safety rope, evaluation of the correlation between the change in the safety rope tension and the change in the worker's position, and detection of whether there is an abnormal situation of a sudden decrease in tension specifically includes: Acquiring worker movement data, where the worker movement data is collected by an inertial measurement unit with a sampling frequency of 100 Hz; Calculating three-dimensional acceleration and angular velocity based on the worker movement data; Smoothing the three-dimensional acceleration and angular velocity to obtain a sequence of acceleration and direction changes of the worker's motion trajectory; For the acceleration and direction change sequence, segmenting is performed by a sliding window method; Selecting acceleration peak value, direction change rate and attitude angle as features from the segmented sequence; Training an abnormal action recognition model, wherein the abnormal action recognition model is used to detect in real time whether a worker has made an abnormal action that causes a sudden break away from the safety rope; Acquire the tension data of the protective rope, wherein the tension data of the protective rope is collected by a tension sensor with a sampling frequency of 50 Hz; Calculate the tension change rate and the position change rate according to the protection rope tension data and the worker position coordinates; evaluating a correlation between the rate of change of tension and the rate of change of position; If the tension change rate exceeds a preset threshold and the correlation is lower than a preset correlation coefficient threshold, it is determined that an abnormal situation exists.

9. The on-site monitoring method based on engineering informationization according to claim 1 is characterized in that: The S6 includes: Obtaining a real-time image from a high-definition surveillance camera, and preprocessing the real-time image to obtain a preprocessed image; Establishing a texture descriptor based on a gray-level co-occurrence matrix according to the preprocessed image, converting the preprocessed image from an RGB color space to an HSV color space to obtain an HSV image; Processing the HSV image, extracting the area that meets the characteristics of the protection rope, and obtaining a preliminary contour image of the protection rope; Processing the preliminary contour image, extracting the precise edge and endpoints of the protection rope, and obtaining an edge endpoint image; If there is a suspected detached protection rope connection point in the edge endpoint image, the edge endpoint image is processed to identify and locate the suspected detached protection rope connection point to obtain a marked monitoring picture.

10. The on-site monitoring method based on engineering informationization according to claim 1 is characterized in that: The S7 includes: Acquire acceleration data collected by the inertial measurement unit, perform spectrum analysis based on the acceleration data, and obtain the main frequency components of the worker's body swing; Acquire vibration data of a tension sensor on the protection rope, calculate the vibration frequency of the protection rope according to the vibration data, and obtain a ratio of the worker's body swing frequency to the protection rope vibration frequency; Fitting the worker's motion trajectory curve, calculating the maximum amplitude and average amplitude of the working action according to the motion trajectory curve, and judging whether the working action amplitude exceeds the preset protection range of the safety rope; The mass distribution and spatial position of each part of the worker's body are estimated based on the simplified human biomechanical model and inertial measurement unit data, and the center of gravity trajectory is smoothed to obtain the offset distance of the worker's center of gravity relative to the connection point of the protective rope; It is determined whether the frequency ratio, the operating motion amplitude and the center of gravity offset distance exceed a preset dynamic safety threshold. If they exceed the dynamic safety threshold, it is determined to be a separation state, and a warning message is sent to the smart device worn by the worker through the wireless communication module.

Images