Intelligent monitoring, prevention and control method and system for operation condition of escalator

Abstract

The invention discloses an intelligent monitoring, prevention and control method and system for the operation condition of an escalator, and the method comprises the steps: determining optimal installation parameters, installing a high-definition camera in an operation region of the escalator, obtaining an image of the operation condition of the escalator, screening scanning data and operation data through a hypersphere, and obtaining abnormal data; determining an early warning result according to abnormal data and a danger threshold value, correcting the operation condition image to obtain an operation condition correction image, inputting the operation condition correction image into a pre-trained escalator danger identification model to obtain an early warning result, and transmitting the early warning result, the abnormal data and the operation condition image to a control center for storage, and the control center executes a prevention and control strategy according to the early warning result. The method not only can improve the running safety and reliability of the escalator and effectively reduce the fault occurrence rate and the accident risk, but also has good interpretability, and can be directly applied to an intelligent monitoring, prevention and control system for the running condition of the escalator.

Description

Technical Field

[0001] The present invention relates to the technical field of escalator safety protection, and particularly to an intelligent monitoring and prevention and control method and system for the operating conditions of escalators. Background Technique

[0002] As an important part of the vertical transportation system, escalators are widely used in densely populated places such as subways, airports, and shopping malls. Their safe operation is directly related to the safety of public travel and the efficiency of urban transportation. However, traditional escalator safety protection technologies are difficult to meet the modern operation and maintenance requirements. Especially in scenarios of abnormal passenger behavior, equipment aging, and high-frequency use, escalators are more likely to cause safety accidents. Therefore, building an intelligent and full-cycle operation monitoring and prevention and control system has become an inevitable trend to improve the safety and reliability of escalators.

[0003] Current escalator monitoring technologies mainly rely on mechanical sensors and regular manual inspections, which have significant limitations: First, single-point monitoring cannot cover complex working conditions; second, offline analysis is difficult to capture instantaneous faults; finally, the early warning model lacks self-adaptability, resulting in insufficient accident prevention capabilities. In recent years, the breakthroughs in Internet of Things, edge computing, and machine learning technologies have provided a new direction for the industry. The present invention proposes an intelligent monitoring and prevention and control method and system for the operating conditions of escalators. By constructing a multi-dimensional sensor network and combining a fault prediction model based on deep learning, it realizes real-time perception of equipment status, intelligent identification of abnormal behaviors, and hierarchical early warning responses, effectively overcoming the deficiencies of traditional technologies, significantly improving the safety and reliability of escalator operation, effectively reducing the failure rate and accident risks, ensuring the personal safety of passengers, improving operation efficiency, and providing solid technical support for the normal operation of urban public transportation and commercial venues. Summary of the Invention

[0004] The object of the present invention is to provide an intelligent monitoring and prevention and control method and system for the operating conditions of escalators.

[0005] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0006] The present invention includes the following steps:

[0007] Obtain escalator scan data and operation data, determine the optimal installation parameters, and install high-definition cameras in the escalator operation area to obtain images of the escalator operation conditions through the cameras;

[0008] Use a hypersphere to screen the scan data and operation data to obtain abnormal data, and determine the early warning result according to the abnormal data and the danger threshold;

[0009] Perform image correction on the operating condition image to obtain an operating condition corrected image, and input the operating condition corrected image into a pre-trained escalator danger recognition model to obtain a warning result;

[0010] Transmit the warning result, the abnormal data, and the operating condition image to the control center for storage, and the control center executes a prevention and control strategy according to the warning result;

[0011] When no warning result is generated, the management personnel send a reset signal to resume the operation of the escalator.

[0012] Furthermore, the method for determining the optimal installation parameters includes:

[0013] Determine the installation optimization goal; the installation optimization goal is that the reciprocal of the monitoring overlapping area of all high-definition cameras is the smallest and the monitoring cost is the lowest; the constraint condition of the installation optimization goal is that the monitoring area of the high-definition cameras covers the entire escalator operation area;

[0014] Randomly determine the initial installation parameters, use the initial installation parameters as the initial population, and use the particle swarm optimization algorithm to optimize the initial installation parameters. Continuously iterate until the installation optimization goal is the smallest or reaches the maximum number of iterations, and obtain the optimal population to output the optimal installation parameters.

[0015] Furthermore, the method for determining the warning result according to the abnormal data and the danger threshold includes:

[0016] Calculate the importance of each feature based on the mutual information method and assign the weight w k Construct a weighted feature vector, dynamically adjust the radius R(t) of the hypersphere according to the spatio-temporal characteristics of the data distribution, and the updated expression of the hypersphere screening objective function is:

[0017]

[0018]

[0019]

[0020] where f aim is the hypersphere screening objective function, (t0, t1) is the sliding window update time period, C1 is the penalty coefficient, m is the number of samples, ζ i is the slack variable, A is the center of the sphere, R0 is the radius of the hypersphere before update, ρ is the attenuation coefficient, k is the density sensitive coefficient, d t is the passenger density at time t, d0 is the initial passenger density, is the weighted feature vector of sample i, a i,k is the value of feature k of sample i, and d is the dimension of the feature;

[0021] Screen the scanned data and the running data according to the objective function of hypersphere screening to obtain abnormal data;

[0022] Set the primary danger threshold and the secondary danger threshold. When the abnormal data of the scanned data is greater than the corresponding primary danger threshold, it is determined as a passenger abnormal behavior and a primary anomaly. When the abnormal data of the running data is greater than the corresponding primary danger threshold, it is determined as a machine running anomaly and a primary anomaly. When the abnormal data of the scanned data is greater than the corresponding secondary danger threshold, it is determined as a passenger dangerous behavior and a secondary risk. When the abnormal data of the running data is greater than the corresponding secondary danger threshold, it is determined as a machine running anomaly and a secondary risk.

[0023] Further, a method for obtaining a warning result by inputting the running condition correction image into a pre-trained escalator danger recognition model includes:

[0024] Perform color and ambient light correction on the running condition image according to the light environment of the escalator running area to obtain the first running condition image. Obtain the camera internal parameter matrix and the distortion coefficient, and perform image distortion correction on the first running condition image according to the inverse mapping algorithm to obtain the second running condition image. For multi-angle images of the same monitoring area, use the SIFT feature matching and RANSAC algorithms to calculate the homography matrix, and obtain the running condition correction image by weighted average fusion processing of the second running condition image;

[0025] The objective function of the color and ambient light correction is:

[0026]

[0027] where I(x,y) is the pixel value of the original image at the coordinate (x,y), L(x,y) is the illumination component at the coordinate (x,y), r(x,y) is the reflection component at the coordinate (x,y), λ is the smoothing coefficient, is the pixel value gradient;

[0028] The expression of the weighted average fusion processing is:

[0029] w i =exp{-.‖H i ·x - x gt ‖ 2 +‖H i ·y - y gt ‖ 2 / / σ 2}

[0030] where w i is the weighted average fusion weight of the viewing angle i, H i is the homography matrix of the viewing angle i, x, y are the point coordinates of the viewing angle i, x gt 、y gtis the reference perspective coordinate, and σ is the Gaussian kernel width;

[0031] Input the operating condition correction image into the pre-trained escalator danger recognition model to obtain the warning result;

[0032] The pre-trained escalator danger recognition model includes an input layer, a first dangerous behavior recognition unit, a second dangerous behavior recognition unit, and an output layer;

[0033] The first dangerous behavior recognition unit constructs a visual model using the YOLOv11 algorithm. The backbone network is CSPDarknet-Enhanced, which extracts multi-scale features through a cross-stage local attention module and outputs the target bounding box and the first warning result;

[0034] The second dangerous behavior recognition unit includes a feature extraction module, a bag-of-words association module, a danger level prediction module, and a first fully connected layer; The feature extraction module uses a VGG-16 convolutional network to extract image features and generates a fixed-dimensional feature vector through global average pooling; The bag-of-words association module performs K-means clustering on the feature vector, constructs a dictionary of visual words, and generates a bag-of-words feature vector weighted by term frequency-inverse document frequency to predict the dangerous situation with the preset dangerous situation; The danger level prediction module inputs the bag-of-words feature vector into a support vector regression model and outputs the predicted danger level by mapping the feature space through a radial basis kernel function; The first fully connected layer connects the bag-of-words association module and the danger level prediction module to output the second warning result;

[0035] The optimization objective and constraint conditions of the danger level prediction module are:

[0036]

[0037] where w j is the weight vector of the model, which is used to define the hyperplane to affect the decision function of the model. C2 is the penalty coefficient, and ξ j , are slack variables, y j is the actual value, is the predicted value, is the representation of the feature vector B j in the high-dimensional space, b is the bias term, and ε is the tolerance error;

[0038] The output layer connects the first dangerous behavior recognition unit and the second dangerous behavior recognition unit through a second fully connected layer, and performs weighted fusion on the first warning result and the second warning result to output the warning result;

[0039] The escalator danger recognition model uses Softmax cross-entropy loss to predict dangerous situations, SmoothL1 loss to predict danger levels, and the AdamW optimizer and cosine annealing strategy to adjust the learning rate.

[0040] Furthermore, the method for transmitting the warning result, the abnormal data, and the operation status image to the control center and storing them includes:

[0041] Transmit the warning result, abnormal data, and operation status image to the control center by combining wired communication and wireless communication for backup and store them in partitions according to data categories; the wireless communication includes adaptive dual-channel scheduling and quantum encryption enhancement; the adaptive dual-channel scheduling is based on the TDMA-LoRaWAN hybrid protocol and dynamically allocates channel resources through the Q-learning algorithm; the quantum encryption enhancement integrates the NTRU post-quantum encryption algorithm in the LoRa physical layer to construct a key matrix K, generates public and private keys according to the key matrix K, and encrypts and decrypts the transmitted plaintext data;

[0042] The expression of the key matrix K is:

[0043]

[0044] where K is the key matrix, α and β are exponential parameters, {p i , q i} is a set of prime numbers, I is the number of set elements, the set of prime numbers is updated every 10 minutes through a hash chain, {p i ′ , q i ′} = SHA3-256({p i , q i} ∥ Timestamp), SHA3-256 is the SHA3-256 hash function, and Timestamp is the corresponding timestamp.

[0045] Furthermore, the method for the control center to execute prevention and control strategies according to the warning result includes:

[0046] Classify the received dangerous situations into passenger-related dangerous situations and equipment-related dangerous situations;

[0047] The control center triggers prevention and control strategies according to the received warning result. The prevention and control strategies specifically include:

[0048] When receiving any dangerous situation signal, trigger the corresponding category voice prompt and transmit the corresponding category voice prompt signal to the voice broadcaster for voice broadcast;

[0049] When receiving passenger - type dangerous situations and level - 1 abnormal signals, only voice announcements are made;

[0050] When receiving passenger - type dangerous situations and level - 2 risk signals, voice announcements, personnel management, and type - two braking are carried out;

[0051] When receiving equipment - type dangerous situations and level - 1 abnormal signals, voice announcements, personnel management, and type - one braking are carried out; the specific operation of the type - one braking is to reduce the elevator running speed by 50%;

[0052] When receiving equipment - type dangerous situations and level - 2 risk signals, voice announcements, personnel management, and type - two braking are carried out; the specific operation of the type - two braking is to perform progressive braking until the elevator stops, and the corresponding cubic spline interpolation speed curve is:

[0053] v(t) = b1(t - t0) 3 +b2(t - t0) 2 +b0

[0054] where v(t) is the elevator running speed, b0, b1, b2 are undetermined coefficients, t0 is the starting time of braking, and t is the target braking time;

[0055] The dangerous situations include abnormal machine operation, machine structure defects, abnormal passenger behavior, and dangerous passenger behavior; the risk levels include level - 1 abnormality and level - 2 risk.

[0056] In a second aspect, an intelligent monitoring and prevention - control system for the operating condition of an escalator includes:

[0057] A camera module: used to determine the optimal installation parameters and install a high - definition camera in the escalator operation area, and obtain the escalator operation condition image through the camera;

[0058] A data monitoring module: used to obtain the escalator scan data and operation data, screen the scan data and operation data using a hypersphere to obtain abnormal data, and determine the warning result according to the abnormal data and the danger threshold;

[0059] An image monitoring module: used to perform image correction on the operation condition image to obtain a corrected operation condition image, and input the corrected operation condition image into a pre - trained escalator danger recognition model to obtain a warning result;

[0060] A protection and control module: used to transmit the warning result, the abnormal data, and the operation condition image to the control center and store them, and the control center executes the prevention - control strategy according to the warning result;

[0061] Management module: used to store, manage, and view the warning results, the abnormal data, and the operation status images, analyze and statistically process the stored data, classify and manage the data to ensure data integrity and security, and perform data backup and recovery operations.

[0062] The beneficial effects of the present invention are as follows:

[0063] The present invention is an intelligent monitoring and prevention and control method and system for the operation status of escalators. Compared with the prior art, the present invention has the following technical effects:

[0064] Through steps such as determining optimal installation parameters, screening abnormal data, image correction, data encryption, and building a model, the present invention can improve the data preprocessing ability and enhance the model adaptability in the intelligent monitoring and prevention and control of escalator operation status, improve the safety and reliability of the intelligent monitoring and prevention and control of escalator operation status, thereby improving the efficiency and accuracy of the intelligent monitoring and prevention and control of escalator operation status, optimize the intelligent monitoring and prevention and control technology of escalator operation status, greatly save resources, improve work efficiency, effectively reduce the failure rate and accident risk, thus ensuring the personal safety of passengers, optimize the maintenance strategy of escalators, reduce maintenance costs to improve operation efficiency, provide a solid technical support for the normal operation of urban public transportation and commercial places, and can adapt to different intelligent monitoring and prevention and control systems for escalator operation status and the intelligent monitoring and prevention and control needs of different users, having a certain universality. Description of the Drawings

[0065] Figure 1 It is a step flowchart of an intelligent monitoring and prevention and control method for the operation status of an escalator according to the present invention. Detailed Embodiments

[0066] The present invention will be further described below through specific embodiments. The illustrative embodiments and explanations of the present invention are used to explain the present invention, but do not limit the present invention.

[0067] An intelligent monitoring and prevention and control method and system for the operation status of an escalator according to the present invention include the following steps:

[0068] As Figure 1 shown, in this embodiment, it includes the following steps:

[0069] Obtain escalator scan data and operation data, determine optimal installation parameters, install a high-definition camera in the escalator operation area, and obtain the escalator operation status image through the camera;

[0070] Use a hypersphere to screen the scan data and operation data to obtain abnormal data, and determine the warning result according to the abnormal data and the danger threshold;

[0071] Perform image correction on the operating condition image to obtain an operating condition corrected image, and input the operating condition corrected image into a pre-trained escalator hazard identification model to obtain a warning result;

[0072] Transmit the warning result, the abnormal data, and the operating condition image to the control center for storage, and the control center executes a prevention and control strategy according to the warning result;

[0073] When no warning result is generated, the management personnel send a reset signal to resume the operation of the escalator.

[0074] In this embodiment, the method for determining the optimal installation parameters includes:

[0075] Determine the installation optimization objective; the installation optimization objective is that the reciprocal of the monitoring overlapping area of all high-definition cameras is the smallest and the monitoring cost is the lowest; the constraint condition of the installation optimization objective is that the monitoring area of the high-definition cameras covers the entire escalator operation area;

[0076] Randomly determine the initial installation parameters, use the initial installation parameters as the initial population, and use the particle swarm optimization algorithm to optimize the initial installation parameters, and continuously iterate until the installation optimization objective is the smallest or reaches the maximum number of iterations, and obtain the optimal population to output the optimal installation parameters;

[0077] In the actual evaluation, the installation parameters include the installation position, installation angle, and installation height;

[0078] Taking the intelligent monitoring and prevention and control of the operating conditions of an escalator in a certain shopping mall as an example, the escalator length is 20m, the height difference is 6m, the required coverage area is the pedal area length * width / 20m * 1.2m, and each side of the handrail belt area is 0.5m. The optimization objective is inversely proportional to the overlapping area of the cameras and directly proportional to the number of cameras;

[0079] Use the particle swarm optimization algorithm to optimize the initial installation parameters. When the 10th iteration is performed, the optimal solution is to install 3 cameras, the positions are (6m, 10m, 16m), the angle of pitch is 15°, the height is 2.8m, the corresponding reciprocal of the overlapping area = 0.12, and the total cost = 3.3; when the 50th iteration is performed, the optimal solution is to install 2 cameras, the positions are (8m, 14m), the angle of pitch is 18°, the height is 3.0m, the corresponding reciprocal of the overlapping area = 0.08, and the total cost = 2.1. At this time, the installation optimization objective is the smallest, and the corresponding optimal installation parameters are output to install cameras to obtain the operating condition image of the escalator.

[0080] In this embodiment, the method for determining the warning result according to the abnormal data and the danger threshold includes:

[0081] Calculate the importance of each feature based on the mutual information method and assign the weight w kConstruct a weighted feature vector, dynamically adjust the radius R(t) of the hypersphere according to the spatio-temporal characteristics of the data distribution, and the updated expression of the hypersphere screening objective function is:

[0082]

[0083] where f aim is the hypersphere screening objective function, (t0, t1) is the sliding window update time period, C1 is the penalty coefficient, m is the number of samples, ζ i is the slack variable, A is the center of the sphere, R0 is the radius of the hypersphere before update, ρ is the decay coefficient, k is the density sensitivity coefficient, d t is the passenger density at time t, d0 is the initial passenger density, is the weighted feature vector of sample i, a i,k is the value of feature k of sample i, and d is the dimension of the feature;

[0084] Screen the scan data and operation data according to the hypersphere screening objective function to obtain abnormal data;

[0085] Set the first-level danger threshold and the second-level danger threshold. When the abnormal data of the scan data is greater than the corresponding first-level danger threshold, it is determined as passenger abnormal behavior and first-level abnormality. When the abnormal data of the operation data is greater than the corresponding first-level danger threshold, it is determined as machine operation abnormality and first-level abnormality. When the abnormal data of the scan data is greater than the corresponding second-level danger threshold, it is determined as passenger dangerous behavior and second-level risk. When the abnormal data of the operation data is greater than the corresponding second-level danger threshold, it is determined as machine operation abnormality and second-level risk;

[0086] In actual evaluation, taking the intelligent monitoring and prevention and control of the operation status of an escalator in a shopping mall as an example, the scan data is obtained: the distance (m) that passengers stretch out of the guardrail [0.2, 0.39, 0.55, 0.51], and the operation data is obtained: {vibration intensity (m / s 2 ) / noise speed (dB) / passenger load (kg / m 2 ) / data stability (variance)} is [3.1, 4.8, 4.5, 6.2] / [75, 82, 80, 95] / [200, 250, 220, 380] / [0.5, 2.8, 1.2, 4.5]. Based on the mutual information method, the weight distribution w k [0.3, 0.25, 0.15, 0.2, 0.1] is obtained. Taking the third group of data as an example, the corresponding weighted feature vector is 72.3;

[0087] Taking the third set of data as an example, with the initial radius R0 being 50, the time decay coefficient ρ being 0.01, the density sensitivity coefficient k being 0.1, and the initial passenger density d0 being 5 passengers per minute, the dynamic radius is calculated to be 60.4 when the current time t is 5 minutes and the passenger density is 10 passengers per minute. At this time, the third set of data Then this set of data is normal, and the second and fourth sets of data are screened out as abnormal data in the same way;

[0088] Taking the second set of data as an example, for the scanned data - the distance that the passenger extends out of the guardrail is 0.39 m, which is greater than the first-level danger threshold of 0.25 m, it is determined as an abnormal passenger behavior and a first-level anomaly. For the operating data - the vibration intensity is 4.8 m / s 2 which is greater than the first-level danger threshold of 4.5 m / s 2 and for the operating data - the variance is 2.8, which is greater than the first-level danger threshold of 2.5, it is determined as an abnormal machine operation and a first-level anomaly;

[0089] Taking the fourth set of data as an example, for the scanned data - the distance that the passenger extends out of the guardrail is 0.51 m, which is greater than the second-level danger threshold of 04 m, it is determined as a dangerous passenger behavior and a second-level risk. For the operating data - the vibration intensity is 6.2 m / s 2 which is greater than the second-level danger threshold of 6.0 m / s 2 and for the operating data - the passenger load is 380 kg / m 2 which is greater than the second-level danger threshold of 380 kg / m 2 and it is determined as an abnormal machine operation and a second-level risk.

[0090] In this embodiment, the method of inputting the operating condition correction image into a pre-trained escalator danger recognition model to obtain a warning result includes:

[0091] Performing color and ambient light correction on the operating condition image according to the light environment of the escalator operating area to obtain a first operating condition image, obtaining the internal camera matrix and distortion coefficient, performing image distortion correction on the first operating condition image according to the inverse mapping algorithm to obtain a second operating condition image, for multi-angle images of the same monitoring area, calculating the homography matrix using the SIFT feature matching and RANSAC algorithms, and obtaining the operating condition correction image through weighted average fusion processing of the second operating condition image;

[0092] The objective function of the color and ambient light correction is:

[0093]

[0094] where I(x,y) is the pixel value of the original image at the coordinate (x,y), L(x,y) is the illumination component at the coordinate (x,y), r(x,y) is the reflection component at the coordinate (x,y), and λ is the smoothing coefficient, is the pixel value gradient;

[0095] The weighted average fusion processing expression is:

[0096] w i = exp{-.‖H i ·x - x gt ‖ 2 +‖H i ·y - y gt ‖ 2 / / σ 2}

[0097] where w i is the weighted average fusion weight of view i, H i is the homography matrix of view i, x and y are the point coordinates of view i, x gt and y gt are the reference view coordinates, and σ is the Gaussian kernel width;

[0098] Input the operating condition corrected image into the pre-trained escalator danger recognition model to obtain a warning result;

[0099] The pre-trained escalator danger recognition model includes an input layer, a first dangerous behavior recognition unit, a second dangerous behavior recognition unit, and an output layer;

[0100] The first dangerous behavior recognition unit constructs a visual model using the YOLOv11 algorithm, with the backbone network being CSPDarknet-Enhanced, extracts multi-scale features through a cross-stage local attention module, and outputs the target bounding box and the first warning result;

[0101] The second dangerous behavior recognition unit includes a feature extraction module, a bag-of-words association module, a danger level prediction module, and a first fully connected layer; the feature extraction module extracts image features using the VGG-16 convolutional network and generates a fixed-dimensional feature vector through global average pooling; the bag-of-words association module performs K-means clustering on the feature vector, constructs a dictionary of visual words, generates a bag-of-words feature vector weighted by term frequency-inverse document frequency, and predicts the dangerous situation; the danger level prediction module inputs the bag-of-words feature vector into a support vector regression model and outputs the predicted danger level by mapping the feature space through a radial basis kernel function; the first fully connected layer connects the bag-of-words association module and the danger level prediction module to output the second warning result;

[0102] The optimization objective and constraint conditions of the danger level prediction module are:

[0103]

[0104] where w jis the weight vector of the model, used to define the hyperplane to affect the decision function of the model, C2 is the penalty coefficient, and ξ j 、 is the slack variable, and y j is the actual value, is the predicted value, is the feature vector B after kernel function mapping j representation in the high-dimensional space, b is the bias term, and ε is the tolerance error;

[0105] The output layer is connected to the first dangerous behavior recognition unit and the second dangerous behavior recognition unit through the second fully connected layer, and the first warning result and the second warning result are weighted and fused to output the warning result;

[0106] The escalator danger recognition model uses Softmax cross-entropy loss to predict the danger situation, uses SmoothL1 loss to predict the danger level, and uses the AdamW optimizer and cosine annealing strategy to adjust the learning rate;

[0107] In the actual evaluation, taking the intelligent monitoring and prevention and control of the operating conditions of an escalator in a shopping mall as an example, the smoothing coefficient λ of the color and ambient light correction objective function is taken as 0.1. After correction, the brightness variance is reduced by 25%. Using the inverse mapping algorithm, the edge error after distortion correction is <0.5px, and the viewing angle weights w i of the two cameras are 0.45 and 0.55 respectively, and the corrected image of the operating condition is obtained through processing,

[0108] Taking the second group of data as an example, inputting the corresponding corrected image of the operating condition into the pre-trained escalator danger recognition model respectively obtains the first warning result and the second warning result: Machine structure defect - foreign object accumulation at the deformed comb plate - first-level anomaly, Machine structure defect - foreign object stuck at the deformed comb plate - first-level anomaly, and the output warning result is: Machine structure defect - anomaly at the deformed comb plate - first-level anomaly. Similarly, inputting the corrected image of the operating condition corresponding to the fourth group of data into the pre-trained escalator danger recognition model outputs the warning result: Passenger dangerous behavior - passenger going against the flow - second-level danger.

[0109] In this embodiment, the method for transmitting the warning result, the abnormal data, and the operating condition image to the control center and storing them includes:

[0110] The early warning results, abnormal data, and operation status images are transmitted to the control center by a combined backup method of wired communication and wireless communication and stored in partitions according to data categories; the wireless communication includes adaptive dual-channel scheduling and quantum encryption enhancement; the adaptive dual-channel scheduling is based on the TDMA-LoRaWAN hybrid protocol, and the channel resources are dynamically allocated through the Q-learning algorithm; the quantum encryption enhancement integrates the NTRU post-quantum encryption algorithm in the LoRa physical layer to construct the key matrix K, generates the public key and private key according to the key matrix K, and encrypts and decrypts the transmitted plaintext data;

[0111] The expression of the key matrix K is:

[0112]

[0113] where K is the key matrix, α and β are exponential parameters, {p i , q i} is a set of prime numbers, I is the number of set elements, the set of prime numbers is updated every 10 minutes through the hash chain, {p i ′ , q i ′} = SHA3-256({p i , q i} ∥ Timestamp), SHA3-256 is the SHA3-256 hash function, and Timestamp is the corresponding timestamp;

[0114] In the actual evaluation, taking the intelligent monitoring and prevention and control of the operation status of an escalator in a shopping mall as an example, and taking the transmission process of the second group of data as an example, the channel parameters of the dual channels are Channel 1 (433 MHz, bandwidth 125 kHz) and Channel 2 (433 MHz, bandwidth 250 kHz) respectively. The learning rate of the Q-learning algorithm is 0.1, the discount factor is 0.9, and the initial Q table is 0. It is calculated that the load of Channel 1 is 70% and the load of Channel 2 is 30%. Therefore, Channel 1 is preferentially selected in the next time period;

[0115] The set of prime numbers {p i , q i} = {3, 5, 7}, the value of the key matrix K is 3688, and the set of prime numbers {p i ′ , q i ′} = SHA3-256(3688 ∥ Timestamp = 1630000000) is updated every 10 minutes, and the new set of prime numbers {11, 13, 17} is output. Taking the transmission of the vibration intensity of 4.8 m / s of the second group of data 2 as an example, the vibration intensity is binary encoded as For the plaintext data Perform NTRU encryption, decrypt using the private key, and restore the inscription to obtain the second set of data;

[0116] Verify the hash value of the second set of data, store the warning results, abnormal data, and operating status images of the second set of data in partitions, store the warning results in the Redis real-time database, store the abnormal data in the InfluxDB time-series database, and store the operating status images in object storage.

[0117] In this embodiment, the method for the control center to execute the prevention and control strategy according to the warning result includes:

[0118] Divide the received dangerous situations into passenger-type dangerous situations and equipment-type dangerous situations;

[0119] The control center triggers the prevention and control strategy according to the received warning result. The prevention and control strategy specifically includes:

[0120] When receiving any dangerous situation signal, trigger the corresponding category voice prompt and transmit the corresponding category voice prompt signal to the voice broadcaster for voice broadcast;

[0121] When receiving passenger-type dangerous situations and level-1 abnormal signals, only perform voice broadcast;

[0122] When receiving passenger-type dangerous situations and level-2 risk signals, perform voice broadcast, personnel management, and type-II braking;

[0123] When receiving equipment-type dangerous situations and level-1 abnormal signals, perform voice broadcast, personnel management, and type-I braking; The specific operation of the type-I braking is to reduce the elevator running speed by 50%;

[0124] When receiving equipment-type dangerous situations and level-2 risk signals, perform voice broadcast, personnel management, and type-II braking; The specific operation of the type-II braking is to perform progressive braking until the elevator stops. The corresponding cubic spline interpolation speed curve is:

[0125] v(t) = b1(t - t0) 3 + b2(t - t0) 2 + b0

[0126] where v(t) is the elevator running speed, b0, b1, and b2 are undetermined coefficients, t0 is the starting time of braking, and t is the target braking time;

[0127] The dangerous situations include abnormal machine operation, machine structure defects, abnormal passenger behavior, and dangerous passenger behavior; The danger levels include level-1 abnormal and level-2 risk;

[0128] In the actual evaluation, taking the intelligent monitoring and prevention and control of the operation status of an escalator in a certain shopping mall as an example, the warning results corresponding to the second group of data are as follows: abnormal passenger behavior (body extending out of the guardrail), abnormal machine operation (excessive vibration intensity and noise), and machine structure defects (abnormal at the deformed comb plate), all of which are first-level abnormalities. The control center triggers the prevention and control strategy according to the received warning results: first, a voice broadcast is made to prompt "Please do not rely on the handrail or extend your body out of the handrail, pay attention to safety" and "The machine is operating abnormally, do not continue to take the elevator". At the same time, a type of braking is carried out to reduce the elevator operating speed by 50%, and the management staff is notified to arrive at the scene to remind passengers and check for machine abnormalities. When no warning results are generated, the management staff sends a reset signal to resume the operation of the escalator;

[0129] The warning results corresponding to the fourth group of data are as follows: dangerous passenger behavior (body extending out of the guardrail, passengers going against the flow), abnormal machine operation (passenger load), both of which are second-level abnormalities. The control center triggers the prevention and control strategy according to the received warning results: first, a voice broadcast is made to prompt "It is strictly prohibited to climb over the handrail, it is very dangerous", "Do not go against the flow when taking the elevator, please turn around immediately", and "You are seriously overweight, do not crowd". At the same time, a type of braking is carried out to perform a progressive braking until the elevator stops (the elevator operating speed v(t)=-0.05t 3 +0.1t 2 +0.6, and the elevator stops after 3 seconds), and the management staff is notified to arrive at the scene to stop the dangerous behavior of passengers. When no warning results are generated, the management staff sends a reset signal to resume the operation of the escalator.

[0130] Secondly, an intelligent monitoring and prevention and control system for the operation status of an escalator includes:

[0131] Camera module: used to determine the optimal installation parameters and install high-definition cameras in the escalator operation area, and obtain the images of the escalator operation status through the cameras;

[0132] Data monitoring module: used to obtain the escalator scan data and operation data, screen the scan data and operation data using a hypersphere to obtain abnormal data, and determine the warning results according to the abnormal data and the danger threshold;

[0133] Image monitoring module: used to perform image correction on the operation status images to obtain corrected operation status images, and input the corrected operation status images into a pre-trained escalator danger recognition model to obtain warning results;

[0134] Protection control module: used to transmit the warning results, the abnormal data, and the operation status images to the control center and store them, and the control center executes the prevention and control strategy according to the warning results;

[0135] Management module: used to store, manage, and view the warning results, the abnormal data, and the operation status images, analyze and statistically process the stored data, classify and manage the data to ensure data integrity and security, and at the same time be able to perform data backup and recovery operations.

[0136] The present invention can automatically stop the escalator or reduce the running speed when solving passenger abnormal behaviors and serious equipment defects. For example, when it is found that an old person has fallen or a child's hands or feet are caught, the escalator can be automatically stopped in time to prevent the expansion of injuries. Due to different control methods, some escalators can reduce the speed until they stop, while some cannot reduce the speed and can only stop. For escalators with variable frequency control, the running speed can be slowly reduced until they stop. For escalators without variable frequency control, voice prompts can be used and they can be directly stopped after a delay (such as 3 seconds). After the escalator fails and stops, only manual on-site confirmation of safety work can be carried out, and it can be manually started.

[0137] The above are only the preferred embodiments of the present invention, and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. An intelligent monitoring and prevention and control method for the operating status of an escalator, characterized in that, It includes the following steps: S1. Obtain the escalator scan data and operation data, determine the optimal installation parameters, install high-definition cameras in the escalator operation area, and obtain the escalator operation status images through the cameras; S2. Use a hypersphere to screen the scan data and the operation data to obtain abnormal data, and determine the warning result according to the abnormal data and the danger threshold; S3. Perform image correction on the operation status images to obtain corrected operation status images, and input the corrected operation status images into a pre-trained escalator danger recognition model to obtain a warning result; S4. Transmit the warning result, the abnormal data, and the operation status images to the control center for storage, and the control center executes the prevention and control strategy according to the warning result; S5. When no warning result is generated, the management personnel send a reset signal to resume the operation of the escalator; The warning result includes a danger status and a danger level; the danger status includes abnormal machine operation, machine structure defect, abnormal passenger behavior, and dangerous passenger behavior; the danger level includes a first-level abnormality and a second-level risk; The prevention and control strategy includes voice broadcast, personnel management, and elevator braking; the elevator braking includes a first type of braking and a second type of braking.

2. The intelligent monitoring and prevention and control method for the operating condition of an escalator according to claim 1, wherein The method for determining the optimal installation parameters includes: Determine the installation optimization goal; the installation optimization goal is that the reciprocal of the monitoring overlapping area of all high-definition cameras is the smallest and the monitoring cost is the lowest; the constraint condition of the installation optimization goal is that the monitoring area of the high-definition cameras covers the entire escalator operation area; Randomly determine the initial installation parameters, use the initial installation parameters as the initial population, and use the particle swarm optimization algorithm to optimize the initial installation parameters, and continuously iterate until the installation optimization goal is the smallest or reaches the maximum iteration number, and obtain the optimal population to output the optimal installation parameters.

3. The intelligent monitoring and prevention and control method for the operating condition of an escalator according to claim 1, characterized in that, The method for determining the warning result according to the abnormal data and the danger threshold includes: Calculate the importance of each feature based on the mutual information method and assign the weight w k Construct a weighted feature vector, dynamically adjust the radius R(t) of the hypersphere according to the spatio-temporal characteristics of the data distribution, and the updated expression of the hypersphere screening objective function is as follows: where f aim is the objective function for hypersphere screening, (t0, t1) is the update time period of the sliding window, C1 is the penalty coefficient, m is the number of samples, ζ i is the slack variable, A is the center of the sphere, R0 is the radius of the hypersphere before update, ρ is the decay coefficient, k is the density sensitivity coefficient, d t is the passenger density at time t, d0 is the initial passenger density, is the weighted feature vector of sample i, a i,k is the value of feature k of sample i, and d is the dimension of the feature; Screen the scan data and the operation data according to the hypersphere screening objective function to obtain abnormal data; Set a first-level danger threshold and a second-level danger threshold. When the abnormal data of the scan data is greater than the corresponding first-level danger threshold, it is determined as abnormal passenger behavior and a first-level abnormality. When the abnormal data of the operation data is greater than the corresponding first-level danger threshold, it is determined as abnormal machine operation and a first-level abnormality. When the abnormal data of the scan data is greater than the corresponding second-level danger threshold, it is determined as dangerous passenger behavior and a second-level risk. When the abnormal data of the operation data is greater than the corresponding second-level danger threshold, it is determined as abnormal machine operation and a second-level risk.

4. The intelligent monitoring and prevention and control method for the operating status of an escalator according to claim 1, characterized in that, The method for inputting the corrected operation status images into a pre-trained escalator danger recognition model to obtain a warning result includes: Perform color and ambient light correction on the operation status images according to the light environment of the escalator operation area to obtain the first operation status images, obtain the internal parameter matrix and distortion coefficient of the camera, perform image distortion correction on the first operation status images according to the inverse mapping algorithm to obtain the second operation status images, for multi-angle images of the same monitoring area, use the SIFT feature matching and RANSAC algorithm to calculate the homography matrix, and obtain the corrected operation status images through weighted average fusion processing of the second operation status images; The objective function of the color and ambient light correction is: where \(I(x,y)\) is the pixel value of the original image at coordinates \((x,y)\), \(L(x,y)\) is the illumination component at coordinates \((x,y)\), \(r(x,y)\) is the reflection component at coordinates \((x,y)\), and \(\lambda\) is the smoothing coefficient. is the pixel value gradient; The weighted average fusion processing expression is as follows: w i = exp{-.‖H i ·x - x gt ‖ 2 +‖H i ·y - y gt ‖ 2 / / σ 2} where w i is the weighted average fusion weight of view i, H i is the homography matrix of view i, x and y are the point coordinates of view i, x gt , y gt are the reference view coordinates, and σ is the Gaussian kernel width; Input the operation condition corrected image into the pre-trained escalator danger recognition model to obtain a warning result; The pre-trained escalator danger recognition model includes an input layer, a first dangerous behavior recognition unit, a second dangerous behavior recognition unit, and an output layer; The first dangerous behavior recognition unit constructs a vision model using the YOLOv11 algorithm. The backbone network is CSPDarknet-Enhanced. It extracts multi-scale features through a cross-stage local attention module and outputs a target bounding box and a first warning result; The second dangerous behavior recognition unit includes a feature extraction module, a bag-of-words association module, a danger level prediction module, and a first fully connected layer; The feature extraction module uses a VGG-16 convolutional network to extract image features and generates a fixed-dimension feature vector through global average pooling; The bag-of-words association module performs K-means clustering on the feature vector, constructs a dictionary of visual words, generates a bag-of-words feature vector weighted by term frequency-inverse document frequency, and predicts the dangerous situation with a preset dangerous situation; The danger level prediction module inputs the bag-of-words feature vector into a support vector regression model and outputs the predicted danger level by mapping the feature space through a radial basis kernel function; The first fully connected layer connects the bag-of-words association module and the danger level prediction module to output a second warning result; The optimization objective and constraint conditions of the danger level prediction module are: where w j is the weight vector of the model, used to define the hyperplane to affect the decision function of the model, C2 is the penalty coefficient, and ξ j , is the slack variable, y j is the actual value, is the predicted value, is the representation of the feature vector B after kernel function mapping j in the high-dimensional space, b is the bias term, and ε is the tolerance error; The output layer is connected to the first dangerous behavior recognition unit and the second dangerous behavior recognition unit through a second fully connected layer, and weights the first warning result and the second warning result to fuse and output a warning result; The escalator danger recognition model uses Softmax cross-entropy loss to predict the dangerous situation, uses Smooth L1 loss to predict the danger level, and uses the AdamW optimizer and the cosine annealing strategy to adjust the learning rate.

5. The intelligent monitoring and prevention and control method for the operating condition of an escalator according to claim 1, characterized in that, The method for transmitting the warning result, the abnormal data, and the operation condition image to the control center and storing them includes: Transmit the warning result, abnormal data, and operation condition image to the control center in a combined backup manner of wired communication and wireless communication and store them in partitions according to data categories; The wireless communication includes adaptive dual-channel scheduling and quantum encryption enhancement; The adaptive dual-channel scheduling is based on the TDMA-LoRaWAN hybrid protocol and dynamically allocates channel resources through the Q-learning algorithm; The quantum encryption enhancement integrates the NTRU post-quantum encryption algorithm in the LoRa physical layer to construct a key matrix K, generates a public key and a private key according to the key matrix K, and encrypts and decrypts the transmitted plaintext data; The expression of the key matrix K is: where K is the key matrix, α and β are exponential parameters, {p i , q i} is the set of prime numbers, I is the number of set elements, the set of prime numbers is updated every 10 minutes through a hash chain, {p i ′ , q i ′} = SHA3-256({p i , q i} ∥ Timestamp), SHA3-256 is the SHA3-256 hash function, and Timestamp is the corresponding timestamp.

6. The intelligent monitoring and prevention and control method for the operating condition of an escalator according to claim 1, characterized in that, The method for the control center to execute a prevention and control strategy according to the warning result includes: Classify the received dangerous situations into passenger-type dangerous situations and equipment-type dangerous situations; The control center triggers a prevention and control strategy according to the received warning result. The prevention and control strategy specifically includes: When receiving any dangerous situation signal, trigger a corresponding category voice prompt and transmit a corresponding category voice prompt signal to the voice broadcaster for voice broadcast; When receiving passenger - type dangerous situations and level - one abnormal signals, only voice announcements are made; When receiving passenger - type dangerous situations and level - two risk signals, voice announcements, personnel management, and type - two braking are carried out; When receiving equipment - type dangerous situations and level - one abnormal signals, voice announcements, personnel management, and type - one braking are carried out; The specific operation of the type - one braking is to reduce the elevator running speed by 50%; When receiving equipment - type dangerous situations and level - two risk signals, voice announcements, personnel management, and type - two braking are carried out; The specific operation of the type - two braking is to perform progressive braking until the elevator stops, and the corresponding cubic spline interpolation speed curve is: v(t) = b1(t - t0) 3 + b2(t - t0) 2 + b0 where \(v(t)\) is the elevator running speed, \(b_0\), \(b_1\), \(b_2\) are undetermined coefficients, \(t_0\) is the starting time of braking, and \(t\) is the target braking time; The dangerous situations include abnormal machine operation, machine structure defects, abnormal passenger behavior, and dangerous passenger behavior; The risk levels include level - one abnormality and level - two risk.

7. An intelligent monitoring and prevention and control system for the operating condition of an escalator, which is used to execute the method described in any one of claims 1-6, characterized in that, It includes: Camera module: Used to determine the optimal installation parameters and install high - definition cameras in the escalator operation area, and obtain images of the escalator operation status through the cameras; Data monitoring module: Used to obtain escalator scan data and operation data, screen the scan data and operation data using a hypersphere to obtain abnormal data, and determine the warning result based on the abnormal data and the danger threshold; Image monitoring module: Used to perform image correction on the operation status image to obtain a corrected operation status image, and input the corrected operation status image into a pre - trained escalator danger recognition model to obtain a warning result; Protection control module: Used to transmit the warning result, the abnormal data, and the operation status image to the control center and store them. The control center executes prevention and control strategies based on the warning result; Management module: Used to store, manage, and view the warning result, the abnormal data, and the operation status image, analyze and statistically process the stored data, classify and manage the data to ensure data integrity and security, and at the same time be able to perform data backup and recovery operations.

Images