Construction elevator safety management and control method and device based on multi-element perception

Abstract

The embodiment of the invention provides a construction elevator safety management and control method and device based on multi-element perception, and the method and device achieve the precise verification of the identity of a person and the state of protection equipment through the multi-mode identity recognition of binocular living body detection and voiceprint collection in combination with a deep neural network. An equipment state evaluation model based on multi-sensor fusion is innovatively designed, and cooperative monitoring of key parameters such as load, position and speed is achieved. The system adopts an intelligent control strategy to carry out operation parameter adaptive adjustment, establishes a real-time fault protection mechanism based on threshold judgment, and realizes remote supervision through the Internet of Things technology. According to the method, the limitation of a traditional elevator management mode is broken through, and an intelligent overall solution is provided for building construction safety management.

Description

Technical Field

[0001] This application relates to the field of data processing, and particularly to a safety control method and device for construction hoists based on multi-sensor perception. Background Art

[0002] Traditional safety control methods for construction hoists mainly rely on simple mechanical safety devices and basic manual supervision, making it difficult to meet the intelligent safety management requirements in modern construction. Existing technologies have obvious deficiencies in personnel identity verification and safety protection detection, and lack the ability to monitor the qualifications of construction personnel and the wearing status of protective equipment in real time.

[0003] At the same time, existing systems also have limitations in equipment status monitoring and fault warning. Traditional methods often use a single sensor for simple parameter monitoring, failing to achieve collaborative analysis of multi-dimensional data and lacking a comprehensive evaluation mechanism for equipment operation status. The system lags behind in fault identification and safety warning, making it difficult to detect faults in a timely manner and respond quickly.

[0004] In addition, existing technologies also need to be improved in intelligent control and safety protection. There is a lack of precise operation control strategies and adaptive adjustment mechanisms, and the dynamic optimization of equipment operation parameters cannot be achieved. The safety protection system often uses fixed threshold judgment, lacking flexible fault protection strategies and remote supervision capabilities.

[0005] In terms of human-machine interaction and remote supervision, existing systems generally have problems such as unintuitive information display and untimely alarm response. There is a lack of a complete Internet of Things communication architecture, making it impossible to achieve real-time transmission and remote supervision of the safety information of construction hoists. Solving these problems is of great significance for improving the safety and management efficiency of construction hoist use. Summary of the Invention

[0006] Aiming at the problems in the prior art, this application provides a safety control method and device for construction hoists based on multi-sensor perception, which can break through the limitations of traditional hoist management models and provide an intelligent overall solution for construction safety management.

[0007] To solve at least one of the above problems, this application provides the following technical solutions: In the first aspect, this application provides a safety control method for construction hoists based on multi-sensor perception, including: Collect the facial images of construction workers through a binocular live detection camera and the voiceprint information through a voiceprint collector. Input the facial images into a personnel verification model for identity feature extraction, and input the voiceprint information into a voiceprint verification model for voiceprint feature extraction. Based on a deep neural network, perform double verification on the identity features and the voiceprint features respectively with the preset information of the construction workers. At the same time, collect the images of the wearing status of the safety equipment of the construction workers, input the images of the wearing status of the safety equipment into a posture recognition model, generate the authority information of the construction workers according to the identity verification result, the voiceprint verification result and the posture recognition result, and input the authority information into the elevator control authorization unit; Install weight sensors, position sensors, speed sensors, vibration sensors, limit sensors, door switch sensors and safety clamp sensors at the bottom of the car, on the side wall of the guide rail and at the edge of the door frame of the construction elevator. Collect the load data, position data, speed data, vibration data, travel data, door status data and braking data of the construction elevator. Input the collected multi-dimensional data into a data preprocessing unit for noise reduction filtering and standardization processing, and input the preprocessed data into an equipment status evaluation model according to preset weights. Generate the operation status information of the construction elevator according to the status parameters output by the equipment status evaluation model; Display the operation status information on the monitoring touch screen. At the same time, input the operation status information and the authority information of the construction workers into an intelligent control unit. The intelligent control unit generates an operation control instruction based on the load data, position data and speed data of the construction elevator, adjusts the braking distance and leveling accuracy of the construction elevator according to the operation control instruction, and triggers a fault protection mechanism when the operation status information deviates from the preset threshold. Control the safety clamp sensor to perform braking through the fault protection mechanism and send an alarm signal to the safety supervision system through the Internet of Things communication module.

[0008] Further, the process of collecting the facial images of construction workers through a binocular live detection camera and the voiceprint information through a voiceprint collector, inputting the facial images into a personnel verification model for identity feature extraction, inputting the voiceprint information into a voiceprint verification model for voiceprint feature extraction, and performing double verification on the identity features and the voiceprint features respectively with the preset information of the construction workers based on a deep neural network includes: Perform image enhancement preprocessing on the facial images collected by the binocular live detection camera. Input the preprocessed facial images into a convolutional neural network for face key point detection, extract the coordinates of the facial feature points including the contours of eyes, nose and mouth, construct a facial geometric feature vector based on the coordinates of the facial feature points, input the facial geometric feature vector into a personnel verification model to generate an identity feature code, and encrypt and store the identity feature code; Collect continuous voice segments of construction workers through a voiceprint collector, perform frame addition and windowing processing on the voice segments, extract Mel-frequency cepstral coefficient features from the voice segments, input the extracted cepstral coefficient features into a voiceprint verification model to generate a voiceprint feature vector, respectively import the identity feature code and the voiceprint feature vector into a deep neural network, calculate the feature similarity score with the preset identity information database through the deep neural network, and compare the similarity score with a preset matching threshold to generate a verification result.

[0009] Further, collect the wearing state images of the protective equipment of construction workers, input the wearing state images of the protective equipment into a pose recognition model, and generate the permission information of construction workers according to the identity verification result, the voiceprint verification result and the pose recognition result, and input the permission information into the elevator control authorization unit, including: Collect the front and side image sequences of construction workers, perform object detection and image segmentation processing on the image sequences, extract the contour features of the protective equipment including safety helmets, reflective vests and safety belts, input the contour features into a pose recognition model based on human skeleton key points, and calculate the spatial position relationship between the protective equipment and the human skeleton key points through the pose recognition model to generate a wearing state feature map of the protective equipment; Input the wearing state feature map of the protective equipment, the identity verification result and the voiceprint verification result into a permission evaluation model, comprehensively analyze the identity level and compliance with the protection specification of construction workers through the permission evaluation model, generate permission information including the operation permission level and operation time limit according to the analysis result, write the permission information into the data register of the elevator control authorization unit, and establish a mapping relationship between the identity of construction workers and the operation permission.

[0010] Further, arrange weight sensors, position sensors, speed sensors, vibration sensors, limit sensors, door switch sensors and safety clamp sensors at the bottom of the car, on the side wall of the guide rail and at the edge of the door frame of the construction elevator, collect the load data, position data, speed data, vibration data, travel data, door state data and braking data of the construction elevator, and input the collected multi-dimensional data into a data preprocessing unit for noise reduction filtering and standardization processing, including: Install a weight sensor at the bottom of the car of the construction elevator, arrange position sensors, speed sensors and vibration sensors on the side wall of the guide rail, and set limit sensors, door switch sensors and safety clamp sensors at the edge of the door frame. Collect the load data through the weight sensor, collect the position data through the position sensor, collect the speed data through the speed sensor, collect the vibration data through the vibration sensor, collect the travel data through the limit sensor, collect the door state data through the door switch sensor, and collect the braking data through the safety clamp sensor; Input the collected multi-dimensional data into the data preprocessing unit, perform noise reduction processing on the multi-dimensional data using wavelet transform, filter out high-frequency interference signals through a Butterworth low-pass filter, perform zero-mean normalization processing on the filtered data, and align the normalized data in time series according to the sampling timestamp to generate a preprocessed data matrix.

[0011] Further, input the preprocessed data into the equipment status evaluation model according to the preset weights, and generate the operation status information of the construction elevator based on the status parameters output by the equipment status evaluation model, including: Construct a preset weight matrix for the equipment status evaluation model, allocate weight coefficients according to the influence degree of each sensor data on the operation status of the construction elevator, perform matrix multiplication operation on the preprocessed data matrix and the preset weight matrix to generate a weighted feature vector, extract time series features of the weighted feature vector through a recurrent neural network, and input the extracted time series features into the equipment status evaluation model; Based on the load parameters, running speed parameters, position parameters, vibration parameters, door status parameters, and braking parameters output by the equipment status evaluation model, combine the parameters to construct a state feature space, and perform non-linear mapping on the state feature space through a deep neural network to generate operation status information including equipment operation status, component wear status, and safety threshold status.

[0012] Further, display the operation status information on the monitoring touch screen, and at the same time input the operation status information and the construction personnel permission information into the intelligent control unit. The intelligent control unit generates an operation control instruction based on the load data, position data, and speed data of the construction elevator, and adjusts the braking distance and leveling accuracy of the construction elevator according to the operation control instruction, including: Construct an operation status display interface of the construction elevator on the monitoring touch screen, convert the equipment operation status, component wear status, and safety threshold status in the operation status information into graphical display data through a data visualization module, and at the same time import the operation permission level and operation time limit in the operation status information and the construction personnel permission information into the data buffer of the intelligent control unit; Input the load data, position data, and speed data into the data analysis module of the intelligent control unit, perform real-time analysis on the data through a fuzzy neural network, establish a speed control model based on the load and displacement, compare the calculation result of the speed control model with the rated parameters of the construction elevator, and generate an operation control instruction including braking torque and leveling compensation value.

[0013] Further, when the operating status information deviates from a preset threshold, a fault protection mechanism is triggered. The fault protection mechanism controls the safety clamp sensor to perform braking and sends an alarm signal to the safety supervision system through the Internet of Things communication module, including: Real-time monitor the equipment operation status, component wear status, and safety threshold status in the operating status information. Compare the monitored data with the preset fault determination threshold. When any status parameter exceeds the preset threshold, start the fault protection mechanism, send a braking control instruction to the safety clamp sensor, and drive the brake to implement emergency braking through the safety clamp sensor; Input the deviation value between the status parameter that triggers the fault protection mechanism and the preset threshold into the Internet of Things communication module. Encrypt and package the deviation data through the Internet of Things communication module, establish a communication link with the safety supervision system, form an alarm data packet with the encrypted deviation data and the equipment number information, and send the alarm data packet to the safety supervision system through the communication link.

[0014] In a second aspect, the present application provides a construction elevator safety control device based on multi-sensor perception, including: A biological detection module, which is used to collect the facial images of construction workers through a binocular live detection camera and the voiceprint information through a voiceprint collector. Input the facial images into a personnel verification model for identity feature extraction, input the voiceprint information into a voiceprint verification model for voiceprint feature extraction, perform double verification on the identity features and the voiceprint features respectively with the preset information of the construction workers based on a deep neural network, and at the same time collect the images of the wearing status of the safety equipment of the construction workers. Input the images of the wearing status of the safety equipment into a posture recognition model, generate the permission information of the construction workers according to the identity verification result, the voiceprint verification result, and the posture recognition result, and input the permission information into the elevator control authorization unit; A status monitoring module, which is used to deploy weight sensors, position sensors, speed sensors, vibration sensors, limit sensors, door switch sensors, and safety clamp sensors at the bottom of the car, on the side wall of the guide rail, and at the edge of the door frame of the construction elevator, collect the load data, position data, speed data, vibration data, travel data, door status data, and braking data of the construction elevator, input the collected multi-dimensional data into a data preprocessing unit for noise reduction filtering and standardization processing, input the preprocessed data into an equipment status evaluation model according to the preset weights, and generate the operating status information of the construction elevator according to the status parameters output by the equipment status evaluation model; The safety control module is used to display the operation status information on the monitoring touch screen, and at the same time input the operation status information and the construction personnel permission information into the intelligent control unit. The intelligent control unit generates an operation control instruction based on the load data, position data and speed data of the construction elevator, adjusts the braking distance and leveling accuracy of the construction elevator according to the operation control instruction, triggers a fault protection mechanism when the operation status information deviates from the preset threshold, controls the safety clamp sensor to perform braking through the fault protection mechanism, and sends an alarm signal to the safety supervision system through the Internet of Things communication module.

[0015] In a third aspect, the present application provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, the steps of the method for safety control of a construction elevator based on multi-sensor perception are implemented.

[0016] In a fourth aspect, the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for safety control of a construction elevator based on multi-sensor perception are implemented.

[0017] In a fifth aspect, the present application provides a computer program product, including a computer program / instructions. When the computer program / instructions are executed by a processor, the steps of the method for safety control of a construction elevator based on multi-sensor perception are implemented.

[0018] As can be seen from the above technical solutions, the present application provides a method and device for safety control of a construction elevator based on multi-sensor perception. Through multi-modal identity recognition of binocular live detection and voiceprint collection, combined with a deep neural network, accurate verification of personnel identity and the status of protective equipment is achieved. An equipment status evaluation model based on multi-sensor fusion is innovatively designed to realize collaborative monitoring of key parameters such as load, position, and speed. The system adopts an intelligent control strategy to adaptively adjust operation parameters, establishes a real-time fault protection mechanism based on threshold judgment, and realizes remote supervision through Internet of Things technology. This method breaks through the limitations of the traditional elevator management mode and provides an intelligent overall solution for construction safety management. Description of the Drawings

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0020] Figure 1One of the schematic flowcharts of the construction elevator safety control method based on multi-sensor in the embodiment of the present application; Figure 2 Another schematic flowchart of the construction elevator safety control method based on multi-sensor in the embodiment of the present application; Figure 3 Another schematic flowchart of the construction elevator safety control method based on multi-sensor in the embodiment of the present application; Figure 4 Another schematic flowchart of the construction elevator safety control method based on multi-sensor in the embodiment of the present application; Figure 5 Another schematic flowchart of the construction elevator safety control method based on multi-sensor in the embodiment of the present application; Figure 6 Another schematic flowchart of the construction elevator safety control method based on multi-sensor in the embodiment of the present application; Figure 7 Another schematic flowchart of the construction elevator safety control method based on multi-sensor in the embodiment of the present application; Figure 8 The structural diagram of the construction elevator safety control device based on multi-sensor in the embodiment of the present application; Figure 9 The structural schematic diagram of the electronic device in the embodiment of the present application.

[0021] Reference numerals: Electronic device 9600, central processing unit 9100, memory 9140, communication module 9110, input unit 9120, audio processor 9130, display 9160, power supply 9170, buffer memory 9141, application / function storage unit 9142, data storage unit 9143, driver program storage unit 9144, antenna 9111, speaker 9131, microphone 9132. Detailed implementation manners

[0022] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Apparently, the described embodiments are some but not all of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the scope of protection of the present application.

[0023] In the technical solutions of the present application, the acquisition, storage, use, processing, etc. of data all comply with the relevant provisions of national laws and regulations.

[0024] In view of the problems existing in the prior art, the present application provides a safety control method and device for construction elevators based on multi-modal perception. Through multi-modal identity recognition of binocular live detection and voiceprint acquisition, combined with a deep neural network, accurate verification of personnel identity and the status of protective equipment is achieved. An equipment status evaluation model based on multi-sensor fusion is innovatively designed to realize collaborative monitoring of key parameters such as load, position, and speed. The system adopts an intelligent control strategy to adaptively adjust operating parameters, establishes a real-time fault protection mechanism based on threshold judgment, and realizes remote supervision through Internet of Things technology. This method breaks through the limitations of traditional elevator management modes and provides an intelligent overall solution for construction safety management.

[0025] In order to break through the limitations of traditional elevator management modes and provide an intelligent overall solution for construction safety management, the present application provides an embodiment of a safety control method for construction elevators based on multi-modal perception. Refer to Figure 1 , the safety control method for construction elevators based on multi-modal perception specifically includes the following content: Step S101: Collect the facial images of construction workers through a binocular live detection camera and the voiceprint information through a voiceprint collector. Input the facial images into a personnel verification model for identity feature extraction, and input the voiceprint information into a voiceprint verification model for voiceprint feature extraction. Based on a deep neural network, double-verify the identity features and the voiceprint features with the preset information of the construction workers respectively. At the same time, collect the images of the wearing status of the protective equipment of the construction workers, input the images of the wearing status of the protective equipment into a pose recognition model, generate the permission information of the construction workers according to the identity verification result, the voiceprint verification result, and the pose recognition result, and input the permission information into the elevator control authorization unit; Optionally, in this embodiment, a binocular live detection camera is installed at the entrance of the construction elevator, and binocular depth imaging technology is used to realize 3D face reconstruction. The resolution of each camera is 2048×1536 pixels, the field of view angle is 85 degrees, and the baseline distance between the two cameras is 12 cm. Binocular calibration is performed to ensure that the optical axes are parallel. The system collects visible light images and infrared images in the 850nm band, and combines depth information to realize live detection. For example, when the face depth is detected to be discontinuous or the depth value is abnormal, it is determined as a flat deception such as a photo.

[0026] In the preprocessing of the facial images in this embodiment, first, adaptive histogram equalization is performed to enhance the contrast. For images with uneven illumination, they are divided into multiple 8×8 local areas for enhancement respectively. Then, a Gaussian low-pass filter is used to eliminate noise, and the standard deviation is set to 1.5 to ensure that facial detail features are retained while denoising. Then, based on the YCbCr skin color model, rough positioning of the face area is performed, and edge detection is combined with the Sobel operator to achieve accurate segmentation.

[0027] In this embodiment, an improved ResNet-50 network structure is adopted in the personnel verification model. The network consists of 5 stages, each stage using a different number of residual blocks, and the residual connections effectively alleviate the problem of gradient disappearance. The attention mechanism is realized through the combination of the spatial attention module and the channel attention module, focusing on key facial regions such as eyes, nose, and mouth. The model is trained using the ArcFace loss function, and the inter-class distance is increased by adding angular margins.

[0028] In this embodiment, in the voiceprint acquisition link, an 8-channel circular microphone array is adopted, and the microphone spacing is 4.5 cm. Through the delay-and-sum beamforming algorithm, the sound source signal in the target direction is enhanced, and the interference in other directions is suppressed. The construction workers need to read preset passwords such as "Construction safety first, work safety is important", and through the double-threshold endpoint detection algorithm, the energy threshold is set to 2 times the background noise energy, and the zero-crossing rate threshold is 25 times per frame, accurately segmenting the effective speech segment.

[0029] In this embodiment, during the voiceprint feature extraction process, the speech signal is pre-emphasized, and the pre-emphasis coefficient is set to 0.97. Frames are segmented with a frame length of 25 ms and a frame shift of 10 ms, and a Hamming window is used to reduce spectral leakage. 13-dimensional Mel-frequency cepstral coefficients are extracted through 32 triangular Mel filter banks, and the first-order and second-order difference coefficients are calculated to obtain a 39-dimensional feature vector. To improve robustness, cepstral mean subtraction is used for channel compensation.

[0030] In this embodiment, in the voiceprint verification model, a temporal convolutional network based on ResNet is adopted. The network contains multiple residual blocks, and each residual block contains two one-dimensional convolutional layers with a kernel size of 3. The receptive field is expanded through dilated convolution, effectively modeling long-term temporal dependencies. The model is trained using the GE2E loss function, optimizing both the speaker recognition and utterance verification tasks simultaneously. The verification threshold is determined through the ROC curve, balancing the false alarm rate and the miss alarm rate.

[0031] In this embodiment, in the detection of protective equipment, an improved YOLOv3 algorithm is adopted. The backbone network uses DarkNet-53, and multi-scale detection is achieved through the feature pyramid structure. The detection head outputs feature maps of three different scales, detecting large, medium, and small targets respectively. For the detection scenario of protective equipment, appropriate anchor box sizes are designed, and 9 optimal anchor boxes are obtained through K-means clustering. The model can detect protective equipment such as safety helmets, reflective vests, and safety belts simultaneously, outputting the bounding box coordinates and confidence levels.

[0032] In this embodiment, an improved HRNet network is adopted in the pose recognition model. Multi-scale features are extracted through parallel multi-resolution sub-networks to maintain high-resolution representation. Seventeen human key points are designed, including the head, torso, and limbs. The spatial relationship between key points is modeled through a graph convolutional network, and the adjacency matrix of the graph is predefined according to the human skeleton structure. The model outputs a heatmap representing the probability distribution of key point positions, and the positioning accuracy is improved by Gaussian kernel smoothing.

[0033] In the permission information generation stage of this embodiment, an attention-based multi-modal feature fusion network is adopted. Facial features, voiceprint features, and pose features are first mapped to the same dimension through independent fully connected layers, and then the correlation weights between features are calculated through a self-attention mechanism. According to the safety management requirements of the construction site, the permission levels are divided into three levels: administrator, operator, and ordinary user, and each level of permission corresponds to different operation scopes and time limits.

[0034] In the elevator control authorization unit of this embodiment, fine-grained permission control based on RBAC is implemented. The permission information includes fields such as user ID, permission level, operation scope, and validity period, and is encrypted and stored using the RSA algorithm. The authorization unit maintains a permission policy table, defines the set of operations that can be executed for different permission levels, and supports dynamic adjustment and real-time verification of permissions. Through this solution, reliable verification of the identities of construction workers and protective equipment is achieved, ensuring the safe use of the lift.

[0035] Step S102: Install weight sensors, position sensors, speed sensors, vibration sensors, limit sensors, door switch sensors, and safety clamp sensors at the bottom of the car, on the side wall of the guide rail, and at the edge of the door frame of the construction lift. Collect the load data, position data, speed data, vibration data, travel data, door status data, and braking data of the construction lift. Input the collected multi-dimensional data into the data preprocessing unit for noise reduction filtering and standardization processing. Input the preprocessed data into the equipment status evaluation model according to the preset weights, and generate the operation status information of the construction lift according to the status parameters output by the equipment status evaluation model; Optionally, in this embodiment, four high-precision tension type weight sensors are installed at the bottom of the construction lift car. The sensors adopt a full-bridge strain gauge design, with a measuring range of 0 - 5000 kg and a resolution of 0.1 kg. The sensors are arranged at the four corners of the bottom of the car to eliminate the influence of eccentric load through multi-point measurement. The output signal of each sensor is sampled by a 24-bit AD converter, and the sampling frequency is set to 100 Hz to achieve real-time acquisition of load data.

[0036] In this embodiment, an optoelectronic encoder is arranged on the side wall of the guide rail as a position sensor, and a combined design of an incremental encoder and an absolute encoder is adopted. The resolution of the incremental encoder is 4096 pulses / revolution, which is used for high-precision displacement measurement; the absolute encoder uses a 17-bit Gray code for power-on positioning. The signals of the two encoders are collected through a dedicated interface circuit to achieve redundant backup of position data.

[0037] In this embodiment, a Hall effect speed sensor is installed on the side wall of the guide rail. The sensor adopts a differential output method to improve the anti-interference ability. The speed signal is processed by a band-pass filter with a cut-off frequency of 0.5 - 50 Hz to filter out power frequency interference and high-frequency noise. The sampling frequency of the sensor is 200 Hz, and speed and acceleration information are obtained through digital integration and differentiation operations.

[0038] In this embodiment, a three-axis MEMS vibration sensor is installed at the connection between the bottom of the car and the guide rail, with a measurement range of ±16g and a bandwidth of 5 kHz. The sensor adopts a distributed layout scheme, and multiple measuring points are arranged at key structural positions to collect the spatial distribution characteristics of vibration signals through a sensor array. Signal acquisition adopts a synchronous trigger mode to ensure the timing consistency of multi-channel data.

[0039] In this embodiment, stroke limit sensors are installed at the top and bottom of the lift guide rail, and a normally closed safety contact design is adopted. The limit sensor is used in conjunction with a mechanical buffer device. When the car approaches the limit position, an emergency brake is triggered through the sensor. The output signal of the sensor undergoes optoelectronic isolation processing to improve electrical safety.

[0040] In this embodiment, Hall door switch sensors are installed at the car door and the landing door, and a dual-redundancy design is adopted to ensure reliability. The sensor detects the meshing state of the door lock mechanism, and the output signal is processed by a logic judgment circuit to achieve real-time monitoring of the door state. At the same time, current detection is set in the door machine drive circuit to monitor the operation state of the door machine.

[0041] In this embodiment, a displacement sensor and a force sensor are installed on the safety gear mechanism to monitor the action state and braking force of the safety gear in real time. The displacement sensor uses a linear potentiometer to detect the position of the safety gear wedge block; the force sensor adopts a piezoelectric design to monitor the magnitude of the braking force. The sensor signals are processed by a signal conditioning circuit to achieve precise monitoring of the braking process.

[0042] In the data preprocessing stage of this embodiment, wavelet packet decomposition is used for signal denoising. The vibration signal is decomposed by wavelet for 5 layers, the db4 wavelet basis is selected, and the high-frequency coefficients are processed by the soft threshold method. The Kalman filtering algorithm is used for the position and speed signals, and the filter parameters are dynamically adjusted through real-time noise estimation. The preprocessed data is subjected to amplitude normalization and time alignment.

[0043] In this embodiment, a deep convolutional neural network structure is adopted in the equipment status evaluation model. The model input includes the time-series data of 7 sensor channels, and the time-domain features are extracted through a one-dimensional convolutional layer. The network designs residual connections and attention mechanisms to enhance the ability to identify key patterns. The model is trained using the historical operation data of the equipment, and the labels include three states: normal, warning, and failure.

[0044] In terms of weight allocation in this embodiment, the importance weights of the data of each sensor are determined based on expert experience and data analysis. The weights of the load and position data are relatively high, which are directly related to the operation safety; the vibration and speed data are the next, which are used for status monitoring; the door status and limit data are used as auxiliary judgment bases. The weight values are optimized and adjusted through the model verification process.

[0045] In this embodiment, through multi-sensor collaborative perception and deep learning analysis, the accurate evaluation of the operation status of the lift is realized. This solution can timely detect equipment abnormalities, predict potential faults, and provide a decision-making basis for preventive maintenance. In practical applications, the reliability and safety of equipment operation are significantly improved.

[0046] Step S103: Display the operation status information on the monitoring touch screen, and at the same time input the operation status information and the construction worker permission information into the intelligent control unit. The intelligent control unit generates an operation control instruction based on the load data, position data, and speed data of the construction lift, adjusts the braking distance and leveling accuracy of the construction lift according to the operation control instruction, triggers a fault protection mechanism when the operation status information deviates from the preset threshold, controls the safety clamp sensor to execute braking through the fault protection mechanism, and sends an alarm signal to the safety supervision system through the Internet of Things communication module.

[0047] Optionally, in this embodiment, a 15.6-inch capacitive touch display screen is installed in the construction lift control room, with a resolution of 1920×1080 pixels and an industrial-grade dustproof and waterproof design. The display interface adopts a partition layout. The main area displays the real-time operation parameters, including the current load, operation speed, position information, and door status; the status area displays the equipment health and fault warning information; the operation area provides emergency braking and remote alarm buttons.

[0048] In this embodiment, a dual-CPU redundancy design is adopted in the intelligent control unit. The main CPU is responsible for operation control, and the standby CPU performs status monitoring. The controller adopts a real-time operating system, and the task scheduling cycle is 10 ms to ensure the timely response of control instructions. The control unit communicates with each sensor module through the CAN bus, and uses differential signal transmission to improve the anti-interference ability.

[0049] In the operation control strategy of this embodiment, an adaptive control algorithm based on a fuzzy neural network is implemented. The controller inputs include load data, position data, and speed data, and the optimal control quantity is calculated through fuzzy inference rules. For example, when the load is large, the acceleration and deceleration are automatically reduced; when approaching the target floor, the braking torque is dynamically adjusted according to the distance to achieve smooth docking.

[0050] In the braking control process of this embodiment, a feedforward-feedback combined control strategy is adopted. The feedforward control estimates the braking force demand based on the load characteristics and position information, and the feedback control adjusts the braking force magnitude through the real-time speed deviation. The controller outputs a PWM signal to drive the brake, and the PWM frequency is 20 kHz. The continuous adjustment of the braking force is achieved by changing the duty cycle.

[0051] In the leveling control of this embodiment, a two-stage deceleration strategy is designed. When the distance to the target floor is 2 meters, it enters the coarse adjustment stage, and the speed is reduced to 30% of the rated speed; when the distance is 0.3 meters, it enters the fine adjustment stage, and the crawling speed is used to approach the target position. The docking error is corrected in real time through the feedback signal of the position sensor to ensure the leveling accuracy.

[0052] In the fault protection mechanism of this embodiment, a multi-level threshold judgment standard is established. The warning threshold and alarm threshold of the operating parameters are set. When the parameter exceeds the warning threshold, the controller starts preventive protection measures; when it exceeds the alarm threshold, the emergency braking is immediately triggered. For example, when the speed is over-limit or the load is overweight, the safety clamp brake is automatically activated.

[0053] In the safety clamp control circuit of this embodiment, a watchdog circuit and redundant power supply design are adopted. The control signal is protected by optical isolation and relay to ensure the reliable execution of the braking instruction. After the safety clamp acts, the braking effect is monitored through a displacement sensor and a force sensor to form a closed-loop control.

[0054] In the Internet of Things communication module of this embodiment, a 4G / 5G dual-mode communication scheme is adopted, which supports the automatic switching of communication networks. The alarm data adopts an encrypted transmission protocol, and the data packet includes the device ID, fault type, fault parameters, and timestamp. The communication module sets a heartbeat packet mechanism to monitor the connection status with the supervision platform in real time.

[0055] In the early warning information processing flow of this embodiment, a hierarchical early warning mechanism is implemented. According to the severity of the fault, it is divided into three levels: prompt, early warning, and alarm. Each level of early warning corresponds to different processing strategies. The prompt-level fault records the log and continues to run; the early warning-level fault limits the running speed; the alarm-level fault executes the emergency braking and reports to the supervision platform.

[0056] In this embodiment, operation control based on permissions is implemented in the human-computer interaction design. Operators with different permission levels have different operation permission scopes. For example, an administrator can modify control parameters, while an ordinary operator can only perform basic operation operations. The operation logs are recorded in real time and uploaded to the supervision platform.

[0057] This embodiment effectively guarantees the operation safety of the construction elevator through a multi-level security protection mechanism and intelligent control strategy. This solution can detect and handle various abnormal situations in a timely manner, and realize the whole-process monitoring of the equipment operation through remote supervision, significantly improving the operation reliability and safety of the construction elevator, and providing a reliable guarantee for the transportation of personnel and materials at the construction site.

[0058] As can be seen from the above description, the multi-sensor perception-based construction elevator safety control method provided by the embodiment of the present application can accurately verify the personnel identity and the status of protective equipment through multi-modal identity recognition of binocular live detection and voiceprint acquisition, and combine a deep neural network. An equipment status evaluation model based on multi-sensor fusion is innovatively designed to realize the collaborative monitoring of key parameters such as load, position, and speed. The system adopts an intelligent control strategy to adaptively adjust operation parameters, establishes a real-time fault protection mechanism based on threshold judgment, and realizes remote supervision through Internet of Things technology. This method breaks through the limitations of the traditional elevator management mode and provides an intelligent overall solution for construction safety management.

[0059] In an embodiment of the multi-sensor perception-based construction elevator safety control method of the present application, see Figure 2 , it may specifically include the following content: Step S201: Perform image enhancement preprocessing on the facial image collected by the binocular live detection camera, input the preprocessed facial image into a convolutional neural network for face key point detection, extract the coordinates of facial feature points including the contours of eyes, nose, and mouth, construct a facial geometric feature vector based on the facial feature point coordinates, input the facial geometric feature vector into a personnel verification model to generate an identity feature code, and encrypt and store the identity feature code; Step S202: Collect continuous voice segments of construction workers through a voiceprint collector, perform frame addition and windowing processing on the voice segments, extract Mel frequency cepstral coefficient features from the voice segments, input the extracted cepstral coefficient features into a voiceprint verification model to generate a voiceprint feature vector, respectively import the identity feature code and the voiceprint feature vector into a deep neural network, calculate the similarity score with the features in a preset identity information database through the deep neural network, and compare the similarity score with a preset matching threshold to generate a verification result.

[0060] Optionally, in this embodiment, a binocular live detection camera is first used to synchronously collect the front and side images of the construction workers. The camera is designed with infrared fill light to ensure the imaging quality in low-light environments. The binocular camera obtains the internal and external parameter matrices through calibration to achieve depth information reconstruction, effectively preventing deception means such as flat photos. The camera resolution is 2048×1536 pixels, the frame rate is 30fps, and it supports the autofocus function.

[0061] In the image enhancement preprocessing stage of this embodiment, first, light compensation is performed, and an adaptive enhancement algorithm based on the brightness histogram is used. For backlight or shadow areas, local details are enhanced through gamma correction. Then, bilateral filtering is used for noise reduction while maintaining edge information. The spatial domain standard deviation of the filter is set to 3, and the value domain standard deviation is set to 25 to achieve a balance between noise suppression and edge preservation.

[0062] In the face detection and alignment stage of this embodiment, an improved MTCNN cascaded convolutional network is used. The network includes three sub-networks: P-Net, R-Net, and O-Net, which respectively complete candidate box generation, candidate box optimization, and key point localization. Each sub-network includes a convolutional layer, a pooling layer, and a fully connected layer, and detects faces of different sizes through a multi-scale feature pyramid.

[0063] In the key point detection process of this embodiment, 68 facial feature points are located, including 12 points for the eye contour, 10 points for the eyebrows, 9 points for the nose, 20 points for the mouth, and 17 points for the face contour. The heat map regression method is used to predict the positions of the feature points, and multi-level features are extracted through a deep residual network. The network uses the wing loss function to improve the training weights of small deviation samples.

[0064] In the feature vector construction stage of this embodiment, geometric features are calculated based on the detected key points. These include the eye spacing ratio, the nose-mouth distance ratio, the symmetry of the face contour, etc. At the same time, local texture features are extracted, and the LBP operator is used to describe the facial details. The geometric features and texture features are concatenated to form a 512-dimensional feature vector.

[0065] In the voiceprint acquisition link of this embodiment, a high-fidelity digital microphone is used, with a sampling rate of 16kHz and a quantization precision of 24 bits. The environmental noise is suppressed through a dual-channel noise reduction algorithm, and the adaptive beamforming technology is used to enhance the target sound source. The construction workers are required to read a preset password, and a 3-5 second voice segment is collected.

[0066] In the voice preprocessing of this embodiment, framing is performed with a frame length of 25ms and a frame shift of 10ms. The Hamming window is used to reduce spectral leakage, and the pre-emphasis coefficient is set to 0.97. Effective voice segments are detected through the double thresholds of short-time energy and zero-crossing rate, and silent segments are removed. Endpoint detection and silence removal are performed on the voice signal.

[0067] In this embodiment, during the feature extraction process, the power spectrum is calculated through a 32-point FFT, and 40 triangular filter banks are used to extract Mel spectrum features. 13-dimensional Mel-frequency cepstral coefficients are obtained through discrete cosine transform. The first-order and second-order difference coefficients are calculated to form a 39-dimensional feature vector. Cepstral mean subtraction is used for channel compensation.

[0068] In this embodiment of the voiceprint verification model, a temporal convolutional network based on ResNet is adopted. The network contains multiple residual blocks, and each residual block contains two one-dimensional convolutional layers and a batch normalization layer. Dilated convolution is used to expand the receptive field and model long temporal dependencies. The triplet loss function is used for model training to increase the inter-class distance and reduce the intra-class distance.

[0069] In the feature fusion stage of this embodiment, facial features and voiceprint features are input into a deep neural network. The network uses an attention mechanism to dynamically adjust the weights of the two features and adaptively fuse them according to the feature reliability. The cosine similarity with the preset feature library is calculated to generate a similarity score. The verification threshold is determined by optimizing the ROC curve to balance the false alarm rate and the missed alarm rate.

[0070] In this embodiment, the verified identity feature code is encrypted and stored using the AES-256 algorithm, and the key is managed by a hardware encryption module. The storage record includes the feature code, timestamp, and verification result, supporting both offline verification and online verification mechanisms. This solution realizes the reliable verification of the construction personnel's identity and effectively prevents the behavior of misusing and substituting for work.

[0071] In an embodiment of the construction elevator safety control method based on multi-sensor perception in this application, refer to Figure 3 , and it may specifically include the following content: Step S301: Collect the front and side image sequences of construction personnel, perform object detection and image segmentation processing on the image sequences, extract the contour features of protective equipment including safety helmets, reflective vests, and safety belts, input the contour features into a pose recognition model based on human skeletal key points, and calculate the spatial position relationship between the protective equipment and the human skeletal key points through the pose recognition model to generate a protective equipment wearing status feature map; Step S302: Input the protective equipment wearing status feature map, the identity verification result, and the voiceprint verification result into a permission evaluation model. Through the permission evaluation model, comprehensively analyze the identity level of construction personnel and the compliance with protection specifications, generate permission information including the operation permission level and operation time limit according to the analysis result, write the permission information into the data register of the elevator control authorization unit, and establish a mapping relationship between the construction personnel's identity and operation permissions.

[0072] Optionally, in this embodiment, two industrial-grade high-speed cameras are arranged at the entrance of the construction lift to collect the front and side images of construction workers respectively. The cameras adopt a global shutter design and are equipped with anti-shake gyroscopes to achieve clear imaging in high-speed motion scenarios. The light compensation system includes an adaptive LED array that dynamically adjusts the light compensation parameters according to the ambient light intensity to ensure the consistency of image quality at different times.

[0073] In this embodiment, the collected image sequence is first preprocessed. Adaptive histogram equalization is used to enhance the contrast, and local enhancement is performed on the shadow and overexposed areas. Bilateral filtering is used for noise reduction, and the standard deviation in the spatial domain changes with the image gradient, effectively suppressing noise while maintaining edge details. The image dynamic range is optimized through gamma correction to enhance the details in the dark areas.

[0074] In the object detection stage of this embodiment, an improved YOLOv5 network is adopted to optimize the network structure for the detection scenario of construction site protection equipment. The backbone network uses CSPDarknet53 to reduce the computational load through the cross-stage partial network. Three detection heads are designed. The large-scale detection head is responsible for large objects such as reflective vests, the medium-scale detection head is responsible for safety helmets, and the small-scale detection head is responsible for fine components such as seat belt fasteners.

[0075] In the training of the detection network in this embodiment, a dataset of protection equipment including different lighting conditions, angles, and occlusions is constructed. A multi-scale training strategy is adopted, and the input image size randomly varies between 416 and 832. The loss function comprehensively considers bounding box regression, confidence prediction, and class classification, and increases the weight for the detection of key parts. The detection network is pre-trained through transfer learning to improve the generalization ability of the model in small-sample scenarios.

[0076] In the image segmentation link of this embodiment, an improved Mask R-CNN model is adopted to achieve precise segmentation of protection equipment. The network uses ResNet101-FPN as the feature extraction backbone and fuses multi-scale features through the feature pyramid network. The ROIAlign layer maintains the spatial accuracy of the feature map and improves the segmentation effect of small objects. The segmentation head adopts a four-layer convolutional structure to output a high-precision mask of 28×28.

[0077] In the contour feature extraction stage of this embodiment, morphological processing and contour extraction are performed on the segmentation mask. Noise points are removed through opening operations, holes are filled through closing operations, and the reconstruction operation maintains the shape of the object. A curvature-based contour simplification algorithm is adopted to retain key contour points. Shape descriptors are calculated, including geometric features such as perimeter ratio, compactness, and eccentricity, for evaluating the standardization of protection equipment.

[0078] In this embodiment, in the pose recognition model, an improved HRNet structure is adopted to implement the detection of human skeleton key points. The network includes four parallel branches to maintain feature maps with different resolutions. Through repeated multi-scale fusion modules, multi-scale context information is fused on the high-resolution feature maps. Heat maps of 17 key points are output, and the positioning accuracy is improved by Gaussian kernel smoothing.

[0079] In this embodiment, a wearing state evaluation method based on spatial geometric constraints is designed. A spatial association model between the protective equipment and human key points is established, such as the relative position and inclination angle between the safety helmet and the head key points, the overlap degree between the reflective vest and the torso contour, and the fixed relationship between the safety belt and the hip position. By calculating the deviation values of geometric features, the standard degree of wearing is evaluated.

[0080] In this embodiment, in the permission evaluation model, a multi-modal feature fusion architecture based on graph neural network is adopted. The state features of protective equipment, authentication results, and voiceprint features are constructed as nodes, and the association strength between nodes is calculated through the attention mechanism. The model iteratively updates the node features and finally outputs the permission level score and operation time limit suggestion. According to the safety management specifications of the construction site, differentiated permission access standards are set.

[0081] In this embodiment, in the ladder control authorization link, dynamic permission management based on spatio-temporal constraints is realized. The permission information includes fields such as user identity, level, operation range, timeliness, etc., and a distributed storage structure is adopted to improve the access efficiency. The authorization policy supports multi-dimensional control based on time window, operation area, and equipment status to ensure the safe and standard use of construction elevators. This solution realizes the intelligent association between the wearing state of protective equipment and operation permissions, and establishes a complete access management mechanism for construction personnel.

[0082] In an embodiment of the construction elevator safety control method based on multi-sensor perception of the present application, refer to Figure 4 , it may also specifically include the following content: Step S401: Install a weight sensor at the bottom of the construction elevator car, arrange position sensors, speed sensors, and vibration sensors on the side wall of the guide rail, and set limit sensors, door switch sensors, and safety clamp sensors at the edge of the door frame. Collect load data through the weight sensor, collect position data through the position sensor, collect speed data through the speed sensor, collect vibration data through the vibration sensor, collect travel data through the limit sensor, collect door state data through the door switch sensor, and collect braking data through the safety clamp sensor; Step S402: Input the collected multi-dimensional data into the data preprocessing unit, perform noise reduction processing on the multi-dimensional data using wavelet transform, filter out high-frequency interference signals through a Butterworth low-pass filter, perform zero-mean normalization processing on the filtered data, and align the data in time series according to the sampling timestamp to generate a preprocessed data matrix.

[0083] Optionally, in this embodiment, high-precision tensiometric load sensors are installed at the four corners of the bottom of the construction elevator car. The sensors adopt a full-bridge strain gauge design and are equipped with a temperature compensation circuit to eliminate the influence of temperature drift. The signals of each sensor are sampled by a 24-bit ADC, and the sampling frequency is set to 100 Hz to achieve multi-point real-time acquisition of load data. The sensor signals are processed by an amplification and conditioning circuit, and the differential transmission method is used to improve the anti-interference ability.

[0084] In this embodiment, a combined position sensor is installed on the side wall of the guide rail, including an incremental photoelectric encoder and an absolute magnetic scale. The incremental encoder is used for high-precision displacement measurement with a resolution of up to 0.1 mm; the absolute magnetic scale is used for position calibration and power-on positioning. The signals of the two sensors are collected through a dedicated interface circuit to achieve redundant backup of position data.

[0085] In this embodiment, Hall effect speed sensors are arranged on the side wall of the guide rail, and the differential output method is used to improve the signal quality. The sensor output is processed by a band-pass filter with a cut-off frequency of 0.5 - 50 Hz to effectively filter out power frequency interference and high-frequency noise. The sampling frequency is set to 200 Hz, and speed and acceleration information are obtained simultaneously through digital integration and differentiation operations.

[0086] In this embodiment, a three-axis MEMS vibration sensor array is arranged at the connection between the bottom of the car and the guide rail. The sensors adopt a distributed layout scheme to collect the spatial distribution characteristics of vibration signals. The sampling frequency of the vibration signals is 1 kHz, and the time series consistency of multi-channel data is ensured through synchronous triggering. The signal conditioning circuit uses a programmable gain amplifier to dynamically adjust the amplification factor according to the vibration amplitude.

[0087] In this embodiment, stroke limit sensors are installed at the top and bottom of the elevator guide rail, adopting a normally closed safety contact design. The limit sensors are used in conjunction with mechanical buffer devices to trigger a braking signal when the car approaches the limit position. The sensor output is processed through optoelectronic isolation to ensure electrical safety, and redundant contacts are set to improve reliability.

[0088] In this embodiment, Hall door switch sensors are installed at the car door and the landing door, adopting a dual-redundancy design to ensure safety. The sensors detect the meshing state of the door lock mechanism, and the output signals are processed by a logic judgment circuit to achieve real-time monitoring of the door state. At the same time, current detection is set in the door machine drive circuit to monitor the operation state of the door machine.

[0089] In this embodiment, a displacement sensor and a force sensor are installed on the safety gear mechanism to achieve precise monitoring of the braking state. A linear potentiometer is used as the displacement sensor to detect the position of the safety gear wedge block; a piezoelectric design is adopted for the force sensor to monitor the braking force. The sensor signals are processed by a signal conditioning circuit to achieve closed-loop control during the braking process.

[0090] In the data preprocessing stage of this embodiment, wavelet transform is first used for noise reduction. The vibration signal is decomposed into 5 layers by wavelet transform, the db4 wavelet basis function is selected, and the high-frequency coefficients are processed by the soft threshold method. An improved threshold function is used when reconstructing the signal to avoid the loss of useful information caused by over-smoothing.

[0091] This embodiment designs a fourth-order Butterworth low-pass filter, and the cut-off frequency is dynamically set according to the characteristics of different sensor signals. The filter adopts a two-way filtering strategy to eliminate phase delay and realizes real-time processing through a buffer queue. For mutation signals, an adaptive filtering algorithm is used to maintain the fast response characteristics of the signal.

[0092] In the data standardization process of this embodiment, a sliding window is used to calculate the local mean and standard deviation, and the window length is 1 second. The dimensional difference of different sensors is eliminated by zero-mean normalization to improve the comparability of data. The median replacement method is used to process outliers to maintain the continuity of the data.

[0093] This embodiment realizes a multi-dimensional data alignment mechanism based on timestamps. A unified clock source is used to provide sampling trigger signals to ensure the synchronization of multi-channel data. The data with inconsistent sampling frequencies is processed by an interpolation algorithm to generate an equally spaced time series data matrix. This scheme realizes high-quality acquisition and preprocessing of multi-source sensing data of construction hoists, providing a reliable data basis for subsequent state assessment.

[0094] In an embodiment of the construction hoist safety control method based on multi-sensor perception of the present application, refer to Figure 5 , and it may specifically include the following content: Step S501: Construct a preset weight matrix for the equipment state evaluation model, allocate weight coefficients according to the influence degree of each sensor data on the operating state of the construction hoist, perform matrix multiplication on the preprocessed data matrix and the preset weight matrix to generate a weighted feature vector, extract the time series features of the weighted feature vector through a recurrent neural network, and input the extracted time series features into the equipment state evaluation model; Step S502: Based on the device status evaluation model, output load parameters, operating speed parameters, position parameters, vibration parameters, door status parameters, and braking parameters. Combine these parameters to construct a state feature space, and perform a non-linear mapping on the state feature space through a deep neural network to generate operation status information including the device operation status, component wear status, and safety threshold status.

[0095] Optionally, in this embodiment, a weight assignment mechanism based on expert knowledge is first constructed, and a preset weight matrix of 8×8 is designed. The weight coefficients are determined according to the influence degree of each sensor data on the safety of the elevator operation, and an analytic hierarchy process is used to establish a judgment matrix. For example, load data and speed data have a greater impact on safety, so higher weights are assigned; vibration data is more critical for evaluating the device wear status, and higher weights are given in the corresponding dimension.

[0096] In the weight matrix design of this embodiment, the correlation and reliability of sensor data are considered. The correlation coefficient between sensors is calculated through historical data analysis, and the weights of highly correlated data are appropriately reduced to avoid information redundancy. At the same time, according to the failure rate and measurement error of the sensors, a reliability compensation coefficient is set to dynamically adjust the weight values.

[0097] In this embodiment, an improved LSTM recurrent neural network is used for time series feature extraction. The network includes three layers of LSTM units, with the number of nodes in each layer being 128, 64, and 32 respectively. The gated mechanism is used to selectively remember long-term dependencies. The input layer receives the weighted feature vector, the hidden layer uses dropout to prevent overfitting, and the output layer extracts the time series feature representation.

[0098] In the training of the LSTM network in this embodiment, a sequence-to-sequence training mode is adopted. The length of the input sequence is 60 seconds, and the step size is 1 second. Training samples are generated through a sliding window. The loss function combines the mean square error and the huber loss to balance the influence of outliers. The Adam optimizer is used to dynamically adjust the learning rate to improve the model convergence speed.

[0099] In this embodiment, a deep neural network enhanced by attention is adopted in the state evaluation model. The attention mechanism dynamically adjusts the importance of different time series features according to the current state, improving the sensitivity of the model to abnormal states. The network includes multiple residual blocks, and each residual block consists of two fully connected layers and a batch normalization layer. The gradient vanishing problem is alleviated through skip connections.

[0100] In the process of constructing the feature space in this embodiment, different parameters are mapped to a unified metric space. The min-max normalization is used to scale the parameter values to the interval [0,1], and the kernel function is used to map the low-dimensional features to the high-dimensional space to enhance the feature expression ability. At the same time, the physical constraint relationship between parameters is maintained to ensure the physical meaning of the feature space.

[0101] In this embodiment, a multi-task learning framework is designed to predict the operating state of the device, the wear state of components, and the safety threshold state simultaneously. Each task branch uses an independent fully connected layer and shares the underlying feature extraction network. Through the task weight adaptive adjustment strategy, the training difficulty of different tasks is balanced, and the generalization ability of the model is improved.

[0102] In the evaluation of the operating state in this embodiment, a multi-level state classification standard is established. According to the combined features of parameters such as load, speed, and position, the operating state is divided into three levels: normal, warning, and fault. The prediction results of multiple classifiers are fused through a soft voting mechanism to improve the reliability of state judgment.

[0103] In the evaluation of the component wear state in this embodiment, a wear degree prediction model is established based on the vibration signal characteristics. The characteristic frequency components are extracted through spectrum analysis and combined with time-domain statistical characteristics to predict the wear degree of key components such as guide rails, gears, and bearings. The model outputs the wear state score and the estimated remaining service life.

[0104] In the evaluation of the safety threshold state in this embodiment, a dynamic threshold adjustment strategy is adopted. Based on the historical operation data of the device, a parameter change trend model is established to predict the future change trend of the parameters. When the parameter approaches the safety threshold, the system issues a warning signal in advance to reserve sufficient response time.

[0105] This embodiment realizes the intelligent mapping from multi-dimensional sensing data to the operating state of the device and establishes a complete state evaluation system. By mining the potential correlations between data features through a deep learning model, the accurate evaluation of the operating state of the lift is achieved. This solution can timely detect device abnormalities and predict component failures, providing reliable guarantees for the safe operation of the construction lift.

[0106] In an embodiment of the safety control method for a construction lift based on multi-source perception in this application, refer to Figure 6 , it may also specifically include the following content: Step S601: Construct a display interface for the operating state of the construction lift on the monitoring touch screen, convert the device operating state, component wear state, and safety threshold state in the operating state information into graphical display data through a data visualization module, and at the same time import the operation permission level and operation time limit in the operating state information and the construction personnel permission information into the data buffer of the intelligent control unit; Step S602: Input the load data, position data, and speed data into the data analysis module of the intelligent control unit, perform real-time analysis on the data through a fuzzy neural network, establish a speed control model based on the load and displacement, compare the calculation results of the speed control model with the rated parameters of the construction lift, and generate an operation control instruction including braking torque and leveling compensation value.

[0107] Optionally, in this embodiment, a hierarchical status display interface is designed on the monitoring touch screen. The main interface uses a fan-shaped dashboard to display core operating parameters, including the current load, operating speed, and position information. The dashboard uses color bands to mark the safety threshold intervals, with green indicating the normal operating interval, yellow indicating the warning interval, and red indicating the dangerous interval. The interface layout adopts a responsive design, supporting display devices with different resolutions.

[0108] In this embodiment, the WebGL graphics engine is adopted in the data visualization module to realize the three-dimensional dynamic display of the elevator operation status. The elevator operation trajectory is simulated through skeletal animation, and the position and attitude of the car are updated in real time. Key components such as guide rails, wire ropes, and brakes are rendered semi-transparently to intuitively display the wear status, and different colors represent different wear degrees.

[0109] This embodiment designs a multi-level data caching mechanism, including a cache and persistent storage. The cache adopts a circular buffer structure to store the operation status data of the most recent 5 minutes, supporting fast read and write access. Persistent storage uses a time-series database to compress and encode the status data, realizing the efficient storage and query of long-term data.

[0110] In this embodiment, two layers of fuzzy rules are designed in the fuzzy neural network. The first layer divides the operation conditions based on the load weight, classifying the load weight into four levels: no load, light load, medium load, and heavy load. The second layer divides the operation stages based on the displacement, including three stages: starting and accelerating, uniform running, and decelerating and stopping. The control strategies for different conditions and stages are determined through fuzzy inference.

[0111] In this embodiment, a hybrid learning algorithm is adopted in network training. First, an initial fuzzy rule base is constructed based on expert experience, then the membership function parameters are optimized through the BP algorithm, and finally the rule weights are optimized using the genetic algorithm. The training data includes the operation trajectories under different load conditions to ensure the adaptability of the model to various working conditions.

[0112] In this embodiment, an adaptive control strategy is implemented in the speed control model. The acceleration limit is dynamically adjusted according to the current load weight, and the acceleration is reduced during heavy loads to avoid impacts. Speed planning is achieved through position feedback, and the deceleration process is entered in advance when approaching the target floor. The model outputs the expected speed curve as the set value of the PID controller.

[0113] This embodiment designs a braking control algorithm based on model prediction. Precise control is achieved through a speed-position double closed-loop structure, with a fuzzy controller in the outer loop and a PID controller in the inner loop. The braking torque command is calculated based on the speed error and position error to achieve smooth braking. When approaching the target floor, the influence of mechanical clearance is eliminated through compensation control to improve the leveling accuracy.

[0114] This embodiment realizes a real-time comparison mechanism for operating parameters. A database containing parameters such as rated load, rated speed, and acceleration limit value is established. Before generating the control instruction, safety verification is carried out to ensure that each parameter does not exceed the limit range. When approaching the limit value, the warning mechanism is started, and safety protection actions are executed if necessary.

[0115] This embodiment adopts a distributed architecture in the execution of the control instruction. The main controller generates the reference control instruction, and the on-site controller performs real-time compensation adjustment. The control instruction and status feedback are transmitted through the real-time bus network to ensure the real-time performance of the control loop. This solution realizes the intuitive display and intelligent control of the operating state of the lift, improving the safety and reliability of the equipment operation.

[0116] This embodiment establishes a complete human-machine interaction system. The operator can monitor the equipment status in real time through the touch interface, and the system dynamically adjusts the control authority according to the operation authority level. When the equipment status is abnormal or approaches the safety threshold, the system reminds the operator in a dual way of vision and sound and automatically records the abnormal events. This solution improves the intelligent level of the construction lift and provides a reliable vertical transportation guarantee for the construction site.

[0117] In an embodiment of the safety control method for a construction lift based on multi-sensor perception of the present application, refer to Figure 7 , and it may specifically include the following content: Step S701: Real-time monitor the equipment operation state, component wear state, and safety threshold state in the operation state information, compare the monitoring data with the preset fault determination threshold, and start the fault protection mechanism when any state parameter exceeds the preset threshold. Send a braking control instruction to the safety clamp sensor, and drive the brake to implement emergency braking through the safety clamp sensor; Step S702: Input the deviation value between the state parameter that triggers the fault protection mechanism and the preset threshold into the Internet of Things communication module, encrypt and package the deviation data through the Internet of Things communication module, establish a communication link with the safety supervision system, and combine the encrypted deviation data with the equipment number information to form an alarm data packet. Send the alarm data packet to the safety supervision system through the communication link.

[0118] Optionally, this embodiment designs a multiple fault monitoring mechanism to ensure the operation safety of the construction lift through hierarchical warning and linkage protection of state parameters. The monitoring indicators include key parameters such as car load, running speed, braking torque, vibration amplitude, and door lock state. Three-level warning thresholds are set for each parameter, corresponding to attention, warning, and danger states respectively. The threshold setting is based on equipment specification requirements and historical operation data statistics.

[0119] In this embodiment, a fuzzy comprehensive evaluation method is adopted for fault determination. A judgment index system including the equipment operation status, component wear status, and safety threshold status is established. The weights of each index are determined by the analytic hierarchy process, considering the correlation and importance degree among the indexes. The evaluation results are divided into three levels: normal, slightly abnormal, and serious fault, and different levels trigger different protection responses.

[0120] In this embodiment, a redundant design is adopted in the safety clamp control system. The controller adopts a dual-CPU architecture. The main control CPU is responsible for normal control logic, and the monitoring CPU is responsible for safety protection functions. The two CPUs exchange data through independent communication buses and monitor each other's working status. When a fault is detected, both CPUs need to confirm simultaneously to trigger an emergency brake to avoid misoperation.

[0121] In this embodiment, adaptive braking force adjustment is realized in braking control. The required braking torque is calculated according to the current running speed and load weight, and the braking pressure is accurately controlled through a proportional solenoid valve. The braking process is divided into two stages: rapid clamping and pressure holding to ensure a smooth and controllable braking process. At the same time, the temperature of the brake is monitored to prevent the brake from failing due to excessive braking.

[0122] In this embodiment, a multi-layer security protection mechanism is adopted in the Internet of Things communication module. Data encryption uses the AES-256 algorithm, and the key is securely exchanged through an asymmetric encryption method. The communication protocol adopts an improved MQTT protocol, supporting message classification and priority transmission. At the network layer, VPN technology is used to establish a secure channel to prevent data from being illegally intercepted or tampered with.

[0123] In this embodiment, an adaptive data compression algorithm is designed. Appropriate compression methods are selected according to the characteristics of different types of parameters. For example, differential coding is used for continuously changing parameters, and run-length coding is used for discrete states. The compression algorithm significantly reduces the amount of transmitted data and improves communication efficiency on the premise of ensuring data accuracy.

[0124] In this embodiment, an automatic reconnection and link backup mechanism is realized during the establishment of the communication link. The primary communication link uses a 4G network, and WiFi is equipped as a backup link at the same time. When the quality of the primary link is detected to decline, it automatically switches to the backup link. Local data caching is enabled during the link switching process to ensure that alarm data is not lost.

[0125] In this embodiment, a modular design is adopted in the construction of the alarm data packet. The data packet includes three parts: basic equipment information, fault status information, and system configuration information. The equipment information includes fixed information such as equipment number, model, installation location, etc.; the fault information contains the parameters, thresholds, and timestamps that trigger the fault; the system information contains the current software version and configuration parameters.

[0126] This embodiment implements an alarm priority management mechanism. Different priorities are assigned to alarm data packets according to the severity of the faults. High-priority packets can interrupt the current transmission and be preferentially sent to the supervision system. After receiving the alarm, the supervision system automatically notifies the relevant personnel according to the preset rules and initiates the emergency response process.

[0127] This embodiment establishes a complete traceability mechanism in the processing of fault data. Detailed status data is recorded for each fault trigger, including the operating parameters and operation records before the fault occurs. These data are stored in the local database and uploaded to the cloud server simultaneously for subsequent fault analysis and preventive maintenance. This solution enables the timely discovery and rapid response to construction lift faults, improving the safety and reliability of equipment operation.

[0128] In order to break through the limitations of the traditional lift management mode and provide an intelligent overall solution for construction safety management, this application provides an embodiment of a multi-sensor-based construction lift safety control device for implementing all or part of the content of the multi-sensor-based construction lift safety control method. See Figure 8 , the multi-sensor-based construction lift safety control device specifically includes the following: A biometric detection module 10, configured to collect the facial images of construction workers through a binocular live detection camera and the voiceprint information through a voiceprint collector, input the facial images into a personnel verification model for identity feature extraction, input the voiceprint information into a voiceprint verification model for voiceprint feature extraction, perform dual verification on the identity features and the voiceprint features respectively with the preset information of the construction workers based on a deep neural network, and simultaneously collect the images of the wearing status of the construction workers' protective equipment. Input the images of the wearing status of the protective equipment into a posture recognition model, generate construction worker permission information according to the identity verification result, the voiceprint verification result, and the posture recognition result, and input the permission information into the elevator control authorization unit; A status monitoring module 20, configured to deploy weight sensors, position sensors, speed sensors, vibration sensors, limit sensors, door switch sensors, and safety clamp sensors at the bottom of the car, the side wall of the guide rail, and the edge of the door frame of the construction lift, collect the load data, position data, speed data, vibration data, travel data, door status data, and braking data of the construction lift, input the collected multi-dimensional data into a data preprocessing unit for noise reduction filtering and standardization processing, input the preprocessed data into an equipment status evaluation model according to a preset weight, and generate construction lift operation status information according to the status parameters output by the equipment status evaluation model; The safety control module 30 is used to display the operation status information on the monitoring touch screen, and at the same time input the operation status information and the construction personnel permission information into the intelligent control unit. The intelligent control unit generates an operation control instruction based on the load data, position data and speed data of the construction lift, adjusts the braking distance and leveling accuracy of the construction lift according to the operation control instruction, and triggers a fault protection mechanism when the operation status information deviates from the preset threshold. The fault protection mechanism controls the safety clamp sensor to perform braking and sends an alarm signal to the safety supervision system through the Internet of Things communication module.

[0129] As can be seen from the above description, the safety control device for construction lifts based on multi-modal perception provided by the embodiments of the present application can accurately verify the personnel identity and the status of protective equipment through multi-modal identity recognition of binocular live detection and voiceprint collection, and combines a deep neural network. An equipment status evaluation model based on multi-sensor fusion is innovatively designed to realize the collaborative monitoring of key parameters such as load, position, and speed. The system adopts an intelligent control strategy to adaptively adjust operation parameters, establishes a real-time fault protection mechanism based on threshold judgment, and realizes remote supervision through Internet of Things technology. This method breaks through the limitations of the traditional lift management mode and provides an intelligent overall solution for construction safety management.

[0130] From the hardware level, in order to break through the limitations of the traditional lift management mode and provide an intelligent overall solution for construction safety management, the present application provides an embodiment of an electronic device for implementing all or part of the content in the safety control method for construction lifts based on multi-modal perception. The electronic device specifically includes the following contents: A processor, a memory, a communication interface, and a bus; wherein, the processor, the memory, and the communication interface complete communication with each other through the bus; the communication interface is used to realize information transmission between the safety control device for construction lifts based on multi-modal perception and related devices such as the core business system, the user terminal, and the relevant database. The logic controller can be a desktop computer, a tablet computer, a mobile terminal, etc., and this embodiment is not limited thereto. In this embodiment, the logic controller can be implemented with reference to the embodiments of the safety control method for construction lifts based on multi-modal perception and the embodiments of the safety control device for construction lifts based on multi-modal perception, and the content is incorporated herein, and the repeated parts will not be described again.

[0131] It can be understood that the user terminal may include a smart phone, a tablet electronic device, a network set-top box, a portable computer, a desktop computer, a personal digital assistant (PDA), a vehicle-mounted device, a smart wearable device, etc. Among them, the smart wearable device may include smart glasses, smart watches, smart bracelets, etc.

[0132] In practical applications, part of the construction elevator safety control method based on multi-sensor perception can be executed on the electronic device side as described above, or all operations can be completed in the client device. Specifically, it can be selected according to the processing capacity of the client device and the limitations of the user usage scenario, etc. This application does not make a limitation in this regard. If all operations are completed in the client device, the client device may further include a processor.

[0133] The above-mentioned client device may have a communication module (i.e., a communication unit), and can be communicatively connected to a remote server to realize data transmission with the server. The server may include a server on the task scheduling center side, and in other implementation scenarios, it may also include a server of an intermediate platform, such as a server of a third-party server platform communicatively linked to the task scheduling center server. The server may include a single computer device, or may include a server cluster composed of multiple servers, or a server structure of a distributed device.

[0134] Figure 9 It is a schematic block diagram of the system composition of the electronic device 9600 according to an embodiment of the present application. As Figure 9 shown, the electronic device 9600 may include a central processing unit 9100 and a memory 9140; the memory 9140 is coupled to the central processing unit 9100. It should be noted that this Figure 9 is exemplary; other types of structures may also be used to supplement or replace this structure to implement telecommunication functions or other functions.

[0135] In one embodiment, the function of the construction elevator safety control method based on multi-sensor perception can be integrated into the central processing unit 9100. Among them, the central processing unit 9100 may be configured to perform the following controls: Step S101: Collect the facial images of construction workers through a binocular live detection camera and the voiceprint information through a voiceprint collector. Input the facial images into a personnel verification model for identity feature extraction, and input the voiceprint information into a voiceprint verification model for voiceprint feature extraction. Based on a deep neural network, perform double verification on the identity features and the voiceprint features respectively with the preset information of the construction workers. At the same time, collect the images of the wearing status of the construction workers' protective equipment, input the images of the wearing status of the protective equipment into a posture recognition model, generate the access permission information of the construction workers according to the identity verification result, the voiceprint verification result and the posture recognition result, and input the access permission information into the elevator control authorization unit; Step S102: Arrange weight sensors, position sensors, speed sensors, vibration sensors, limit sensors, door switch sensors and safety clamp sensors at the bottom of the car, on the side wall of the guide rail and at the edge of the door frame of the construction elevator. Collect the load data, position data, speed data, vibration data, travel data, door status data and braking data of the construction elevator. Input the collected multi-dimensional data into a data preprocessing unit for noise reduction filtering and standardization processing. Input the preprocessed data into an equipment status evaluation model according to preset weights, and generate the operation status information of the construction elevator according to the status parameters output by the equipment status evaluation model; Step S103: Display the operation status information on the monitoring touch screen. At the same time, input the operation status information and the access permission information of the construction workers into an intelligent control unit. The intelligent control unit generates an operation control instruction based on the load data, position data and speed data of the construction elevator, adjusts the braking distance and leveling accuracy of the construction elevator according to the operation control instruction. When the operation status information deviates from the preset threshold, trigger a fault protection mechanism, control the safety clamp sensor to perform braking through the fault protection mechanism and send an alarm signal to the safety supervision system through the Internet of Things communication module.

[0136] As can be seen from the above description, the electronic device provided by the embodiment of the present application realizes the accurate verification of personnel identity and the status of protective equipment through multi-modal identity recognition of binocular live detection and voiceprint collection, and combines a deep neural network. An equipment status evaluation model based on multi-sensor fusion is innovatively designed to realize the collaborative monitoring of key parameters such as load, position and speed. The system adopts an intelligent control strategy to adaptively adjust operation parameters, establishes a real-time fault protection mechanism based on threshold judgment, and realizes remote supervision through Internet of Things technology. This method breaks through the limitations of the traditional elevator management mode and provides an intelligent overall solution for construction safety management.

[0137] In another embodiment, the safety control device for construction elevators based on multi-sensor perception can be separately configured from the central processor 9100. For example, the safety control device for construction elevators based on multi-sensor perception can be configured as a chip connected to the central processor 9100, and the functions of the safety control method for construction elevators based on multi-sensor perception can be realized through the control of the central processor.

[0138] As Figure 9 shown, the electronic device 9600 may further include: a communication module 9110, an input unit 9120, an audio processor 9130, a display 9160, and a power supply 9170. It should be noted that the electronic device 9600 does not necessarily have to include Figure 9 all the components shown in Figure 9 ; in addition, the electronic device 9600 may further include

[0139] As Figure 9 shown, the central processor 9100 is sometimes also referred to as a controller or an operation control, and may include a microprocessor or other processor devices and / or logic devices. The central processor 9100 receives inputs and controls the operations of the various components of the electronic device 9600.

[0140] Among them, the memory 9140 may be, for example, one or more of a buffer, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory, or other suitable devices. The above information related to failures can be stored, and in addition, programs for executing relevant information can also be stored. And the central processor 9100 can execute the programs stored in the memory 9140 to implement information storage or processing, etc.

[0141] The input unit 9120 provides inputs to the central processor 9100. The input unit 9120 is, for example, a key or a touch input device. The power supply 9170 is used to supply power to the electronic device 9600. The display 9160 is used to display display objects such as images and texts. The display may be, for example, an LCD display, but is not limited thereto.

[0142] The memory 9140 can be a solid-state memory, for example, a read-only memory (ROM), a random access memory (RAM), a SIM card, etc. It can also be a memory that stores information even when the power is off, can be selectively erased and has more data. Examples of such a memory are sometimes referred to as EPROMs, etc. The memory 9140 can also be some other type of device. The memory 9140 includes a buffer memory 9141 (sometimes referred to as a buffer). The memory 9140 can include an application / function storage unit 9142, which is used to store application programs and function programs or the processes for operating the electronic device 9600 by the central processing unit 9100.

[0143] The memory 9140 can also include a data storage unit 9143, which is used to store data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 9144 of the memory 9140 can include various drivers of the electronic device for communication functions and / or for performing other functions of the electronic device (such as a messaging application, an address book application, etc.).

[0144] The communication module 9110 is a transmitter / receiver that transmits and receives signals via the antenna 9111. The communication module 9110 (transmitter / receiver) is coupled to the central processing unit 9100 to provide input signals and receive output signals, which can be the same as in the case of a conventional mobile communication terminal.

[0145] Based on different communication technologies, multiple communication modules 9110 can be provided in the same electronic device, such as a cellular network module, a Bluetooth module, and / or a wireless local area network module, etc. The communication module 9110 (transmitter / receiver) is also coupled to the speaker 9131 and the microphone 9132 via the audio processor 9130 to provide an audio output via the speaker 9131 and receive an audio input from the microphone 9132, so as to achieve normal telecommunication functions. The audio processor 9130 can include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 9130 is also coupled to the central processing unit 9100, so that it is possible to record on the device via the microphone 9132 and play the sounds stored on the device via the speaker 9131.

[0146] An embodiment of the present application also provides a computer-readable storage medium that can implement all steps of the multi-sensor-based construction elevator safety control method where the execution entity in the above embodiment is a server or a client. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, it implements all steps of the multi-sensor-based construction elevator safety control method where the execution entity in the above embodiment is a server or a client. For example, when the processor executes the computer program, the following steps are implemented: Step S101: Collect the facial image of the construction worker through a binocular live detection camera and the voiceprint information through a voiceprint collector. Input the facial image into a personnel verification model for identity feature extraction, and input the voiceprint information into a voiceprint verification model for voiceprint feature extraction. Based on a deep neural network, perform double verification on the identity feature and the voiceprint feature with the preset information of the construction worker. At the same time, collect the image of the wearing state of the construction worker's protective equipment, input the image of the wearing state of the protective equipment into a pose recognition model, generate the permission information of the construction worker according to the identity verification result, the voiceprint verification result, and the pose recognition result, and input the permission information into the elevator control authorization unit; Step S102: Install weight sensors, position sensors, speed sensors, vibration sensors, limit sensors, door switch sensors, and safety clamp sensors at the bottom of the car, the side wall of the guide rail, and the edge of the door frame of the construction elevator. Collect the load data, position data, speed data, vibration data, travel data, door state data, and braking data of the construction elevator. Input the collected multi-dimensional data into a data preprocessing unit for noise reduction filtering and standardization processing. Input the preprocessed data into an equipment state evaluation model according to the preset weights, and generate the operation state information of the construction elevator according to the state parameters output by the equipment state evaluation model; Step S103: Display the operation state information on a monitoring touch screen. At the same time, input the operation state information and the permission information of the construction worker into an intelligent control unit. The intelligent control unit generates an operation control instruction based on the load data, position data, and speed data of the construction elevator, adjusts the braking distance and leveling accuracy of the construction elevator according to the operation control instruction. When the operation state information deviates from the preset threshold, trigger a fault protection mechanism, control the safety clamp sensor to perform braking through the fault protection mechanism, and send an alarm signal to the safety supervision system through the Internet of Things communication module.

[0147] As can be seen from the above description, the computer-readable storage medium provided by the embodiments of the present application achieves precise verification of personnel identity and the status of protective equipment through multi-modal identity recognition of binocular live detection and voiceprint acquisition, and combines a deep neural network. An equipment status evaluation model based on multi-sensor fusion is innovatively designed to achieve collaborative monitoring of key parameters such as load, position, and speed. The system uses an intelligent control strategy to adaptively adjust operating parameters, establishes a real-time fault protection mechanism based on threshold judgment, and realizes remote supervision through Internet of Things technology. This method breaks through the limitations of the traditional elevator management mode and provides an intelligent overall solution for construction safety management.

[0148] An embodiment of the present application also provides a computer program product capable of implementing all steps of the multi-sensor-based construction elevator safety control method with the execution subject being a server or a client in the above embodiments. When the computer program / instructions are executed by a processor, the steps of the multi-sensor-based construction elevator safety control method are implemented. For example, the computer program / instructions implement the following steps: Step S101: Collect the facial image of the construction worker through a binocular live detection camera and the voiceprint information through a voiceprint collector. Input the facial image into a personnel verification model for identity feature extraction, input the voiceprint information into a voiceprint verification model for voiceprint feature extraction, and perform double verification on the identity feature and the voiceprint feature respectively with the preset information of the construction worker based on a deep neural network. At the same time, collect the image of the wearing state of the construction worker's protective equipment, input the image of the wearing state of the protective equipment into a pose recognition model, generate the permission information of the construction worker according to the identity verification result, the voiceprint verification result, and the pose recognition result, and input the permission information into the elevator control authorization unit; Step S102: Arrange weight sensors, position sensors, speed sensors, vibration sensors, limit sensors, door switch sensors, and safety clamp sensors at the bottom of the car, the side wall of the guide rail, and the edge of the door frame of the construction elevator. Collect the load data, position data, speed data, vibration data, travel data, door state data, and braking data of the construction elevator. Input the collected multi-dimensional data into a data preprocessing unit for noise reduction filtering and standardization processing. Input the preprocessed data into an equipment state evaluation model according to preset weights, and generate the operation state information of the construction elevator according to the state parameters output by the equipment state evaluation model; Step S103: Display the operation status information on the monitoring touch screen. At the same time, input the operation status information and the construction worker permission information into the intelligent control unit. The intelligent control unit generates an operation control instruction based on the load data, position data, and speed data of the construction lift, adjusts the braking distance and leveling accuracy of the construction lift according to the operation control instruction, triggers a fault protection mechanism when the operation status information deviates from a preset threshold, controls the safety clamp sensor to execute braking through the fault protection mechanism, and sends an alarm signal to the safety supervision system through the Internet of Things communication module.

[0149] As can be seen from the above description, the computer program product provided by the embodiments of the present application realizes the accurate verification of personnel identity and the status of protective equipment through multi-modal identity recognition of binocular live detection and voiceprint collection, and combines a deep neural network. An equipment status evaluation model based on multi-sensor fusion is innovatively designed to realize the collaborative monitoring of key parameters such as load, position, and speed. The system adopts an intelligent control strategy to adaptively adjust operation parameters, establishes a real-time fault protection mechanism based on threshold judgment, and realizes remote supervision through Internet of Things technology. This method breaks through the limitations of the traditional lift management mode and provides an intelligent overall solution for construction safety management.

[0150] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, apparatus, or computer program product. Therefore, the present invention can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0151] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (apparatus), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for realizing the functions specified in one Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0152] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction device that implements the functions specified in one or more of the processes and / or blocks Figure 1 in one or more of the processes and / or blocks Figure 1 specified in the block or blocks.

[0153] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one or more of the processes and / or blocks Figure 1 in one or more of the processes and / or blocks Figure 1 specified in the block or blocks.

[0154] Specific embodiments are used in the present invention to illustrate the principles and implementation manners of the present invention. The descriptions of the above embodiments are only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. A construction elevator safety management and control method based on multi-sensing, characterized in that: The method comprises: The facial image of the construction personnel is collected by a binocular liveness detection camera and the voiceprint information is collected by a voiceprint collector, the facial image is input into a personnel verification model to extract identity features, the voiceprint information is input into a voiceprint verification model to extract voiceprint features, the identity features and the voiceprint features are double-verified with the preset information of the construction personnel based on a deep neural network, and at the same time, an image of the wearing state of the protective equipment of the construction personnel is collected, the image of the wearing state of the protective equipment is input into a posture recognition model, and the authority information of the construction personnel is generated according to the identity authentication result, the voiceprint verification result and the posture recognition result, and the authority information is input into the elevator control authorization unit; Weight sensors, position sensors, speed sensors, vibration sensors, limit sensors, door switch sensors and safety clamp sensors are arranged at the bottom of the car, the side walls of the guide rails and the edges of the door frames of the construction elevator to collect the load data, position data, speed data, vibration data, travel data, door status data and braking data of the construction elevator, and the collected multi-dimensional data is input into the data preprocessing unit for noise reduction filtering and standardization processing, and the preprocessed data is input into the equipment status evaluation model according to the preset weights, and the operation status information of the construction elevator is generated according to the status parameters output by the equipment status evaluation model; The operation status information is displayed on the monitoring touch screen, and the operation status information and the construction personnel authority information are input into the intelligent control unit. The intelligent control unit generates an operation control instruction based on the load data, position data and speed data of the construction elevator, and adjusts the braking distance and leveling accuracy of the construction elevator according to the operation control instruction. When the operation status information deviates from the preset threshold, a fault protection mechanism is triggered. The safety clamp sensor is controlled by the fault protection mechanism to perform braking and send an alarm signal to the safety supervision system through the Internet of Things communication module.

2. The construction elevator safety management and control method based on multi-sensing according to claim 1 is characterized in that: The method collects the facial image of the construction personnel through the binocular liveness detection camera and the voiceprint information through the voiceprint collector, inputs the facial image into the personnel verification model to extract the identity feature, inputs the voiceprint information into the voiceprint verification model to extract the voiceprint feature, and performs double verification on the identity feature and the voiceprint feature with the preset information of the construction personnel based on the deep neural network, including: Performing image enhancement preprocessing on the facial image collected by the binocular liveness detection camera, inputting the preprocessed facial image into a convolutional neural network to detect key points of the face, extracting the coordinates of facial feature points including the contours of eyes, nose and mouth, constructing a facial geometric feature vector based on the facial feature point coordinates, inputting the facial geometric feature vector into a personnel verification model to generate an identity feature code, and encrypting and storing the identity feature code; Continuous speech segments of construction workers are collected through a voiceprint collector, and the speech segments are framed and windowed, and Mel-frequency cepstral coefficient features are extracted from the speech segments. The extracted cepstral coefficient features are input into a voiceprint verification model to generate a voiceprint feature vector, and the identity feature code and the voiceprint feature vector are respectively introduced into a deep neural network. The deep neural network calculates a feature similarity score with a preset identity information library, and the similarity score is compared with a preset matching threshold to generate a verification result.

3. The construction elevator safety management and control method based on multi-sensing according to claim 1 is characterized in that: The collecting of the protective equipment wearing state image of the construction personnel, inputting the protective equipment wearing state image into the posture recognition model, generating the construction personnel authority information according to the identity authentication result, the voiceprint verification result and the posture recognition result, and inputting the authority information into the elevator control authorization unit, comprises: Collect front and side image sequences of construction workers, perform target detection and image segmentation on the image sequences, extract contour features of protective equipment including helmets, reflective clothing and safety belts, input the contour features into a posture recognition model based on key points of the human skeleton, calculate the spatial position relationship between the protective equipment and the key points of the human skeleton through the posture recognition model, and generate a characteristic map of the wearing state of the protective equipment; The protective equipment wearing status characteristic diagram, the identity authentication result, and the voiceprint verification result are input into the authority assessment model, and the identity level of the construction personnel and the compliance with the protection specifications are comprehensively analyzed through the authority assessment model. According to the analysis results, authority information including the operation authority level and the operation time limit is generated, and the authority information is written into the data register of the elevator control authorization unit to establish a mapping relationship between the construction personnel identity and the operation authority.

4. The construction elevator safety management and control method based on multi-sensing according to claim 1 is characterized in that: The weight sensor, position sensor, speed sensor, vibration sensor, limit sensor, door switch sensor and safety clamp sensor are arranged at the bottom of the car, the side wall of the guide rail and the edge of the door frame of the construction elevator to collect the load data, position data, speed data, vibration data, travel data, door status data and brake data of the construction elevator, and the collected multi-dimensional data is input into the data preprocessing unit for noise reduction filtering and standardization processing, including: A weight sensor is installed at the bottom of the construction elevator car, a position sensor, a speed sensor and a vibration sensor are arranged on the side wall of the guide rail, and a limit sensor, a door switch sensor and a safety clamp sensor are set on the edge of the door frame. The weight sensor is used to collect load data, the position sensor is used to collect position data, the speed sensor is used to collect speed data, the vibration sensor is used to collect vibration data, the limit sensor is used to collect travel data, the door switch sensor is used to collect door status data, and the safety clamp sensor is used to collect brake data; The collected multidimensional data is input into a data preprocessing unit, the multidimensional data is subjected to noise reduction processing using wavelet transform, high-frequency interference signals are filtered out using a Butterworth low-pass filter, the filtered data is subjected to zero-mean normalization processing, the normalized data is time-series aligned according to the sampling timestamps, and a preprocessed data matrix is ​​generated.

5. The construction elevator safety management and control method based on multi-sensing according to claim 1 is characterized in that: The pre-processed data is input into the equipment status evaluation model according to the preset weights, and the operation status information of the construction elevator is generated according to the status parameters output by the equipment status evaluation model, including: Constructing a preset weight matrix of the equipment status assessment model, allocating weight coefficients according to the degree of influence of each sensor data on the operating status of the construction elevator, performing matrix multiplication operation on the preprocessed data matrix and the preset weight matrix to generate a weighted feature vector, extracting time series features from the weighted feature vector through a recursive neural network, and inputting the extracted time series features into the equipment status assessment model; Based on the load parameters, running speed parameters, position parameters, vibration parameters, door state parameters and braking parameters output by the equipment state assessment model, the parameters are combined to construct a state feature space, and the state feature space is nonlinearly mapped through a deep neural network to generate operating state information including equipment operating state, component wear state and safety threshold state.

6. The construction elevator safety management and control method based on multi-sensing according to claim 1 is characterized in that: The operation status information is displayed on the monitoring touch screen, and the operation status information and the construction personnel authority information are input into the intelligent control unit. The intelligent control unit generates an operation control instruction based on the load data, position data and speed data of the construction elevator, and adjusts the braking distance and leveling accuracy of the construction elevator according to the operation control instruction, including: Constructing a construction elevator operation status display interface on the monitoring touch screen, converting the equipment operation status, component wear status and safety threshold status in the operation status information into graphical display data through a data visualization module, and importing the operation status information and the operation authority level and operation time limit in the construction personnel authority information into the data cache area of ​​the intelligent control unit; The load data, the position data and the speed data are input into the data analysis module of the intelligent control unit, the data are analyzed in real time through a fuzzy neural network, a speed control model based on the load and displacement is established, the calculation result of the speed control model is compared with the rated parameters of the construction elevator, and an operation control instruction including a braking torque and a leveling compensation value is generated.

7. The construction elevator safety management and control method based on multi-sensing according to claim 1 is characterized in that: When the operating status information deviates from a preset threshold, a fault protection mechanism is triggered, and the safety clamp sensor is controlled by the fault protection mechanism to perform braking and an alarm signal is sent to the safety supervision system through the Internet of Things communication module, including: Real-time monitoring of the equipment operation status, component wear status and safety threshold status in the operation status information, comparing the monitoring data with the preset fault judgment threshold, and when any state parameter exceeds the preset threshold, starting the fault protection mechanism, sending a brake control instruction to the safety clamp sensor, and driving the brake to implement emergency braking through the safety clamp sensor; The deviation value between the state parameter that triggers the fault protection mechanism and the preset threshold is input into the Internet of Things communication module, the deviation data is encrypted and encapsulated through the Internet of Things communication module, a communication link with the security supervision system is established, the encrypted deviation data and the equipment number information are combined into an alarm data packet, and the alarm data packet is sent to the security supervision system through the communication link.

8. A construction elevator safety control device based on multi-sensing, characterized in that: The device comprises: A biological detection module is used to collect facial images of construction personnel through a binocular liveness detection camera and voiceprint information through a voiceprint collector, input the facial images into a personnel verification model to extract identity features, input the voiceprint information into a voiceprint verification model to extract voiceprint features, and perform double verification on the identity features and the voiceprint features with the preset information of the construction personnel based on a deep neural network, and at the same time collect images of the wearing status of the construction personnel's protective equipment, input the images of the wearing status of the protective equipment into a posture recognition model, generate construction personnel authority information according to the identity authentication result, the voiceprint verification result, and the posture recognition result, and input the authority information into an elevator control authorization unit; A state monitoring module is used to arrange weight sensors, position sensors, speed sensors, vibration sensors, limit sensors, door switch sensors and safety clamp sensors at the bottom of the car, the side walls of the guide rails and the edges of the door frames of the construction elevator, collect the load data, position data, speed data, vibration data, travel data, door state data and brake data of the construction elevator, input the collected multi-dimensional data into the data preprocessing unit for noise reduction filtering and standardization processing, input the preprocessed data into the equipment state evaluation model according to the preset weights, and generate the operation state information of the construction elevator according to the state parameters output by the equipment state evaluation model; The safety management and control module is used to display the operating status information on the monitoring touch screen, and at the same time input the operating status information and the construction personnel authority information into the intelligent control unit. The intelligent control unit generates an operating control instruction based on the load data, position data and speed data of the construction elevator, and adjusts the braking distance and leveling accuracy of the construction elevator according to the operating control instruction. When the operating status information deviates from the preset threshold, the fault protection mechanism is triggered. The safety clamp sensor is controlled by the fault protection mechanism to perform braking and send an alarm signal to the safety supervision system through the Internet of Things communication module.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the construction elevator safety management and control method based on multi-sensing described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the construction elevator safety management and control method based on multi-sensing described in any one of claims 1 to 7 are implemented.

Images