Pumped storage power station safety construction monitoring method based on image recognition

Abstract

The invention discloses a pumped storage power station safety construction monitoring method based on image recognition, and relates to the technical field of pumped storage power station safety construction monitoring, and the method comprises the steps: carrying out the intelligent monitoring of a construction site through a high-definition camera, an unmanned plane, and a multispectral imaging device, and optimizing the target capturing capability through an environment adaptive adjustment strategy. And in combination with a deep learning and behavior analysis model, identifying construction personnel, equipment and environment states, detecting behaviors such as no safety helmet wearing, illegal retention, equipment abnormity and the like, and performing real-time tracking and updating. Based on a risk level calculation model, a multi-level alarm mechanism is constructed, acousto-optic early warning, remote alarm and task scheduling are triggered, and construction safety is optimized in combination with cloud collaborative management. And self-supervised learning and cross-scene transfer learning are adopted, so that the recognition precision and the system adaptability in a complex environment are improved. According to the method, better effects are achieved in the aspects of real-time performance and accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of safety construction monitoring of pumped storage power stations, and specifically to a method for safety construction monitoring of pumped storage power stations based on image recognition. Background Art

[0002] With the rapid development of intelligent technologies, the applications of computer vision, artificial intelligence (AI), and deep learning in the field of industrial safety monitoring have become increasingly widespread. In high-risk working environments such as construction and energy facility construction, traditional safety monitoring methods mainly rely on manual inspections and fixed monitoring devices. However, with the expansion of project scale, the complexity of construction environments, and the increasing demand for refined safety management, traditional monitoring means are difficult to meet the requirements of efficient, accurate, and safe construction supervision. In recent years, the development of computer vision and deep learning technologies has promoted the application of intelligent monitoring methods based on image recognition. The combination of drones, multispectral imaging, edge computing, and cloud storage enables construction safety monitoring to have a higher level of automation. Especially during the construction process of pumped storage power stations, which involve high-altitude operations, large machinery operation, and construction in complex terrains, accurate monitoring of construction workers, equipment, and the environment, as well as risk early warning, have become key requirements for ensuring construction safety. Intelligent safety monitoring technologies based on image recognition have become the focus of current research and are gradually applied to construction safety management systems.

[0003] Currently, during the construction process of major projects such as pumped storage power stations, safety monitoring mainly relies on manual inspections, fixed camera monitoring, and sensor monitoring. Manual inspections rely on the experience and responsibility of safety officers, are easily affected by subjective factors, and cannot provide real-time and full-coverage monitoring. Although fixed camera monitoring can provide partial monitoring coverage, it is limited by the fixed installation position of the cameras and is difficult to dynamically adapt to changes in the construction environment, with a limited monitoring range. In addition, single-sensor monitoring methods can only collect environmental parameters (such as temperature, humidity, wind speed, etc.), but cannot perform intelligent analysis and cannot comprehensively evaluate the behavior of construction workers and the operating status of equipment.

[0004] In the existing applications of image recognition and deep learning technologies, there are also several deficiencies: the existing methods have poor adaptability to factors such as lighting changes, occlusion, and dust interference in complex construction environments, resulting in reduced recognition accuracy; most existing construction safety monitoring methods are based on single-modal data (such as a single RGB image or infrared image), and do not fully utilize the advantages of multi-modal data fusion (such as visible light + infrared + depth imaging), which limits the accuracy of target detection; the existing systems are relatively simple in target tracking and behavior analysis, and cannot accurately identify violations of construction workers, such as not wearing a safety helmet and staying in a dangerous area for a long time; the risk assessment mechanism is not perfect, most are based on simple rule matching or threshold judgment, lacking a data-driven intelligent risk prediction model, and it is difficult to give early warnings; there is a lack of an effective cloud collaborative management and remote scheduling mechanism, resulting in imperfect functions such as data storage, permission management, and remote supervision, and it is difficult to support the intelligent management of large-scale construction sites. Summary of the Invention

[0005] In view of the above problems, the present invention is proposed.

[0006] Therefore, the technical problems solved by the present invention are: the existing construction safety monitoring methods have the problems of limited monitoring scope and difficulty in covering a dynamic construction environment in real time, the target recognition accuracy is affected by complex factors at the construction site, and it is difficult to accurately judge the status of construction workers and equipment, the risk assessment mechanism is imperfect, and it is impossible to achieve data-driven accurate early warning, and how to realize the automatic and intelligent monitoring of construction site safety through means such as multi-modal data fusion, intelligent behavior analysis, and cloud collaborative management.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: a safety construction monitoring method for a pumped storage power station based on image recognition, including using high-definition cameras, drones, and multi-spectral imaging equipment to conduct all-round monitoring of the construction site, and combining an environment adaptive adjustment strategy to optimize the target capture ability in complex environments; Adopting a deep learning neural network and dynamic feature extraction technology, combined with a behavior analysis model, to identify the status of construction workers, equipment, and the environment, detect abnormal behaviors, including not wearing a safety helmet, personnel staying in a dangerous area for a long time, and abnormal operation of mechanical equipment, and conduct target tracking and real-time update; Based on a risk level calculation model, a multi-level alarm mechanism is constructed to trigger local audible and visual warnings, remote management alarms, and automatic task scheduling, handle them at different levels, and give prediction results for cloud collaborative management and remote scheduling, and adopt self-supervised learning and cross-scene transfer learning to optimize the model.

[0008] As a preferred solution of the safety construction monitoring method for pumped-storage power stations based on image recognition according to the present invention, wherein: the environmental adaptive adjustment strategy includes dynamically adjusting the exposure parameters, noise suppression, contrast enhancement and image compensation of the camera according to the factors of illumination, weather, air quality and mechanical vibration in the construction area.

[0009] As a preferred solution of the safety construction monitoring method for pumped-storage power stations based on image recognition according to the present invention, wherein: the deep learning neural network adopts multi-modal data input including construction worker pose estimation, equipment operation state analysis and environmental sensor data, and combines object classification, behavior recognition and time series modeling.

[0010] As a preferred solution of the safety construction monitoring method for pumped-storage power stations based on image recognition according to the present invention, wherein: the hierarchical processing includes setting the alarm level based on the risk level calculation model, triggering local audible and visual warnings for low-level alarms, pushing medium-level alarms to the management platform, and triggering remote task scheduling and emergency response for high-level alarms.

[0011] As a preferred solution of the safety construction monitoring method for pumped-storage power stations based on image recognition according to the present invention, wherein: the cloud collaborative management includes distributed storage and access permission control, encrypting and storing the construction monitoring data, and allocating construction tasks and equipment resources based on the remote scheduling strategy.

[0012] As a preferred solution of the safety construction monitoring method for pumped-storage power stations based on image recognition according to the present invention, wherein: the self-supervised learning includes unsupervised feature alignment and transfer learning, using few-shot learning for object detection, and adjusting the neural network parameters.

[0013] As a preferred solution of the safety construction monitoring method for pumped-storage power stations based on image recognition according to the present invention, wherein: the remote scheduling includes the system adopting a scheduling optimization strategy based on construction environment parameters, personnel distribution and equipment status, allocating construction tasks, and adjusting the scheduling plan in combination with real-time monitoring data.

[0014] Another object of the present invention is to provide a safety construction monitoring system for pumped-storage power stations based on image recognition, which can solve the problems of limited monitoring range and influence of illumination change on recognition accuracy in the current construction safety monitoring technology through an intelligent object detection method based on multi-modal data fusion.

[0015] As a preferred solution of the pumped storage power station safety construction monitoring system based on image recognition described in the present invention, it includes: a target capture module, a behavior recognition and analysis module, and a risk assessment and early warning module; the target capture module is used to use high-definition cameras, drones and multi-spectral imaging equipment to conduct all-round monitoring of the construction site, and combine the environmental adaptive adjustment strategy to optimize the target capture capability in a complex environment; the behavior recognition and analysis module is used to use deep learning neural networks and dynamic feature extraction technology, combined with behavior analysis models, to identify construction personnel, equipment and environmental status, detect abnormal behaviors, including not wearing safety helmets, personnel staying in dangerous areas for a long time, and abnormal operation of mechanical equipment, and perform target tracking and real-time updates; the risk assessment and early warning module is used to build a multi-level alarm mechanism based on a risk level calculation model, trigger local sound and light warnings, remote management alarms and automatic task scheduling, hierarchical processing, and give prediction results for cloud-based collaborative management and remote scheduling, and use self-supervised learning and cross-scenario transfer learning to optimize the model.

[0016] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a step of a method for monitoring the safe construction of a pumped storage power station based on image recognition.

[0017] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for monitoring the safe construction of a pumped-storage power station based on image recognition.

[0018] Beneficial effects of the present invention: The image recognition-based pumped storage power station safety construction monitoring method provided by the present invention adopts deep learning target detection, combined with posture estimation and time series behavior analysis model, which can accurately identify the construction personnel wearing safety equipment and illegal operation behaviors, and combined with target tracking technology, real-time monitoring of the dynamic status of construction site personnel and equipment, and construction. A multi-dimensional risk assessment model is constructed, and the risk index of the construction site is calculated by comprehensively considering the construction personnel, equipment operation, and environmental factors. A three-level response mechanism of local sound and light warning, remote management alarm, and automatic task scheduling is constructed to achieve data-driven accurate safety warning. The present invention achieves better results in terms of real-time and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0020] Figure 1This is the overall flowchart of a safety construction monitoring method for pumped storage power stations based on image recognition provided by the first embodiment of the present invention. Detailed implementation manners

[0021] To make the above objects, features, and advantages of the present invention more apparent and understandable, the following will describe in detail the specific implementation manners of the present invention with reference to the accompanying drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0022] Example 1, referring to Figure 1 , which is an embodiment of the present invention, provides a safety construction monitoring method for pumped storage power stations based on image recognition, including: S1: Use high-definition cameras, unmanned aerial vehicles (UAVs), and multispectral imaging devices to conduct comprehensive monitoring of the construction site, and combine an environment adaptive adjustment strategy to optimize the target capture ability in complex environments.

[0023] Furthermore, the environment adaptive adjustment strategy includes dynamically adjusting the exposure parameters, noise suppression, contrast enhancement, and image compensation of the camera according to factors such as light, weather, air quality, and mechanical vibration in the construction area.

[0024] Specifically, complex environments are usually common operating environments in pumped storage power stations such as low light and dust. It should be noted that data collection is carried out through fixed high-definition cameras, UAV inspection systems, and multispectral imaging devices.

[0025] Fixed high-definition cameras: Deployed in high-altitude construction operation areas, large equipment areas, and personnel-intensive areas to provide high-frame-rate and high-definition visible light images.

[0026] UAV inspection system: Deploy flight path optimization and automatic navigation algorithms to collect construction images at high altitudes or in complex environments.

[0027] Multispectral imaging device: Combine infrared light and depth imaging technologies to obtain construction data in low-light and dusty environments.

[0028] During data collection, automatic exposure adjustment is performed to optimize the image acquisition quality in different lighting environments, noise suppression and contrast enhancement are carried out to improve the recognizability of image details, and multi-scale target enhancement is performed to ensure the accuracy of detecting construction personnel and equipment at a long distance.

[0029] S2: Apply deep learning neural networks and dynamic feature extraction techniques, combined with a behavior analysis model, to identify the status of construction workers, equipment, and the environment, detect abnormal behaviors, including not wearing safety helmets, personnel staying in dangerous areas for a long time, and abnormal operation of mechanical equipment, and perform target tracking and real-time updates.

[0030] Furthermore, the deep learning neural network uses multi-modal data input, including construction worker pose estimation, equipment operation status analysis, and environmental sensor data, combined with object classification, behavior recognition, and time series modeling.

[0031] It should be noted that the personnel identification and behavior analysis at the construction site adopt a multi-modal method of deep learning object detection, pose analysis, and authentication to ensure the intelligence and accuracy of construction safety supervision.

[0032] Adopt object detection algorithms such as YOLO and Faster R-CNN to detect construction workers in real time, and filter redundant targets based on non-maximum suppression to improve the detection confidence.

[0033] Combine multi-modal input (RGB + infrared + depth map) to optimize the detection stability in low-light, high-dust, and complex background environments.

[0034] Perform personnel authentication through FaceNet and ArcFace to manage the identity of personnel entering and leaving the construction site, ensuring that unauthorized personnel cannot enter the dangerous operation area.

[0035] Combine with the construction management database to compare the identities of construction workers in real time and record the entry and exit times to achieve access management at the construction site.

[0036] Adopt the OpenPose human pose detection algorithm to extract the key joint points of construction workers and calculate their behavior states.

[0037] The key nodes are set as the arms, legs, and hip joints.

[0038] Conduct trajectory anomaly analysis, expressed as: Where, is the equipment trajectory deviation index. is the monitoring time window. is the number of monitored devices. is the current speed of the equipment, is the normal working speed. is the current acceleration of the equipment, is the normal acceleration. is the current attitude angle of the equipment, is the standard angle. They are the speed, acceleration, and angle weighting parameters respectively. is the delay impact coefficient. is the complexity level of the construction site, is the normalization coefficient.

[0039] To perform the recognition, an intelligent recognition formula for the construction site is established. Combining object detection, pose analysis, behavior tracking, and anomaly analysis, the comprehensive risk factors of construction workers, equipment, and environmental status are calculated, expressed as: Among them, is the intelligent recognition risk coefficient of the construction site. is the monitoring period. is the total number of all personnel and equipment currently detected. is the behavior risk index of the is the th construction worker, calculated based on pose analysis, object detection, and violation behavior counting. is the th equipment status index, calculated based on operating status, vibration frequency, and temperature change. are the weight factors respectively, used to adjust the contributions of different parameters. is the historical number of violations of construction workers or equipment.

[0040] S3: Based on the risk level calculation model, construct a multi-level alarm mechanism, trigger local acoustic and optical warnings, remote management alarms, and task automatic scheduling, perform hierarchical processing, and give the prediction results for cloud collaborative management and remote scheduling. Adopt self-supervised learning and cross-scenario transfer learning to optimize the model.

[0041] Furthermore, the hierarchical processing includes setting the alarm level based on the risk level calculation model. Low-level alarms trigger local acoustic and optical warnings, medium-level alarms are pushed to the management platform, and high-level alarms trigger remote task scheduling and emergency responses.

[0042] Adopt a multi-dimensional risk assessment method to comprehensively analyze construction workers, equipment operation, and environmental factors, and calculate the risk level, expressed as: Among them, is the comprehensive risk index of the construction area. is the monitoring period. is the total number of all personnel and equipment currently detected. is the number of historical data windows, As a time trend correction function, it is used to calculate the impact factors of different time periods.

[0043] Eradicate The risk level is determined by the preset risk threshold and is divided into three levels: high, medium and low.

[0044] Low-level warnings are identified as violations of construction safety regulations but do not directly cause safety risks, including but not limited to: construction workers not wearing safety helmets, construction workers not wearing reflective vests, construction workers not wearing safety shoes or other personal protective equipment as required, construction workers smoking in non-smoking areas, construction workers entering non-operating areas, construction workers not wearing protective glasses or other protective equipment as required, construction workers not fastening safety belts or protective ropes as required, equipment operators not checking equipment status as required, equipment not inspected and maintained in accordance with safety standards, debris or obstacles appearing in the construction area, affecting normal operations, construction workers not walking along the prescribed route or walking around the work area at will, and other violations.

[0045] The early warning responds to on-site sound and light prompts, such as "Please wear a safety helmet", "Please leave non-working areas", and "Please wear protective equipment". Voice broadcast reminders, such as "Smoking is strictly prohibited at the construction site" and "Please walk according to the prescribed route". Violations are recorded in the management system, such as statistics on the number of personnel violations and equipment maintenance abnormalities. If the cumulative number of violations by construction personnel exceeds the set threshold, the system will automatically upgrade to a medium-level warning.

[0046] The medium-level warning identifies behaviors or equipment abnormalities that may lead to safety accidents, including but not limited to: construction workers staying in high-risk areas for a long time, construction workers staying around dangerous equipment for a long time, construction workers entering confined spaces such as machine rooms, tunnels or confined spaces without authorization, construction workers not using safety protection devices in high-altitude working areas, construction equipment experiencing abnormal vibrations, excessive temperatures, abnormal operating noises during operation, construction machinery overloading or overspeeding, sudden environmental changes at the construction site, such as abnormal weather conditions such as high temperature, high humidity, strong winds, construction workers approaching or touching live equipment, construction workers illegally using hand-held tools or other dangerous equipment in high-risk areas, construction workers standing on unsecured objects to perform operations, such as unsecured ladders and scaffolding without guardrails, etc.

[0047] The early warning response is a remote alarm to notify the management personnel, "A construction worker has stayed in the dangerous area for a long time", "The equipment is vibrating abnormally, please check", and the voice broadcast system reminds, "Please stay away from running mechanical equipment", "Please perform operations in accordance with safety regulations", and the electronic alarm screen in the construction area displays real-time warning information. The construction management system records violations, and the safety officer needs to check and confirm the handling within the specified time. If the medium-level warning is not lifted in time, the system automatically upgrades to a high-level warning.

[0048] A high-level warning refers to situations where a serious safety accident has occurred or is about to occur, which are deemed to exist, including but not limited to: construction workers falling from a height or having a fall accident, construction workers losing their ability to move, such as fainting, suffocating or losing balance, construction workers being involved in or trapped in mechanical equipment, construction machinery and equipment losing control, such as brake failure, sudden power outage or abnormal swinging, construction equipment failure, such as overload operation, hydraulic system failure, electrical failure, etc., abnormal gas leakage or excessive dust concentration at the construction site, open flame, smoke or explosion risks in the construction area, structural damage or collapse of the construction area due to unstable foundation, equipment collision and other factors, extreme environmental conditions at the construction site, such as thunderstorms, high temperatures, and high wind speeds, which may have serious impacts on personnel or equipment, and construction personnel failing to evacuate the warned high-risk area in accordance with safety regulations.

[0049] The early warning response is to stop the construction equipment urgently and issue an alarm "The equipment is out of control and the emergency brake mechanism has been triggered". The remote alarm notification management platform is immediately notified, and the real-time monitoring screen and accident location are pushed to trigger the on-site emergency rescue process, "People have fallen, please rescue immediately" and "Fire is discovered, please evacuate quickly". SMS or APP push notifications are sent to construction personnel and management personnel, and fire protection systems, emergency broadcasts, emergency lighting and other safety measures are linked to issue evacuation orders, "All personnel are requested to evacuate the construction area immediately". Construction accident information is automatically archived for subsequent analysis and improvement of safety measures.

[0050] It should be noted that cloud-based collaborative management includes distributed storage and access permission control, encrypted storage of construction monitoring data, and allocation of construction tasks and equipment resources based on remote scheduling strategies.

[0051] Distributed cloud storage is used to encrypt and store construction safety data to ensure data integrity. Combined with the authority management mechanism, the access rights of different users are controlled. Construction management personnel can access the system through PC or mobile terminals to view real-time monitoring data. Safety officers are dispatched remotely, and combined with historical data analysis, an optimized construction safety plan is formulated.

[0052] It should also be noted that self-supervised learning includes unsupervised feature alignment and transfer learning, using few-shot learning for object detection, and adjusting the parameters of the neural network. Traditional deep learning object detection requires a large amount of manually labeled data for supervised training, while the data collection cost in the construction environment is high and the labeling workload is large. The present invention introduces a self-supervised learning method into the object detection model, and through contrastive learning, generative pre-training, and rotation prediction, enables the model to adaptively optimize on unlabeled data, improving the generalization ability in the construction environment. The construction environment is complex and changeable, such as day and night, sunny and hazy weather, lighting and background in different construction sites, etc., resulting in a decline in the generalization ability of the model in some environments. The core idea of transfer learning is to utilize the knowledge of a pre-trained model and transfer it to a new task to improve the training efficiency. The present invention adopts the method of parameter transfer + fine-tuning, applies the model pre-trained on a large-scale dataset to the construction environment, and performs weight fine-tuning according to expert experience for the construction scenario, thereby reducing the data requirements and improving the recognition accuracy.

[0053] Furthermore, remote scheduling includes the system based on construction environment parameters, personnel distribution, and equipment status, adopting a scheduling optimization strategy, allocating construction tasks, and adjusting the scheduling plan in combination with real-time monitoring data.

[0054] Remote scheduling requires real-time monitoring of the location, working status, and type matching of construction personnel, and making optimized scheduling in combination with the construction plan.

[0055] When the current total number of personnel in the construction area exceeds the standard, is insufficient, or key types of work are not operating in the designated area, or the inspection personnel work overtime, a personnel transfer instruction is issued.

[0056] Identify whether there is an available device to execute the current construction task; whether the device is under maintenance or repair, and whether it meets the safe operating conditions. Whether the device is running at high load for a long time, whether the device is in an idle state, and it can be scheduled to other construction tasks. If there is a qualified device, device scheduling is performed.

[0057] Embodiment 2, an embodiment of the present invention, provides a safety construction monitoring system for a pumped storage power station based on image recognition, including a target capture module, a behavior recognition and analysis module, and a risk assessment and early warning module.

[0058] Among them, the target capture module is used to comprehensively monitor the construction site by using high-definition cameras, drones, and multispectral imaging devices, and optimize the target capture ability in complex environments in combination with the environment adaptive adjustment strategy; the behavior recognition and analysis module is used to adopt deep learning neural networks and dynamic feature extraction technologies, and combine with the behavior analysis model to identify the construction personnel, equipment, and environmental status, detect abnormal behaviors, including not wearing safety helmets, personnel staying in dangerous areas for a long time, and abnormal operation of mechanical equipment, and perform target tracking and real-time updates; the risk assessment and early warning module is used to build a multi-level alarm mechanism based on the risk level calculation model, trigger local audible and visual alarms, remote management alarms, and automatic task scheduling, perform hierarchical processing, give prediction results for cloud collaborative management and remote scheduling, and optimize the model by using self-supervised learning and cross-scenario transfer learning.

[0059] If a function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0060] The logic and / or steps represented in the flowchart or described in other ways herein, for example, can be considered as a definite sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch instructions from the instruction execution system, apparatus, or device and execute the instructions), or in combination with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device.

[0061] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection (electronic device) having one or more wirings, a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable media can even be paper or other suitable media on which a program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or otherwise processing as appropriate, and then storing it in a computer memory.

[0062] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having suitable combinational logic gate circuits, programmable gate arrays (PGA), field programmable gate arrays (FPGA), etc. It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

[0063] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A safety construction monitoring method for pumped storage power stations based on image recognition, characterized in that, It includes: Utilize high-definition cameras, drones, and multi-spectral imaging equipment to conduct all-round monitoring of the construction site, and combine with the environmental adaptive adjustment strategy to optimize the target capture ability in complex environments; Adopt deep learning neural networks and dynamic feature extraction technology, combined with the behavior analysis model, to identify the status of construction personnel, equipment, and the environment, detect abnormal behaviors, including not wearing safety helmets, personnel staying in dangerous areas for a long time, and abnormal operation of mechanical equipment, and conduct target tracking and real-time updates; Based on the risk level calculation model, construct a multi-level alarm mechanism to trigger local audible and visual warnings, remote management alarms, and automatic task scheduling, conduct hierarchical processing, give prediction results for cloud collaborative management and remote scheduling, and adopt self-supervised learning and cross-scenario transfer learning to optimize the model.

2. The safety construction monitoring method for pumped storage power stations based on image recognition according to claim 1, characterized in that: The environmental adaptive adjustment strategy includes dynamically adjusting the exposure parameters, noise suppression, contrast enhancement, and image compensation of the camera according to factors such as light, weather, air quality, and mechanical vibration in the construction area.

3. The safety construction monitoring method of the pumped storage power station based on image recognition according to claim 2, characterized in that: The deep learning neural network adopts multi-modal data input, including construction personnel pose estimation, equipment operation status analysis, and environmental sensor data, combined with target classification, behavior recognition, and time series modeling.

4. The safety construction monitoring method of the pumped storage power station based on image recognition according to claim 3, characterized in that: The hierarchical processing includes setting alarm levels based on the risk level calculation model. Low-level alarms trigger local audible and visual warnings, medium-level alarms are pushed to the management platform, and high-level alarms trigger remote task scheduling and emergency responses.

5. The method for monitoring the safe construction of a pumped-storage power station based on image recognition according to claim 4, characterized in that: The cloud collaborative management includes distributed storage and access permission control, encrypts the storage of construction monitoring data, and based on the remote scheduling strategy, allocates construction tasks and equipment resources.

6. The method for monitoring the safe construction of a pumped storage power station based on image recognition according to claim 5, wherein: The self-supervised learning includes unsupervised feature alignment and transfer learning, uses few-shot learning for target detection, and adjusts the neural network parameters.

7. The safety construction monitoring method for pumped storage power stations based on image recognition according to claim 6, characterized in that: The remote scheduling includes that the system, based on construction environment parameters, personnel distribution, and equipment status, adopts a scheduling optimization strategy to allocate construction tasks and adjusts the scheduling plan in combination with real-time monitoring data.

8. A system adopting the image recognition-based safety construction monitoring method for pumped storage power stations as described in any one of claims 1 to 7, characterized in that: It includes a target capture module, a behavior recognition and analysis module, and a risk assessment and warning module; The target capture module is used to utilize high-definition cameras, drones, and multi-spectral imaging equipment to conduct all-round monitoring of the construction site, and combine with the environmental adaptive adjustment strategy to optimize the target capture ability in complex environments; The behavior recognition and analysis module is used to adopt deep learning neural networks and dynamic feature extraction technology, combined with the behavior analysis model, to identify the status of construction personnel, equipment, and the environment, detect abnormal behaviors, including not wearing safety helmets, personnel staying in dangerous areas for a long time, and abnormal operation of mechanical equipment, and conduct target tracking and real-time updates; The risk assessment and warning module is used to based on the risk level calculation model, construct a multi-level alarm mechanism to trigger local audible and visual warnings, remote management alarms, and automatic task scheduling, conduct hierarchical processing, give prediction results for cloud collaborative management and remote scheduling, and adopt self-supervised learning and cross-scenario transfer learning to optimize the model.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the image recognition-based safety construction monitoring method for pumped storage power stations 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 image recognition-based safety construction monitoring method for pumped storage power stations described in any one of claims 1 to 7 are implemented.

Images