Hydraulic power plant security monitoring method and system

Abstract

The invention relates to the technical field of monitoring and security, in particular to a hydraulic power plant security monitoring method and system, and the method comprises the steps: dividing a monitoring region into a core region and a non-core region; collecting first image information of the core area at a first frequency; collecting second image information of the non-core area at a second frequency, wherein the first frequency is higher than the second frequency; face features in the first image information are extracted through a first face recognition model, similarity calculation is conducted on the face features and feature images in a first preset image library, if the similarity is larger than a first preset threshold value, the first detection result is marked as normal, and if not, the first detection result is marked as abnormal; the identification precision, the processing efficiency and the safety protection capability of the hydraulic power plant safety protection monitoring system are effectively improved through the technical means of area grading acquisition, differentiation model identification, human body detection pre-screening, target segmentation optimization and the like, and the method has good practical value and popularization prospect.

Description

Technical Field

[0001] This invention relates to the field of security monitoring, specifically to a security monitoring method and system for a hydropower plant. Background Technology

[0002] With the continuous expansion of hydropower plant infrastructure and the increasing demands for production safety, the safety and security of hydropower plants have become a key factor in ensuring the stable operation of power production. Traditional hydropower plant security monitoring systems mostly employ a single-frequency video acquisition method, collecting and analyzing video images from all areas within the plant at a uniform frequency. However, different areas within a hydropower plant have varying importance and security levels. For example, control rooms and areas with important equipment are core areas requiring focused monitoring, while non-core areas and living quarters have relatively lower monitoring requirements.

[0003] Existing technologies generally suffer from the following problems: First, the monitoring frequency is fixed and cannot be dynamically adjusted according to the regional security level, resulting in insufficient real-time monitoring of core areas or wasting computing resources on non-core areas; second, there is a lack of effective automatic identification mechanisms for abnormal behavior or personnel, relying solely on a single face recognition model without designing differentiated recognition models for different regional characteristics, which easily leads to reduced recognition accuracy or false alarms and missed alarms; third, the lack of pre-screening processing of human targets in monitoring images results in invalid recognition of irrelevant images, increasing the system's computational burden and affecting processing efficiency and response speed.

[0004] Therefore, there is a need for a security monitoring method and system that can adopt differentiated processing strategies for different areas of a hydropower plant, and combine human body detection and facial recognition technologies to improve recognition accuracy and monitoring efficiency. Summary of the Invention

[0005] Based on the above problems, this invention proposes a method and system for security monitoring of hydropower plants.

[0006] This invention is achieved through the following technical solution: A method for security monitoring of a hydropower plant includes: The monitoring area is divided into core areas and non-core areas; First image information of the core region is acquired at a first frequency; Second image information of the non-core region is acquired at a second frequency, wherein the first frequency is higher than the second frequency; The first image information is input into a pre-constructed first face recognition model, and a first detection result is output. The output of the first detection result includes: extracting facial features from the first image information through the first face recognition model, calculating the similarity between the facial features and feature images in a first preset image library, and if the similarity is greater than a first preset threshold, then the first detection result is marked as normal; otherwise, it is marked as abnormal. Based on the first detection result, determine whether there are any abnormal personnel in the core area; The second image information is input into a pre-constructed second face recognition model, and a second detection result is output. The output of the second detection result includes: extracting facial features from the second image information through the second face recognition model, calculating the similarity between the facial features and feature images in a second preset image library, and if the similarity is greater than a second preset threshold, the second detection result is marked as normal; otherwise, it is marked as abnormal. Based on the second detection result, it is determined whether there are any abnormal personnel in the non-core area.

[0007] Furthermore, before inputting the first image information into the pre-built first face recognition model, the process also includes: The first image information is input into a pre-built human body recognition model, and the first recognition result is output. Based on the first recognition result, it is determined whether a first target human body exists in the first image information; If the first target human body exists, then the step of inputting the first image information into a pre-built first face recognition model is performed.

[0008] Furthermore, if it is determined that a first target human body exists in the first image information, the method further includes: The first target human body is segmented to obtain a first human body image; The step of inputting the first image information into the pre-constructed first face recognition model specifically involves inputting the first human image into the pre-constructed first face recognition model and outputting the first detection result.

[0009] Furthermore, before inputting the second image information into the pre-built second face recognition model, the method further includes: The second image information is input into a pre-built human body recognition model, and the second recognition result is output. Based on the second recognition result, determine whether there is a second target human body in the second image information; If the second target human body exists, then the step of inputting the second image information into the pre-built second face recognition model is performed.

[0010] Furthermore, if it is determined that a second target human body exists in the second image information, the method further includes: The second target human body is segmented to obtain a second human body image; The step of inputting the second image information into the pre-constructed second face recognition model specifically involves inputting the second human image into the pre-constructed second face recognition model and outputting the second detection result.

[0011] This invention also proposes a security monitoring system for hydropower plants, comprising: The region division module is used to define the core and non-core areas of a hydropower plant. Image acquisition module: configured to acquire first image information of the core region at a first frequency and acquire second image information of the non-core region at a second frequency, wherein the first frequency is higher than the second frequency; The core area processing module includes: First face recognition unit: used to receive the first image information and process it using a pre-built first face recognition model; the first face recognition unit is configured as follows: Extract facial features from the first image information, calculate the similarity between the facial features and feature images in the first preset image library, and output the first detection result marked as normal if the similarity is greater than the first preset threshold, otherwise output the first detection result marked as abnormal. Core area judgment unit: used to determine whether there are abnormal personnel in the core area based on the first detection result; Non-core area processing modules include: The second face recognition unit is used to receive the second image information and process it using a pre-built second face recognition model; the second face recognition unit is configured as follows: Extract facial features from the second image information, calculate the similarity between the facial features and feature images in the second preset image library, and output a second detection result marked as normal if the similarity is greater than the second preset threshold, otherwise output a second detection result marked as abnormal. Non-core area judgment unit: used to determine whether there are abnormal personnel in the non-core area based on the second detection result.

[0012] Furthermore, the core area processing module also includes: The first human body recognition unit is used to receive the first image information before the first image information is input into the first face recognition unit, process it using a pre-built human body recognition model, and output the first recognition result. The first human body determination unit is used to determine whether a first target human body exists in the first image information based on the first recognition result; The first face recognition unit only receives and processes the first image information when the first human body judgment unit determines that a first target human body exists.

[0013] Furthermore, the core area processing module also includes: The first human body segmentation unit is used to perform image segmentation on the first target human body in the first image information when the first human body judgment unit determines that there is a first target human body, so as to obtain a first human body image. Specifically, the first face recognition unit is used to receive and process the first human body image output by the first human body segmentation unit.

[0014] Furthermore, the non-core area processing module also includes: The second human body recognition unit is used to receive the second image information before the second image information is input into the second face recognition unit, process it using a pre-built human body recognition model, and output the second recognition result. The second human body determination unit is used to determine whether a second target human body exists in the second image information based on the second recognition result; The second face recognition unit only receives and processes the second image information when the second human body judgment unit determines that a second target human body exists.

[0015] Furthermore, the non-core area processing module also includes: The second human body segmentation unit is used to perform image segmentation on the second target human body in the second image information to obtain a second human body image when the second human body judgment unit determines that there is a second target human body. Specifically, the second face recognition unit is used to receive and process the second human body image output by the second human body segmentation unit.

[0016] The beneficial effects of this invention are: (1) The present invention proposes a method for security monitoring of hydropower plants, which significantly improves the real-time monitoring capability of the core area by dividing the monitoring area into core area and non-core area and using different image acquisition frequencies, while avoiding the waste of computing resources in non-core area and realizing the optimal allocation of resources. (2) The present invention proposes a security monitoring method for hydropower plants, which adopts a first face recognition model and a second face recognition model trained separately for different areas, so that the face recognition of core areas and non-core areas can be optimized according to the actual scene characteristics, thereby improving the overall recognition accuracy and robustness and reducing the risk of false alarms and missed alarms. (3) The present invention proposes a method for security monitoring of hydropower plants, which introduces a human body recognition model before face recognition, and first determines whether there is a human body in the image, avoiding redundant face feature extraction and comparison operations for images without human bodies, thereby improving image processing efficiency and reducing system load. (4) The present invention proposes a method for security monitoring of hydropower plants. After identifying the target human body, it performs image segmentation to extract a clearer and more accurate human body image for subsequent face recognition, avoiding background interference and the influence of non-target areas, thereby improving the quality of face feature extraction and the accuracy of similarity calculation. (5) The hydropower plant security monitoring system proposed in this invention can determine whether there are abnormal personnel based on the face recognition results, and promptly detect unauthorized personnel entering the core or non-core areas, providing a higher level of security for the hydropower plant and helping to prevent the occurrence of safety accidents such as illegal intrusion and misoperation. This invention effectively improves the identification accuracy, processing efficiency, and security protection capabilities of the hydropower plant security monitoring system through regional hierarchical data acquisition, differentiated model recognition, human body detection pre-screening, and target segmentation optimization. It has good practical value and promising prospects for promotion. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used 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 those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a hydropower plant security monitoring method proposed in this invention; Figure 2 This is a schematic diagram of the terminal equipment for a hydropower plant security monitoring method proposed in this invention; Figure 3 A schematic diagram of a readable storage medium for a hydropower plant security monitoring method proposed in this invention; In the diagram, 200 is the terminal device, 210 is the memory, 211 is the RAM, 212 is the cache, 213 is the ROM, 214 is the program / utility, 215 is the program module, 220 is the processor, 230 is the bus, 240 is the external device, 250 is the I / O interface, 260 is the network adapter, and 300 is the program product. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0020] Example 1 refer to Figure 1 A large hydropower plant divides its production area into a core area, including the generator room, central control room, high-voltage transformer area, and dam gate control station, and a non-core area, including administrative office buildings, general warehouses, peripheral passages, and living quarters. After system startup, the area division module first completes geospatial mapping, marking the coordinate range of core equipment as a red alert zone and the remaining areas as yellow regular zones. The image acquisition module then operates according to a preset strategy: the 2-megapixel wide-angle cameras in the core area generate high-density temporal first image information at a first frequency (30 frames / second) to ensure the capture of rapid personnel movement; the general cameras in the non-core areas output second image information at a second frequency (5 frames / second), significantly reducing storage and transmission load.

[0021] When the first image information flows into the core region processing module, it first reaches the first human body recognition unit. This unit uses a deep learning algorithm based on Faster R-CNN: the backbone network uses ResNet-101 to extract global image features, the Region Proposal Network (RPN) generates approximately 2000 candidate human body regions, and then the RoI pooling layer uniformly maps candidate boxes of different sizes to fixed-size feature maps. Finally, the classifier outputs the coordinates and confidence scores of the human body bounding boxes, forming the first recognition result. Based on this, the first human body judgment unit confirms the existence of the first target human body. If a human body bounding box with a confidence score of 98% is detected at the entrance of the control room, the first human body segmentation unit is activated. This unit uses the Mask R-CNN instance segmentation model: a mask branch is added to the Faster R-CNN framework, and each candidate human body region is segmented at the pixel level through a fully convolutional network (FCN). Its backbone network also uses ResNet-101, and after multi-scale fusion by the Feature Pyramid Network (FPN), a binary mask accurate to the human body contour is output, thereby removing background interference and generating a clean first human body image. The segmented first human body image is then transmitted to the first face recognition unit. The first face recognition model loaded in this unit adopts a Convolutional Neural Network (CNN) architecture: the input image is processed through 5 sets of convolutional-pooling layers, where the convolutional layers use 3×3 kernels for local feature extraction and enhance nonlinear expression capabilities through the ReLU activation function. The first convolutional kernel learns edge detector functions such as horizontal / vertical filters during training, while deeper convolutional kernels capture complex textures, including eye and nose contours. Finally, a 512-dimensional face feature vector is output through a global average pooling layer. This vector is compared with the cosine similarity features of high-definition registration photos of 500 core authorized personnel in a first preset image library. If the similarity to a registered personnel's features exceeds a first preset threshold of 0.92 (e.g., 0.95 when matching the chief engineer's features), a "normal" first detection result is output; if the highest similarity is only 0.75 (e.g., a stranger's face), it is marked as "abnormal." The core region judgment unit receives this result in real time; an abnormal signal triggers an audible and visual alarm and automatically closes the corresponding explosion-proof door.

[0022] In the synchronously running non-core region processing module, the second image information enters the second human body recognition unit at a low pace. A lightweight YOLOv5 algorithm is used here: the image is scaled to 640×640 resolution, features are extracted through the CSPDarknet53 backbone network, and multi-scale feature fusion is achieved through the path aggregation network PANet. The final output layer directly predicts the bounding box coordinates and class probabilities. When a human target (the second target human) is detected in the office building corridor with a confidence level of 90%, the second human body judgment unit activates the second human body segmentation unit. This unit uses a simplified version of the U-Net segmentation network: the encoder consists of four convolutional downsampling layers, and the decoder gradually restores the resolution through deconvolution, with skip connections fusing deep and shallow features. Although the accuracy is slightly lower than Mask R-CNN, the processing speed is increased by 3 times, and the output second human body image meets basic requirements. This second human body image is input into the second face recognition unit, whose second face recognition model uses the MobileFaceNet convolutional network: based on the inverse residual structure of MobileNetV2, linear bottleneck layers reduce computation, and depthwise separable convolutions replace standard convolution operations. The extracted 128-dimensional feature vector is compared with a second preset image library of 2000 people using a second preset threshold of 0.85. When a similarity of 0.88 is found between the features of administrative personnel and the target, a "normal" second detection result is output; if the similarity is 0.72 (e.g., the visitor is not registered), an anomaly is marked. The non-core area judgment unit generates a work order for the abnormal event and pushes it to the security PDA terminal.

[0023] In this embodiment, the core area of ​​the human detection stage uses a two-stage Faster R-CNN detector. The first stage, RPN, generates candidate boxes, such as 1200 anchor boxes in the control room scene. The second stage fine-tunes the bounding boxes and classifies them, ensuring a recall rate of 98.5% in dense scenes. The non-core area uses a single-stage YOLOv5 detector. The detection is completed in a single forward propagation, such as the corridor scene, which takes 15ms to process, meeting the real-time requirements.

[0024] In the human body segmentation stage, the core region is enhanced by Mask R-CNN by adding a mask prediction branch on the bounding box. Each RoI outputs a 28×28 binary mask, such as accurately separating overlapping people in front of a console. For non-core regions, U-Net is used based on an encoder-decoder structure, which preserves details through skip connections, such as segmenting distant human contours in warehouse surveillance.

[0025] In the face feature extraction stage, the core region CNN uses a deep structure of 5 sets of convolutional layers. The first layer has 64 7×7 convolutional kernels to learn basic edge features, and the deep layer has 256 3×3 kernels to encode the spatial relationship of facial features. In the non-core region CNN, MobileFaceNet uses depth-separable convolution, and the standard convolutional kernel is split into depthwise convolution (channels are processed independently) and point convolution (channels are fused), reducing the computational cost to 1 / 8.

[0026] The data flow in this embodiment includes: first image information → first human body recognition unit (Faster R-CNN) → first recognition result → first human body judgment unit → first human body segmentation unit (Mask R-CNN) → first human body image → first face recognition unit (deep CNN) → first detection result → core region judgment unit → access control linkage; Second image information → Second human body recognition unit (YOLOv5) → Second recognition result → Second human body judgment unit → Second human body segmentation unit (U-Net) → Second human body image → Second face recognition unit (lightweight CNN) → Second detection result → Non-core region judgment unit → Mobile alarm.

[0027] Through the synergy of CNN feature extraction, two-level detection algorithms, and segmentation models, the system achieves pixel-level human body separation and high-precision face matching with a false recognition rate of <0.1% in core areas, while balancing speed and accuracy by processing 35 frames per second in non-core areas. When anomalies are triggered consecutively in a non-core area, such as three unauthorized personnel detections at night around a warehouse perimeter, the system automatically increases the sampling frequency for that area to 15 frames per second and calls the core area segmentation model for verification, forming a dynamic security closed loop.

[0028] Example 2 This embodiment proposes a refined scheme for a hydropower plant security monitoring method based on embodiment 1.

[0029] Specifically, the facial feature extraction process in the core region using a Convolutional Neural Network (CNN) is as follows: Input the segmented face region (112×112 pixels) → First layer: 64 7×7 convolutional kernels (stride 2) to extract edge responses → Max pooling layer for dimensionality reduction → Second layer: 192 3×3 convolutional kernels to capture texture → Subsequent three layers: 256 3×3 convolutional kernels to model facial features → Global average pooling layer to compress spatial dimensions → Fully connected layer to output a 512-dimensional feature vector.

[0030] Feature visualization: Shallow convolution kernels activate facial contours such as the edge of the cheekbone, while deep convolution kernels respond to complex patterns such as the opening and closing of the eyes.

[0031] Differentiated implementations of human detection algorithms include: Faster R-CNN (Core Region): The Region Proposal Network (RPN) slides 9 anchor boxes on the feature map with a 3-scale × 3 aspect ratio. A positive and negative sample balancing strategy (IoU > 0.7 is positive, < 0.3 is negative) ensures training stability. The RoI pooling layer converts variable-sized candidate boxes into 7×7 fixed feature maps.

[0032] YOLOv5 (non-core region): The image is divided into a 20×20 grid, and each grid predicts 3 bounding boxes. The CIOU loss function optimizes the box position prediction, and the Focus structure downsampling reduces the amount of computation (e.g., 608×608 → 304×304).

[0033] Comparison of human body segmentation technologies: Mask R-CNN (Core Region): Based on Faster R-CNN, a mask branch is added. Each RoI outputs a 28×28 binary mask through FCN. The mask loss function adopts binary cross-entropy (e.g., to accurately separate overlapping people with a spacing of 0.5 meters).

[0034] U-Net (non-core region): The encoder downsamples 4 times (channel number 64→512) to capture the context, the decoder upsamples 4 times to restore the resolution, skip connections fuse local details, and the output layer uses sigmoid activation to generate a probability map.

[0035] This embodiment, by clearly defining the technical implementation path, deeply integrates technologies such as CNN feature extraction, YOLO / Faster R-CNN detection, and Mask R-CNN / U-Net segmentation into the hydropower plant security scenario while maintaining the first / second data flow mainline, forming an industrial-grade solution that combines accuracy and efficiency.

[0036] Example 3 refer to Figure 2 Based on Example 1, this example proposes a terminal device for a hydropower plant security monitoring method. The terminal device 200 includes at least one memory 210, at least one processor 220, and a bus 230 connecting different platform systems.

[0037] The memory 210 may include a readable medium in the form of volatile memory, such as RAM 211 and / or cache memory 212, and may further include ROM 213.

[0038] The memory 210 also stores a computer program, which can be executed by the processor 220, causing the processor 220 to perform any of the above-described applications of a hydropower plant security monitoring method in this application. The specific implementation method and the achieved technical effects are consistent with those described in the above-described application embodiments, and some details will not be repeated here. The memory 210 may also include a program / utility 214 having a set (at least one) of program modules 215. Such program modules include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment.

[0039] Accordingly, processor 220 can execute the aforementioned computer program, as well as executable program / utility 214.

[0040] Bus 230 can represent one or more of several types of bus structures, including a memory bus or memory controller, peripheral bus, graphics acceleration port, processor, or a local bus using any of the various bus structures.

[0041] Terminal device 200 can also communicate with one or more external devices 240, such as keyboards, pointing devices, Bluetooth devices, etc., and with one or more devices capable of interacting with it, and / or with any device that enables it to communicate with one or more other computing devices (e.g., routers, modems, etc.). This communication can be performed via I / O interface 250. Furthermore, terminal device 200 can communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via network adapter 260. Network adapter 260 can communicate with other modules of terminal device 200 via bus 230. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with terminal device 200, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms.

[0042] Example 4 refer to Figure 3 This embodiment proposes a readable storage medium for a hydropower plant security monitoring method. The computer-readable storage medium stores instructions that, when executed by a processor, implement any of the above-mentioned hydropower plant security monitoring methods. The specific implementation method and the technical effects achieved are consistent with those described in the above-mentioned application embodiments, and some details will not be repeated.

[0043] Figure 3The present embodiment illustrates a program product 300 for implementing the above-described applications. This product may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may run on a terminal device, such as a personal computer. However, the program product 300 of the present invention is not limited thereto. In this embodiment, the readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device. The program product 300 may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0044] Computer-readable storage media may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable storage medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof. Program code for performing operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on a user computing device, partially on a user device, as a standalone software package, partially on a user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing devices can be connected to user computing devices via any type of network, including local area networks (LANs) or wide area networks (WANs), or they can be connected to external computing devices (e.g., via the Internet using an Internet service provider).

[0045] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for security monitoring in a hydropower plant, characterized in that, include: The monitoring area is divided into core areas and non-core areas; First image information of the core region is acquired at a first frequency; Second image information of the non-core region is acquired at a second frequency, wherein the first frequency is higher than the second frequency; The first image information is input into a pre-constructed first face recognition model, and a first detection result is output. The output of the first detection result includes: extracting facial features from the first image information through the first face recognition model, calculating the similarity between the facial features and feature images in a first preset image library, and if the similarity is greater than a first preset threshold, then the first detection result is marked as normal; otherwise, it is marked as abnormal. Based on the first detection result, determine whether there are any abnormal personnel in the core area; The second image information is input into a pre-constructed second face recognition model, and a second detection result is output. The output of the second detection result includes: extracting facial features from the second image information through the second face recognition model, calculating the similarity between the facial features and feature images in a second preset image library, and if the similarity is greater than a second preset threshold, the second detection result is marked as normal; otherwise, it is marked as abnormal. Based on the second detection result, it is determined whether there are any abnormal personnel in the non-core area.

2. The method for security monitoring of a hydropower plant according to claim 1, characterized in that, Before inputting the first image information into the pre-built first face recognition model, the method further includes: The first image information is input into a pre-built human body recognition model, and the first recognition result is output. Based on the first recognition result, it is determined whether a first target human body exists in the first image information; If the first target human body exists, then the step of inputting the first image information into a pre-built first face recognition model is performed.

3. The method for security monitoring of a hydropower plant according to claim 2, characterized in that, If it is determined that a first target human body exists in the first image information, the method further includes: The first target human body is segmented to obtain a first human body image; The step of inputting the first image information into the pre-constructed first face recognition model specifically involves inputting the first human image into the pre-constructed first face recognition model and outputting the first detection result.

4. The method for security monitoring of a hydropower plant according to claim 1, characterized in that, Before inputting the second image information into the pre-built second face recognition model, the method further includes: The second image information is input into a pre-built human body recognition model, and the second recognition result is output. Based on the second recognition result, determine whether there is a second target human body in the second image information; If the second target human body exists, then the step of inputting the second image information into the pre-built second face recognition model is performed.

5. A method for security monitoring of a hydropower plant according to claim 4, characterized in that, If it is determined that a second target human body exists in the second image information, the method further includes: The second target human body is segmented to obtain a second human body image; The step of inputting the second image information into the pre-constructed second face recognition model specifically involves inputting the second human image into the pre-constructed second face recognition model and outputting the second detection result.

6. A security monitoring system for a hydropower plant, characterized in that, include: The region division module is used to define the core and non-core areas of a hydropower plant. Image acquisition module: configured to acquire first image information of the core region at a first frequency and acquire second image information of the non-core region at a second frequency, wherein the first frequency is higher than the second frequency; The core area processing module includes: First face recognition unit: used to receive the first image information and process it using a pre-built first face recognition model; the first face recognition unit is configured as follows: Extract facial features from the first image information, calculate the similarity between the facial features and feature images in the first preset image library, and output the first detection result marked as normal if the similarity is greater than the first preset threshold, otherwise output the first detection result marked as abnormal. Core area judgment unit: used to determine whether there are abnormal personnel in the core area based on the first detection result; Non-core area processing modules include: The second face recognition unit is used to receive the second image information and process it using a pre-built second face recognition model; the second face recognition unit is configured as follows: Extract facial features from the second image information, calculate the similarity between the facial features and feature images in the second preset image library, and output a second detection result marked as normal if the similarity is greater than the second preset threshold, otherwise output a second detection result marked as abnormal. Non-core area judgment unit: used to determine whether there are abnormal personnel in the non-core area based on the second detection result.

7. A hydropower plant security monitoring system according to claim 6, characterized in that, The core region processing module also includes: The first human body recognition unit is used to receive the first image information before the first image information is input into the first face recognition unit, process it using a pre-built human body recognition model, and output the first recognition result. The first human body determination unit is used to determine whether a first target human body exists in the first image information based on the first recognition result; The first face recognition unit only receives and processes the first image information when the first human body judgment unit determines that a first target human body exists.

8. A hydropower plant security monitoring system according to claim 7, characterized in that, The core region processing module also includes: The first human body segmentation unit is used to perform image segmentation on the first target human body in the first image information when the first human body judgment unit determines that there is a first target human body, so as to obtain a first human body image. Specifically, the first face recognition unit is used to receive and process the first human body image output by the first human body segmentation unit.

9. A hydropower plant security monitoring system according to claim 6, characterized in that, The non-core area processing module also includes: The second human body recognition unit is used to receive the second image information before the second image information is input into the second face recognition unit, process it using a pre-built human body recognition model, and output the second recognition result. The second human body determination unit is used to determine whether a second target human body exists in the second image information based on the second recognition result; The second face recognition unit only receives and processes the second image information when the second human body judgment unit determines that a second target human body exists.

10. A hydropower plant security monitoring system according to claim 9, characterized in that, The non-core area processing module also includes: The second human body segmentation unit is used to perform image segmentation on the second target human body in the second image information to obtain a second human body image when the second human body judgment unit determines that there is a second target human body. Specifically, the second face recognition unit is used to receive and process the second human body image output by the second human body segmentation unit.

Images