Implementation method of sound-light alarm linkage terminal device under external force construction monitoring

Abstract

The invention provides an implementation method of a sound-light alarm linkage terminal device under external force construction monitoring. The implementation method comprises the following specific steps that a low-power-consumption image sensing terminal deployed along a power transmission line continuously collects channel environment images; performing target detection and behavior modeling on the uploaded image, identifying whether an external force construction behavior exists or not, and judging whether the image enters a high-voltage line protection area or not; when the system judges that the invasion risk exists, an acousto-optic warning signal is sent out, and field intervention is carried out on constructors; and after completing the acousto-optic intervention, the mobile terminal uploads the event to a remote platform. The method has the advantages that the construction behavior can be accurately identified and the risk level can be evaluated in real time; after potential threats are identified, on-site illegal behaviors are quickly stopped; the response speed, the intervention effect and the intelligent management level of the power grid to external force damage behaviors are improved, and an efficient, intelligent and sustainable evolution early warning protection system is constructed for high-voltage line channel safety.

Description

Technical Field

[0001] This invention relates to the field of power system operation and maintenance technology, and in particular to a method for implementing an audible and visual alarm linkage terminal device under external construction monitoring. Background Technology

[0002] As a crucial component of the power system, the safe and stable operation of transmission lines directly impacts the overall reliability of the power grid. However, with the accelerating pace of urbanization, numerous municipal construction projects, infrastructure developments, and agricultural operations frequently occur near transmission line corridors, making external damage a major contributing factor to transmission line accidents. Common external construction activities include excavation, drilling, hoisting, pile driving, and vehicle collisions. If these activities approach or accidentally contact transmission equipment, they can easily trigger tripping, line breaks, or even widespread power outages, resulting in significant safety hazards and economic losses.

[0003] Currently, the monitoring and early warning of external damage to transmission lines in power systems still mainly rely on manual inspections, fixed-point video surveillance, and traditional audible and visual alarms. Manual inspections suffer from significant lag and uncertainty, failing to cover the operational status of lines around the clock and across all areas. While traditional video surveillance equipment can acquire image information, most lack intelligent recognition capabilities, unable to automatically determine whether external construction activities pose a threat, and their accuracy drops sharply at night or under complex weather conditions. Furthermore, existing audible and visual alarm devices are mostly triggered at fixed points or manually controlled, lacking intelligent linkage with external construction activity recognition systems. This makes it difficult to achieve proactive response and risk level differentiation, leading to frequent false alarms and missed alarms, severely impacting the warning effect and efficiency.

[0004] Furthermore, traditional systems typically lack data logging and remote management capabilities. When an alarm occurs, they cannot automatically generate event data or transmit it back to the backend system for analysis and tracing, which is particularly lagging behind in the context of increasingly important intelligent operation and maintenance of power grids. The lack of closed-loop control and self-learning mechanisms also prevents the system from continuously optimizing its identification model and alarm strategies based on historical data, making it difficult to adapt to the diversity of construction activities and the complexity of the power transmission environment.

[0005] In view of the above-mentioned problems and limitations, this invention proposes a method for implementing an audible and visual alarm linkage terminal device under external force construction monitoring, which overcomes the limitations of traditional transmission line external damage prevention technologies in terms of identification delay, intervention lag, and incomplete control loop. This device deploys low-power image sensing terminals to collect real-time images of the environment surrounding the transmission channel and uses an IoT communication module to remotely upload the image data to the monitoring platform. Based on image recognition and behavior analysis algorithms, the monitoring center accurately judges the approach behavior of construction machinery or personnel and automatically identifies the existence of intrusion risks by combining preset line protection boundary rules. Once an anomaly is confirmed, the system immediately sends audible and visual control commands to the on-site linkage terminal, driving a 360° high-brightness flashing light and a high-volume warning horn to work together, emitting strong visual and audible signals to intervene in construction behavior and enhance safety warning intensity. The backend system synchronously records and uploads event images, trigger times, and response data, enabling traceability of construction behavior, risk documentation, and continuous optimization of the identification model. This invention forms a closed-loop mechanism covering the entire process of "image acquisition - intelligent recognition - sound and light linkage - data feedback", which not only enables proactive prevention and control of external intrusion risks, but also significantly improves the safety, intelligence and response efficiency of power grid operation. Summary of the Invention

[0006] The purpose of this invention is to provide a method for implementing an audible and visual alarm linkage terminal device under external force construction monitoring. This not only enables all-weather, low-power, and intelligent monitoring and identification of external force construction activities around high-voltage transmission lines, but also achieves proactive intervention and real-time early warning of illegal operations through an audible and visual linkage mechanism, effectively reducing the threat of external damage events to the safe operation of transmission lines.

[0007] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0008] This invention provides a method for implementing an audible and visual alarm linkage terminal device under external force construction monitoring, the specific steps of which are as follows:

[0009] S1: Low-power image sensing terminals deployed along the transmission line continuously collect images of the channel environment and remotely transmit real-time data to the monitoring center through the Internet of Things communication module, providing an image basis for subsequent analysis;

[0010] S2: The monitoring center uses a big data analysis platform and artificial intelligence image recognition algorithms to perform target detection and behavior modeling on uploaded images, identify whether there is external construction behavior, and determine whether it has entered the high-voltage line protection zone.

[0011] S3: When the system determines that there is an intrusion risk, it immediately sends a control command to the front-end linkage terminal, triggering the 360° high-brightness flashing light and the 110dB high-volume alarm to work synchronously, emitting an audible and visual warning signal to intervene on-site for construction personnel.

[0012] S4: After completing the audio-visual intervention, the mobile terminal uploads the image clips, trigger time, warning level and response log of this event to the remote platform for background recording and archiving, behavior analysis and self-learning optimization of the recognition model.

[0013] In step S1, this invention employs a low-power image acquisition terminal with a built-in image frame buffering mechanism and differential processing module, supporting continuous frame analysis of channel scenes. The system enhances the contour recognition effect of external moving targets through inter-frame image differences, and extracts image feature vectors using a convolutional neural network (CNN) to construct candidate detection boxes. To improve recognition accuracy, the image input is first processed by a background modeling and noise filtering module to remove low-risk interference targets such as rustling grass and flying birds, retaining only significant targets with engineering operation characteristics such as human bodies, robotic arms, and engineering vehicles. All image acquisition and recognition are performed at low frequency and low computing power to ensure the terminal's continuous working capability.

[0014] In step S2, after the image recognition results are uploaded to the monitoring center, they are further classified and analyzed using artificial intelligence algorithms. The platform introduces a motion trend modeling mechanism to dynamically fuse and judge information such as the target's posture, direction of travel, and acceleration. The system sets logical criteria to identify the continuous state of "approaching - staying - working," forming a multi-stage judgment model for external force construction behavior. Simultaneously, the platform supports setting line safety protection boundaries using custom rules, without relying on GIS data structures. It can construct alarm trigger ranges based on parameters such as the vertical distance from the equipment installation point to the conductor and safety tolerance, achieving spatial intrusion judgment under simplified deployment conditions.

[0015] In step S3, this invention constructs a multi-level sound and light linkage strategy control model, which can automatically match the intervention intensity according to the risk level in the identification results. For suspected approaching behavior, only a high-brightness flashing light is activated for early warning; if the behavior continues and approaches the set trigger range, a 110dB high-volume alarm is activated simultaneously to implement forced intervention. The system supports remote parameter distribution and dynamic adjustment, and the frequency, rhythm, duration, etc. of the on-site sound and light equipment can be configured as needed. To prevent personnel from becoming desensitized to repeated alarms, the system integrates a short-term "silent interval" mechanism, intermittently triggering the sound and light equipment to enhance the intervention effect and control environmental interference.

[0016] In step S4, after the event concludes, the linked terminal encapsulates the entire process data in a structured manner, forming a complete record containing fields such as target image fragments, trigger timestamps, risk levels, and response action status. The data is uploaded to the remote platform using a lightweight encryption format (such as JSON+AES). The platform has the capability to resume interrupted uploads and reconnect offline, adapting to complex communication environments. The backend system performs clustering statistics and behavioral label training on the uploaded events, analyzes false positives and false negatives based on actual handling results, and continuously optimizes the recognition model parameters, achieving a self-learning and evolutionary closed loop for the platform's recognition performance.

[0017] The present invention has the following beneficial effects:

[0018] This invention proposes a method for implementing an audible and visual alarm linkage terminal device under external force construction monitoring, overcoming the technical bottlenecks of traditional transmission line external damage early warning systems, such as single identification methods, delayed response, crude intervention methods, and insufficient data closure. The device deploys low-power image sensing terminals and utilizes IoT communication technology to upload environmental images of the transmission channel to the monitoring center in real time. Based on artificial intelligence image recognition algorithms and behavioral modeling mechanisms, the system can accurately identify construction behaviors and assess their risk levels in real time. After identifying potential threats, it can link control terminals to trigger multi-mode audible and visual intervention signals, achieving rapid cessation of on-site violations. Simultaneously, the system possesses a comprehensive event archiving and model self-optimization mechanism, supporting intelligent analysis of historical data and continuous iterative training of the front-end identification model, forming a closed-loop protection path of "identification—intervention—learning—re-identification." Through these technical means, this invention significantly improves the power grid's response speed, intervention effect, and intelligent management level against external force damage, constructing an efficient, intelligent, and sustainably evolving early warning and protection system for high-voltage transmission line safety. Attached Figure Description

[0019] Figure 1 This is a flowchart of a remote millimeter-wave ranging and AI recognition system applicable to power transmission lines. Detailed Implementation

[0020] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0021] Example

[0022] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0023] This invention provides a method for implementing an audible and visual alarm linkage terminal device under external force construction monitoring, with reference to... Figure 1 As shown, the method includes:

[0024] S1: Low-power image sensing terminals deployed along the transmission line continuously collect images of the channel environment and remotely transmit real-time data to the monitoring center through the Internet of Things communication module, providing an image basis for subsequent analysis;

[0025] In this invention, low-power image sensing terminals deployed at equal intervals (e.g., every 300 meters) along the transmission line possess image acquisition, preprocessing, judgment, uploading, and wireless communication capabilities. These terminals are used to continuously acquire images of the high-voltage line's surrounding area and determine whether there are signs of external intrusion. The terminal structure is as follows: Figure 1 As shown, the core components include an image acquisition module, an MCU control unit, a Flash cache module, and an NB-IoT wireless communication module.

[0026] 1. Image acquisition and change detection modeling In each image acquisition cycle, the terminal compares the global differences between two consecutive frames. The definition is as follows: in, For the image in frame t, at pixel position The grayscale value at that location; The width and height of the image; This represents the total number of pixels in the image. This is a mean grayscale difference index used to measure the intensity of image changes.

[0027] when (The threshold is set based on experimental experience, with a typical value of 10~15). If a significant change is detected, the image must be uploaded to the monitoring platform; otherwise, it will not be uploaded to save communication traffic and energy consumption.

[0028] 2. Image Compression and Bandwidth Estimation Using the JPEG-LS compression algorithm, the image compression ratio can reach: in, The original image size (e.g., 320×240×8bit = 614.4KB); This refers to the compressed image size (e.g., 50KB); common image compression ratios. .

[0029] If the communication frequency is twice per minute, the total amount of image data uploaded per day is: The platform's traffic load can be dynamically controlled by configuring the image compression level and upload trigger threshold.

[0038] 3. Energy control and system runtime prediction

[0039] The terminal uses a solar panel + lithium battery power supply system. The power consumption per unit cycle (upload or sleep) is calculated as follows:

[0040] E cycle =P acq ·t acq +P tx ·t tx +P sleep ·t sleep (4)

[0041] Among them, P acq The power consumption for image acquisition is approximately 30mW; P tx For NB-IoT transmission power consumption, approximately 150mW; P sleep Power consumption in sleep mode, <1mW; t * The duration (in seconds) of each state.

[0042] The average daily energy consumption is:

[0043] E daily =N cycle ·E cycle (5)

[0044] If the battery capacity is 2000mAh, the system energy budget is:

[0045]

[0046] The voltage is U = 3.7V and the capacity is C = 2000mAh to ensure that the system can operate for more than 7 consecutive days during rainy weather.

[0047] S2: The monitoring center uses a big data analysis platform and artificial intelligence image recognition algorithms to perform target detection and behavior modeling on uploaded images, identify whether there is external construction behavior, and determine whether it has entered the high-voltage line protection zone.

[0048] In step S1, after image acquisition and remote uploading are completed, the monitoring center receives image data streams from multiple terminals and inputs them into an artificial intelligence recognition system built on a big data platform. This system integrates target detection algorithms, behavior state modeling mechanisms, and line area boundary rule judgment modules to achieve intelligent recognition and intrusion determination of external force construction behavior.

[0049] 1. Target Detection Module

[0050] The image recognition system uses a lightweight convolutional neural network (such as YOLOv5-Lite or MobileNet-SSD) to perform forward inference on the uploaded images, outputting the location information and classification labels of all salient targets. Each image generates a target set after recognition.

[0051]

[0052] Among them, b i The target bounding box coordinates (center point + width and height); c i For target categories (e.g., construction workers, excavators, tower cranes, etc.); s i Confidence score. The system only retains scores with a confidence level greater than a set threshold s. i Targets with a value >0.6 proceed to the next step of analysis to filter out false positives.

[0053] 2. Behavioral Modeling and Risk Assessment Mechanism

[0054] To identify construction activities, the platform establishes a trajectory sequence by analyzing the displacement paths of the same target in consecutive frame images:

[0055]

[0056] Where (x) k y k ) represents the coordinates of the target center point, t k This is the timestamp of the image frame.

[0057] The velocity vector of the trajectory is:

[0058]

[0059] The platform performs dynamic feature extraction on the target, including: approach speed. Whether to rapidly approach the line; duration of stay T s The duration of continuous occurrence within a certain area; directional trend: whether the target's movement direction points towards the center of the channel.

[0060] When the following comprehensive behavioral criteria are met, it is considered to have the characteristics of external force construction:

[0061] v n >v th T s >T min cosθ>cosθ th (10)

[0062] Among them, v th To be close to the speed threshold; T min Cosθ represents the minimum dwell time; θ is the angle between the target's direction of motion and the guide wire direction. this the direction threshold, usually set to cos(30°) or less.

[0063] 3. Channel boundary matching and intrusion determination

[0064] To replace GIS, the platform presets a fixed two-dimensional boundary area according to the distance between the device deployment point and the wire, which is represented as the protection zone rectangular area R safe , and sets the following boundaries:

[0065] R safe ={(x, y)|x min ≤x≤x max , y min ≤y≤y max (11)

[0066] Once the center point of the target trajectory falls into this area and the behavior model determines that it is continuously approaching, it is determined that the target "has entered the protection area" and an alarm signal is triggered.

[0067] S3: When the system determines that there is an intrusion risk, it immediately sends a control instruction to the front-end linkage terminal, triggering the synchronous operation of the 360° high-brightness flashing light and the 110dB high-pitched warning device to emit an audible and visual warning signal to implement on-site intervention on the construction personnel;

[0068] Among them, after identifying that the target has entered the high-voltage line protection area and completing the risk level determination, the system generates audible and visual linkage control parameters according to the alarm level and drives the terminal execution device to perform active intervention. The trigger logic, parameter adjustment method and dynamic response process of this audible and visual linkage mechanism are described as follows:

[0069] First, the system adjusts the audible and visual output parameters in real time according to the risk score value . The adjustment of each control parameter is mapped by a linear or non-linear function to ensure that the linkage behavior is finely adjusted according to the risk level. Specifically as follows:

[0070] 1. LED flashing frequency adjustment:

[0071] The flashing frequency f LED [[ID=​​​​​​​​​​​​​​​​

[0075] The speaker volume is increased on a risk-based basis, using the following formula:

[0076]

[0077] Here, L0 = 90dB is the base sound pressure level, and k = 20dB is the increment. This method ensures that lower-risk targets receive only a mild alert, while high-risk targets can trigger a strong intervention sound wave of 110dB.

[0078] 3. Response time adjustment mechanism:

[0079] Alarm duration T alarm Use a square function to enhance high-risk latency and improve alarm persistence. The function is as follows:

[0080] Among them, T min =3 seconds, T max =10 seconds, ensuring that response time increases non-linearly with high-risk behavior.

[0081] For example, in a typical case: when the identified risk score is The system will then output an LED flashing frequency of approximately 16.25Hz and a speaker sound pressure level of 105dB, and will continuously issue a warning for approximately 6.9 seconds, creating sufficiently strong visual and auditory stimulation to achieve the purpose of strong intervention.

[0082] Secondly, to avoid frequent triggering that could cause worker fatigue or on-site disruption, this invention also incorporates an alarm scheduling and control strategy. The system sets a minimum trigger interval Δt for the same construction target. safe If the current time is less than the interval since the last warning, the alarm will not be triggered again; at the same time, the maximum number of triggers N per unit time is limited. max This ensures that alarm-related behaviors are characterized as "focused, limited, and controllable".

[0083] In actual operation, the system also incorporates an "incremental linkage response mechanism." If the target fails to evacuate after the initial audio-visual intervention, the system will increase the LED flashing frequency and speaker volume, progressively strengthening the intervention intensity. This mechanism can be described by the following principle: Where n is the current alarm count, , The incrementing step size is used for each round.

[0086] In addition, to address recognition errors caused by changes in ambient background noise, this system can incorporate background noise N measured by on-site sensors. envIt adjusts the volume output. When ambient noise increases, it automatically boosts the speaker output; when the background is quiet, it automatically reduces the sound pressure level to avoid disturbing unrelated people in the vicinity, ensuring the human-centered adaptability of the acoustic intervention.

[0087] S4: After completing the audio-visual intervention, the mobile terminal uploads the image clips, trigger time, warning level and response log of this event to the remote platform for background recording and archiving, behavior analysis and self-learning optimization of the recognition model.

[0088] In step S3, after the audible and visual alarm linkage intervention is completed, the terminal enters the event closed-loop recording and data uploading stage. The terminal designed in this invention has local event encapsulation, structured log construction, breakpoint resume upload mechanism, and platform-side incremental optimization function for the identification model, realizing a complete closed-loop control from "perception response" to "data accumulation" and "intelligent evolution".

[0089] 1. Event data recording and structured encapsulation

[0090] After the audio-visual response is completed, the linkage terminal immediately calls the internal MCU control logic to package the multi-source information within this event cycle into a structured record. The encapsulated content includes:

[0091] The event ID (EID) is an automatically generated unique event identifier; the target classification label is C. obj Such as personnel, excavators, tower cranes, etc.; image summary data I t Three keyframe images before, during, and after the event; trigger timestamp T start T end The start and end times of the early warning; risk score. Output from step S2; Acousto-optic response parameter f LED L sound , t alarm Intervention status label S response To determine whether the intervention was effective; a snapshot of the device's operating status (such as battery level, signal strength, and ambient brightness).

[0092] The event log data is organized in JSON format as follows:

[0093]

[0094] To ensure log reliability, all event data is cached locally on the terminal in a Flash circular buffer, with a set cache depth (e.g., supporting up to 72 hours of continuous events).

[0095] 2. Data upload mechanism and fault tolerance design

[0096] The terminal transmits event data uplink via MQTT or HTTP protocols through an NB-IoT or 4G communication module. To improve robustness, the system employs a breakpoint resumption mechanism and a link confirmation mechanism: data packet numbering is... Mapping; for each uploaded event log packet, the platform returns an ACK flag; if the terminal fails to upload continuously for more than N times... timeout If the request is repeated, the user will be automatically returned to the local retry queue.

[0097] Each uploaded packet size should be controlled between 10 and 50KB, with a compression ratio of:

[0098]

[0099] 3. Platform-side data parsing and archiving management

[0100] After receiving the uploaded data, the platform writes all event logs into the event database and establishes the following index fields: time dimension (search by minute / hour); target type; risk level; terminal ID and geographical location; the platform can perform cluster analysis, generate risk distribution heat maps, and assess equipment operating status.

[0101] 4. Incremental Model Optimization Mechanism

[0102] The platform of this invention has self-learning capabilities. Based on continuously accumulated event data samples, the system executes the following identification and optimization strategies:

[0103] (1) Sample label extraction: "True positive / false positive / missed positive" is automatically labeled by manual verification or based on intervention results;

[0104] (2) Incremental training mechanism: new samples are appended to the training dataset. Fine-tuning training is performed periodically using the sliding window method;

[0105] (3) Model redeployment mechanism: The platform regularly distributes the optimized recognition model weights or parameters to the front-end devices to complete the seamless version iteration;

[0106] Example of optimizing the feedback function:

[0107]

[0108] Where θ t To identify model parameters, Let η be the loss function and η be the learning rate.

[0109] This invention proposes a method for implementing an audible and visual alarm linkage terminal device under external force construction monitoring, overcoming the technical bottlenecks of traditional transmission line external damage early warning systems, such as single identification methods, delayed response, crude intervention methods, and insufficient data closure. The device deploys low-power image sensing terminals and utilizes IoT communication technology to upload environmental images of the transmission channel to the monitoring center in real time. Based on artificial intelligence image recognition algorithms and behavioral modeling mechanisms, the system can accurately identify construction behaviors and assess their risk levels in real time. After identifying potential threats, it can link control terminals to trigger multi-mode audible and visual intervention signals, achieving rapid cessation of on-site violations. Simultaneously, the system possesses a comprehensive event archiving and model self-optimization mechanism, supporting intelligent analysis of historical data and continuous iterative training of the front-end identification model, forming a closed-loop protection path of "identification—intervention—learning—re-identification." Through these technical means, this invention significantly improves the power grid's response speed, intervention effect, and intelligent management level against external force damage, constructing an efficient, intelligent, and sustainably evolving early warning and protection system for high-voltage transmission line safety.

[0110] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention. The purpose of the present invention is to

[0111] Matters not covered in this invention are common knowledge. Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions, and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for implementing an audible and visual alarm linkage terminal device under external force construction monitoring, characterized in that: The specific steps are as follows: S1: The low-power image sensing terminal deployed along the power transmission line continuously acquires channel environment images, and transmits real-time data to the monitoring center through the Internet of Things communication module, providing image basis for subsequent analysis; S2: The monitoring center based on the big data analysis platform and artificial intelligence image recognition algorithm, target detection and behavior modeling are performed on the uploaded images to identify whether there is external construction behavior and whether it enters the high-voltage line protection area; S3: When the system determines that there is an invasion risk, control instructions are immediately sent to the front-end linkage terminal to trigger the 360° high-brightness strobe light and 110dB high-pitched alarm to work synchronously, and sound and light warning signals are issued to implement on-site intervention on the construction personnel; S4: After the terminal completes the sound and light intervention, the image segments, trigger time, warning level and response log of the event are uploaded to the remote platform for background record archiving, behavior analysis and self-learning optimization of the recognition model.

2. The implementation method of the terminal device of the acoustic-optic alarm linkage under the external force construction monitoring according to claim 1, characterized in that: In step S1, a low-power image acquisition terminal is used, which is internally provided with an image frame buffer mechanism and a difference processing module, which supports continuous frame analysis of the channel scene; the system enhances the contour recognition effect of external moving targets through inter-frame image difference, extracts image feature vectors combined with convolutional neural network CNN, and constructs candidate detection boxes; to improve recognition accuracy, the image input is first processed by the background modeling and noise filtering module, and low-risk interference targets such as wind and grass movement, flying birds, etc. are removed, and only significant targets such as human bodies, mechanical arms, engineering vehicles, etc. with engineering operation characteristics are retained; all image acquisition and recognition are based on low frequency and low algorithm, to ensure the terminal's continuous working ability.

3. The implementation method of the terminal device of the acoustic-optic alarm linkage under the external force construction monitoring according to claim 1, characterized in that: In step S2, after the image recognition result is uploaded to the monitoring center, further classification and behavior analysis are performed through artificial intelligence algorithm, and the platform introduces action trend modeling mechanism to dynamically fuse and judge the information such as the posture, direction of travel and acceleration of the target; the system sets logical criteria to identify the continuous state of "approaching-staying-operation", forming a multi-stage judgment model of external construction behavior; at the same time, the platform supports setting the line safety protection boundary according to self-defined rules, without relying on GIS data structure, and can construct the alarm trigger range according to parameters such as the vertical distance from the equipment installation point to the conductor and the safety tolerance, to realize spatial intrusion judgment under simplified deployment conditions.

4. The implementation method of the terminal device of the acoustic-optic alarm linkage under the external force construction monitoring according to claim 1, characterized in that: In step S3, the application constructs a multi-level sound and light linkage strategy control model, which can automatically match the intervention intensity according to the risk level in the recognition result; for suspected approaching behavior, only the high-brightness strobe light is started to give a warning; If the behavior continues and approaches the set trigger range, the 110dB high-pitched alarm is started synchronously to implement forced intervention; the system supports remote parameter issuance and dynamic adjustment, and the frequency, rhythm and duration of the on-site sound and light equipment can be configured as needed; to prevent people from developing a numb psychology to repeated alarms, the system integrates a short "silent interval" mechanism to intermittently trigger the sound and light equipment, enhancing the intervention effect and controlling the environmental interference.

5. The implementation method of the terminal device of the acoustic-optic alarm linkage under the external force construction monitoring according to claim 1, characterized in that: In step S4, the linkage terminal encapsulates the whole process data after the event ends, forms a complete record containing target image segment, trigger timestamp, risk level, response action state and other fields, and uploads the data to the remote platform in a lightweight encryption format. The platform has the ability of breakpoint resume and offline reconnection, and is suitable for complex communication environment. The backend system clusters and counts the uploaded events and trains behavior labels, analyzes false positives and false negatives combined with actual handling effect, continuously optimizes recognition model parameters, and realizes self-learning and evolution closed loop of platform recognition performance.

Images