Multi-channel transmission optical fiber alarm system based on end side large model

Abstract

The invention relates to the technical field of transmission optical fibers, in particular to a multi-channel transmission optical fiber alarm system based on an end-side large model, and the system comprises a multi-source data fusion module which is used for collecting optical fiber vibration signals, environment data, video monitoring data and sound data; the laser signal transmitting power dynamic adjusting module comprises a laser, a U15 chip and a main chip; the multi-channel switching module comprises six polarizers which are divided into three groups of working areas, and each group of working areas is composed of two orthogonal polarizers; the end side large model module is arranged in the main chip; linkage and automatic adjustment are carried out through the end side large model, the laser emission signal adjusting circuit and the multi-channel switching module, the background noise influence in a complex environment can be reduced, the signal characteristic value of a vibration event is improved, the signal-to-noise ratio is improved, and therefore the intrusion behavior type can be rapidly and accurately detected.

Description

Technical Field

[0001] The present invention relates to the technical field of transmission optical fibers, and in particular to a multi-channel transmission optical fiber alarm system based on an end-side large model. Background Art

[0002] As the infrastructure of modern communication, optical fiber transmission has the characteristics of high bandwidth, long transmission distance, strong anti-interference ability, etc., and is widely used in fields such as telecommunications, security, and industrial control. With the acceleration of the digitalization process, its coverage and transmission capacity requirements continue to grow, and the requirements for network stability are becoming increasingly strict.

[0003] At present, for the optical fiber vibration intrusion alarm devices on the market, mainly two-core single-mode optical fibers are used to form a balanced optical fiber device. When a coherent laser is used to emit a beam of laser light to it, the device composed of these two optical fibers outputs an interference optical signal. When the optical fiber is disturbed by the outside world, such as excavation, touch, knocking, etc., the output waveform of the interference light changes and an interference image is generated. By detecting this waveform change with a photodetector and then analyzing the characteristics of the changed waveform, the true situation of the event is distinguished, achieving the effect of "intrusion mode recognition". However, the high-sensitivity use of the optical fiber in the system is easily interfered by various factors, such as external factors like strong wind blowing and rain hitting, resulting in false alarms. How to improve the recognition accuracy of intrusion behavior and reduce false alarms and missed alarms has become the top priority of the product.

[0004] Existing optical fiber vibration intrusion alarm devices analyze the waveform of the vibration signal to extract the vibration characteristic data of different vibration events, thereby realizing event alarm. The above system can achieve a relatively high recognition accuracy in a simple environment. However, when encountering complex weather environments, such as strong winds of level 7-8 and heavy rain, the spectral signals triggered by intrusion behaviors are easily submerged by the background noise, the characteristic values are not obvious, and false alarms and missed alarms are likely to occur.

[0005] The existing system realizes intrusion recognition by analyzing the change of the interference optical waveform, and has the following defects: environmental noises such as strong wind and rainfall will submerge the spectral characteristics of the intrusion signal, resulting in false alarms or missed alarms; in extreme weather such as strong winds of level 7-8 and heavy rainstorms, the signal-to-noise ratio of the system is significantly reduced, and it is difficult to extract features; the existing solutions cannot automatically adjust parameters according to environmental changes, rely on static analysis models, and have a high misjudgment rate.

[0006] Aiming at the problem that in complex weather environments such as strong winds of level 7-8 and heavy rain, the spectral signals triggered by intrusion behaviors are easily submerged by the background noise, the characteristic values are not obvious, and false alarms and missed alarms are likely to occur, a multi-channel transmission optical fiber alarm system based on an end-side large model can be designed. Summary of the Invention

[0007] To overcome the problem that the spectral signals triggered by intrusion behaviors in complex weather environments are easily submerged by the background noise, with unclear eigenvalue and prone to false alarms and missed alarms, the present invention proposes a multi-channel transmission fiber optic alarm system based on an edge large model.

[0008] The technical solution of the present invention is as follows: A multi-channel transmission fiber optic alarm system based on an edge large model, including:

[0009] A multi-source data fusion module: used to collect fiber optic vibration signals, environmental data, video surveillance data, and sound data, where the environmental data includes temperature, humidity, and wind speed;

[0010] A laser signal emission power dynamic adjustment module: including a laser, a U15 chip, and a main chip. The U15 chip changes the output power of the laser by dynamically adjusting the resistance, and the main chip has an edge large model built-in and interacts with the U15 chip through IIC signals to achieve automatic adjustment of the laser output power;

[0011] A multi-channel switching module: including 6 polarizers divided into 3 working domains, and each working domain consists of two orthogonal polarizers, used to filter polarized light in a specific direction to improve the sensitivity and accuracy of vibration signal detection;

[0012] An edge large model module: built into the main chip, used to analyze multi-channel data in real time, dynamically schedule the multi-channel switching frequency and laser power, and support self-supervised learning to update model parameters.

[0013] Preferably, the multi-source data fusion module includes:

[0014] A fiber optic vibration signal acquisition unit: used to collect data on the change of the interference light waveform generated by fiber optic vibration in real time;

[0015] An environmental data acquisition unit: used to collect meteorological data such as temperature, humidity, and wind speed, and transmit it to the main chip through a sensor network;

[0016] A video surveillance data acquisition unit: used to obtain the real-time video stream of the monitored area and extract the moving target features through image processing algorithms;

[0017] A sound data acquisition unit: used to collect environmental sound signals and extract abnormal sound features through spectral analysis;

[0018] A data fusion unit: used to align the above multi-source data according to the time stamp and generate a multi-dimensional feature vector as the input of the edge large model.

[0019] Preferably, the laser signal emission power dynamic adjustment module is connected to the DS_W0 and DS_L0 signal lines and is used to receive the power adjustment instruction sent by the main chip. Among them, the U15 digital potentiometer chip is connected to the main chip through the IIC bus and is used to dynamically adjust the resistance value according to the output of the edge large model. The laser signal emission power dynamic adjustment module monitors the working current of the laser in real time and feeds back the data to the edge large model, and calculates the optimal emission power according to the vibration signal eigenvalue and the environmental noise level.

[0020] Preferably, the multi-channel switching module includes: 6 high-precision polarizers, which are divided into 3 working domains, and each working domain consists of two polarizers with orthogonal polarization directions, and the polarizer combinations of different working domains can be quickly switched.

[0021] Preferably, the system initializes the laser emission power and the multi-channel working domain through the multi-source data fusion module, and realizes the rapid identification of intrusion behaviors and the adaptive adjustment under complex environments through the linkage of the edge large model module with the laser signal emission power dynamic adjustment module and the multi-channel switching module.

[0022] Preferably, the multi-source data fusion module also dynamically adjusts the initial laser emission power and the initial multi-channel working domain according to the environmental data to optimize the startup state of the system under complex environments.

[0023] Preferably, when the waveform characteristics of the detected intrusion behavior are not obvious, the edge large model increases the output power of the laser through the U15 chip to enhance the eigenvalue of the vibration event.

[0024] Preferably, polarizer one and polarizer four are orthogonal to form working domain one, polarizer two and polarizer five are orthogonal to form working domain two, and polarizer three and polarizer six are orthogonal to form working domain three. Each working domain corresponds to the sensitive direction of a specific intrusion behavior.

[0025] Preferably, the switching between multiple channels is carried out quickly. Each time the data of the switched working domain is sent to the fiber optic sensing AI large model for data analysis. When an abnormal energy waveform is detected while there is no abnormality in other passes, the switching mode is triggered to stop, and waveform analysis and data preprocessing are carried out for a certain frame rate in this working domain, and a certain channel in this working domain is quickly locked to quickly detect the intrusion behavior.

[0026] Preferably, the main chip of the system detection device has a large model built in at the edge side. Through the analysis ability of the large model at the edge side, automatic adjustment and scheduling are carried out with the laser emission signal adjustment circuit and the multi-channel switching module, so that the multi-channel switching module works in the optimal working range, the laser emission signal is self-adaptive, and the start and stop, switching frequency, and laser output signal intensity of multiple channels all come from the self-feedback of the model, so that the model can quickly identify the vibration event characteristics in the fiber optic sensing data.

[0027] Advantages of the present invention:

[0028] 1. Through the switching of multiple channels and polarization filtering, background noise can be effectively dispersed, thereby improving the signal-to-noise ratio in complex environments.

[0029] 2. The large model at the edge side is linked with the hardware to achieve self-adaptive optimization of laser power and channel switching, thereby enhancing the feature extraction ability.

[0030] 3. The system supports online update of the model, thereby continuously reducing the false alarm rate and missed alarm rate.

[0031] 4. By combining environmental and video data, a more comprehensive basis for intrusion behavior analysis is provided. Description of the Drawings

[0032] Figure 1 Shown is a schematic circuit diagram of the laser signal emission power dynamic adjustment module of the present invention;

[0033] Figure 2 Shown is the circuit schematic diagram of the laser signal emission power dynamic adjustment module of the present invention;

[0034] Figure 3 Shown is a schematic diagram of the multi-channel principle of the present invention;

[0035] Figure 4 Shown is a schematic diagram of the working process of the present invention.

[0036] Description of the reference numerals: 1, first polarizer; 2, second polarizer; 3, third polarizer; 4, fourth polarizer; 5, fifth polarizer; 6, sixth polarizer. Detailed Embodiments

[0037] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, rather than all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present invention.

[0038] The present invention provides an embodiment: a multi-channel transmission optical fiber alarm system based on an edge large model, including:

[0039] A multi-source data fusion module: used to collect optical fiber vibration signals, environmental data, video surveillance data, and sound data, where the environmental data includes temperature, humidity, and wind speed;

[0040] A laser signal transmission power dynamic adjustment module: includes a laser, a U15 chip, and a main chip. The U15 chip changes the output power of the laser by dynamically adjusting the resistance, and the main chip has an edge large model built-in and interacts with the U15 chip through IIC signals to achieve automatic adjustment of the laser output power;

[0041] A multi-channel switching module: includes 6 polarizers divided into 3 working domains, and each working domain consists of two orthogonal polarizers, used to filter polarized light in a specific direction to improve the sensitivity and accuracy of vibration signal detection;

[0042] An edge large model module: built into the main chip, used to analyze multi-channel data in real time, dynamically schedule the multi-channel switching frequency and laser power, and support self-supervised learning to update model parameters.

[0043] Please refer to Figure 1 and Figure 2 , in this embodiment, the transmission power of the laser signal can be adjusted. When the intrusion type cannot be quickly distinguished in detail, the transmission power can be adjusted to change the input waveform so that the optical fiber sensing large model can quickly distinguish the intrusion behavior type. The detailed principle is as follows: the output power is changed by changing the current of the laser. DS_W0 and DS_L0 are connected to the U15 chip, and the main function of the U15 chip is to dynamically adjust the resistance, thereby changing the laser output power; the U15 interacts with the main chip through IIC signals, and the main chip has an edge large model built-in and automatically adjusts in interaction with the U15 chip.

[0044] When a waveform similar to an intrusion behavior is detected and the characteristic values of the waveform are not very obvious, the output power can be adjusted to continue to distinguish the intrusion type and highlight the waveform characteristic values.

[0045] The system initializes the laser transmission power and the multi-channel working domain through the multi-source data fusion module, and realizes the rapid identification of intrusion behaviors and the adaptive adjustment in complex environments through the linkage of the edge large model module with the laser signal transmission power dynamic adjustment module and the multi-channel switching module.

[0046] Preferably, the multi-source data fusion module also dynamically adjusts the initial laser transmission power and the initial multi-channel working domain according to the environmental data to optimize the startup state of the system in complex environments.

[0047] Preferably, when the waveform features of the intrusion behavior are not obvious, the edge large model increases the output power of the laser through the U15 chip to enhance the eigenvalue of the vibration event.

[0048] Please refer to Figure 3 , in this embodiment, the multi-channel switching mainly uses a polarizer to filter out unnecessary polarization components, making the detection system only sensitive to the polarization light changes in a specific direction, thereby improving the sensitivity and accuracy of the fiber vibration measurement, helping to more accurately detect tiny vibration signals, that is, the intrusion behavior generates specific energy waveforms in a specific direction. The hardware includes 6 polarizers, divided into 3 working domains (channels 1-6 are pairwise orthogonal), each group corresponding to a specific polarization direction. Polarizer 1 (channel 1) and polarizer 4 (channel 4) are perpendicular to each other, forming working domain 1; polarizer 2 (channel 2) and polarizer 5 (channel 5) are perpendicular to each other, forming working domain 2; polarizer 3 (channel 3) and polarizer 6 (channel 6) are perpendicular to each other, forming working domain 3. The multi-channels switch quickly, and the data of each switched working domain is sent to the fiber sensing AI large model for data analysis. When an abnormal energy waveform is detected (no abnormality in other channels), the switching mode is triggered to stop, and waveform analysis and data preprocessing are performed at a certain frame rate in this working domain, and a certain channel in this working domain is quickly locked to quickly detect the intrusion behavior. Through polarization component filtering, multi-directional background noise is suppressed, and the signal-to-noise ratio is improved.

[0049] In complex scenarios, interference factors such as strong winds cause the energy to quickly accumulate due to the light speed transmission, and background noise is generated in multiple directions. Through the switching of multi-channels, interference factors can be blocked, the background noise can be reduced, and the signal-to-noise ratio of the signal can be improved, making the data waveform features generated by the intrusion behavior more obvious.

[0050] The working process of the multi-channel switching module is as follows:

[0051] The multi-channels are scanned at a preset frequency, and at the same time, the edge large model analyzes the data of each channel in real time.

[0052] When an abnormal waveform is detected, the system locks the corresponding working domain, stops scanning, and performs high-frame-rate waveform analysis.

[0053] The main chip of the detection device of this system has an edge large model built in. Through the analysis ability of the edge large model, automatic adjustment and scheduling are carried out with the laser emission signal adjustment circuit and the multi-channel switching module, enabling the multi-channel switching module to work in the best working domain and the laser emission signal to be adaptive. The start and stop, switching frequency, and laser output signal intensity of the multi-channels all come from the self-feedback of the model, so that the model can quickly identify the vibration event characteristics in the fiber sensing data, weaken the interference factors, reduce the background noise, and improve the signal intensity of the vibration event;

[0054] The multi-channel switching module initializes and scans at a specific frequency, and the model conducts corresponding inspections (the multi-channel module can directly weaken interference factors, making the system more sensitive to vibration characteristic signals). When abnormal data is detected in a certain channel, the model quickly stops scanning, switches to that channel for observation and analysis for a period of time, and focuses on feature extraction to analyze the type of intrusion behavior.

[0055] If an intrusion behavior is detected but the specific type of intrusion cannot be identified, the output signal intensity of the laser can be automatically adjusted to enhance the characteristic value of the vibration event.

[0056] This model can perform in-sensor memory, computing, and training integration. After false alarm data and missed alarm data are added with one key, the model conducts self-supervised learning and model update, and continues to adjust the parameters of the model, such as the learning rate, number of layers, number of neurons, etc., to automatically improve the accuracy of the system and reduce false alarms and missed alarms.

[0057] The present invention proposes a new system for multi-channel transmission fiber vibration intrusion alarm analysis based on an in-sensor large model. It can not only inherit the advantages of existing solutions but also overcome all the disadvantages of existing solutions. Through the linkage and automatic adjustment of the in-sensor large model, the laser emission signal adjustment circuit, and the multi-channel switching module, it can reduce the influence of background noise in complex environments, enhance the signal characteristic value of vibration events, improve the signal-to-noise ratio, and quickly and accurately detect the type of intrusion behavior.

[0058] Please refer to Figure 4 , the working process of the present invention is as follows:

[0059] For the initialization of multi-source data fusion, it is necessary to load environmental sensor data, such as temperature, humidity, and wind speed as reference parameters, initialize the laser emission power, and set the multi-channel scanning mode. For example, the initial frequency is set to 100 Hz, and 3 working domains are scanned cyclically.

[0060] The system starts to work. First, the laser emits coherent light to the sensing optical fiber, and the photodetector receives the interference light signal and converts it into an electrical signal. The signal is sent to the main chip for preprocessing, such as filtering and noise reduction, after being converted by the ADC. At the same time, the temperature and humidity sensor collects the temperature and humidity in the environment, the wind speed sensor detects the current environmental wind speed, the camera monitors the target area, and the microphone collects the environmental sound. The main chip receives the data from the temperature and humidity sensor, the wind speed sensor, the data collected by the camera and the microphone. At the same time, the laser outputs a signal and passes through the working area of the polarizer array. The polarizer array switches in a preset order (working domain 1→2→3→1…), and each working domain stays for 1 ms to ensure sufficient data sampling time. The in-sensor large model calculates the signal-to-noise ratio of each channel in real time, and the optical signal detects the vibration signal in the fence detection optical fiber after passing through the polarizer array;

[0061] After detecting the vibration signal, the signal is interferometrically adjusted, and the signal acquisition continues while providing initialization parameters for the AI fiber optic sensing large model;

[0062] The collected raw data needs to be preprocessed and feature-extracted for analysis by the AI model;

[0063] When an intrusion behavior is detected, a stop switching signal is sent to the multi-channel switch, and at the same time, the types of intrusion are started to be subdivided. If it can be distinguished, the alarm event is recorded, such as personnel climbing, digging, or vehicle rolling, etc.; if it cannot be distinguished, the emission power of the laser signal is dynamically adjusted, and the laser emits a signal output.

[0064] Among them, the false alarm data and the misreported data of this system will be self-supervised learning, and the model will be updated, while providing initialization parameters for the AI fiber optic sensing large model.

[0065] In view of the shortcomings of the existing solutions, the present invention proposes a system based on an edge large model and a hardware innovation circuit (laser emission power adjustment) and multi-channel module linkage analysis, which does not require human intervention, can perform adaptive adjustment and supervised learning, and quickly adapts to fiber optic intrusion detection in complex environments; by using the physical property of the interferometer to filter the polarization components to innovate the multi-channel switching module, the energy of the background noise is dispersed, the overall signal-to-noise ratio of the system is improved, and the detection of intrusion events by the system is faster and more sensitive; in parallel, through the adaptive adjustment and enhancement of the laser emission power, the characteristics of intrusion events are strengthened, and false positives and missed detections are reduced.

Claims

1. A multi-channel transmission optical fiber alarm system based on a large end-side model, characterized in that: Included are: Multi-source data fusion module: used to collect optical fiber vibration signals, environmental data, video monitoring data and sound data, the environmental data including temperature, humidity and wind speed; Laser signal transmission power dynamic adjustment module: includes laser, U15 chip and main chip. The U15 chip changes the laser output power by dynamically adjusting the resistance. The main chip has a built-in end-side large model and interacts with the U15 chip through IIC signals to achieve automatic adjustment of the laser output power. Multi-channel switching module: includes 6 polarizers divided into 3 working areas. Each working area consists of two orthogonal polarizers, which are used to filter polarized light in a specific direction to improve the sensitivity and accuracy of vibration signal detection. Large model module on the end: Built into the main chip, it is used to analyze multi-channel data in real time, dynamically schedule multi-channel switching frequency and laser power, and support self-supervised learning to update model parameters.

2. According to claim 1, a multi-channel transmission optical fiber alarm system based on a large end-side model is characterized in that: The multi-source data fusion module includes: Optical fiber vibration signal acquisition unit: used to collect the interference light waveform change data generated by optical fiber vibration in real time; Environmental data acquisition unit: used to collect temperature, humidity, wind speed meteorological data, and transmit them to the main chip through the sensor network; Video surveillance data acquisition unit: used to obtain real-time video stream of the surveillance area and extract moving target features through image processing algorithms; Sound data acquisition unit: used to collect environmental sound signals and extract abnormal sound features through spectrum analysis; Data fusion unit: used to align the above multi-source data by timestamp and generate a multi-dimensional feature vector as the input of the end-side large model.

3. The multi-channel transmission optical fiber alarm system based on the end-side large model according to claim 1 is characterized in that: The laser signal transmission power dynamic adjustment module is connected to the DS_W0 and DS_L0 signal lines, and is used to receive power adjustment instructions sent by the main chip. The U15 digital potentiometer chip is connected to the main chip through the IIC bus, and is used to dynamically adjust the resistance value according to the output of the end-side large model. The laser signal transmission power dynamic adjustment module monitors the laser operating current in real time and feeds back the data to the end-side large model, and calculates the optimal transmission power according to the vibration signal characteristic value and the environmental noise level.

4. The multi-channel transmission optical fiber alarm system based on the end-side large model according to claim 1 is characterized in that: The multi-channel switching module comprises: 6 high-precision polarizers, which are divided into 3 groups of working domains, each group of working domains is composed of two polarizers with orthogonal polarization directions, and the polarizer combinations of different working domains can be quickly switched.

5. The multi-channel transmission optical fiber alarm system based on the end-side large model according to claim 4 is characterized in that: The system initializes the laser emission power and multi-channel working domain through a multi-source data fusion module, and links the terminal-side large model module with the laser signal emission power dynamic adjustment module and the multi-channel switching module to achieve rapid identification of intrusion behavior and adaptive adjustment in complex environments.

6. The multi-channel transmission optical fiber alarm system based on the end-side large model according to claim 5 is characterized in that: The multi-source data fusion module also dynamically adjusts the initial emission power of the laser and the initial working domain of the multi-channel according to the environmental data to optimize the startup state of the system in a complex environment.

7. The multi-channel transmission optical fiber alarm system based on the end-side large model according to claim 6 is characterized in that: When it is detected that the waveform characteristics of the intrusion behavior are not obvious, the end-side large model increases the laser output power through the U15 chip to enhance the characteristic value of the vibration event.

8. The multi-channel transmission optical fiber alarm system based on the end-side large model according to claim 7 is characterized in that: The polarizer 1 (1) and the polarizer 4 (4) are orthogonally formed into a working domain 1, the polarizer 2 (2) and the polarizer 5 (5) are orthogonally formed into a working domain 2, and the polarizer 3 (3) and the polarizer 6 (6) are orthogonally formed into a working domain 3. Each group of working domains corresponds to a sensitive direction of a specific intrusion behavior.

9. The multi-channel transmission optical fiber alarm system based on the end-side large model according to claim 8 is characterized in that: Rapid switching between multiple channels. Each time the working domain is switched, the data is transmitted to the fiber optic sensor AI large model for data analysis. When an abnormal energy waveform is detected and other channels pass without abnormalities, the switching mode is triggered to stop, and waveform analysis and data preprocessing at a certain frame rate are performed in the working domain. A channel under the working domain is quickly locked to quickly detect intrusion behavior.

10. The multi-channel transmission optical fiber alarm system based on the end-side large model according to claim 9, characterized in that: The main chip of the detection equipment of this system has a built-in end-side large model. Through the analysis capability of the end-side large model and the laser emission signal adjustment circuit and multi-channel switching module, automatic adjustment and scheduling are performed to make the multi-channel switching module work in the optimal working domain and the laser emission signal self-adaptive. The start and stop of multiple channels, the switching frequency, and the laser output signal strength all come from the self-feedback of the model, so that the model can quickly identify the characteristics of vibration events in the optical fiber sensing data.

Images