Intelligent vehicle perimeter safety fiber sensing early warning method and system

Abstract

This invention discloses a fiber optic sensing and early warning method and system for intelligent vehicle perimeter safety, relating to the fields of intelligent transportation and vehicle safety technology. The method includes: acquiring fiber optic scattering signals and determining the spatial location of disturbances; continuously reconstructing the multi-point fiber optic scattering signals in space and time according to the spatial location of the disturbances to obtain a disturbance energy density field; constructing a fiber optic topology graph using the fiber optics deployed around the vehicle as nodes, local disturbance energy as node states, and spatial propagation intensity between nodes as edge weights; introducing physical constraints based on the disturbance energy density field as the optimization objective of the fiber optic topology graph to update the node states; and fusing the disturbance energy density field, vehicle operation information, and driver physiological signals, combined with the local disturbance energy of the fiber optics to obtain a disturbance event type identification result. By constructing a continuous fiber optic sensing network around the vehicle, real-time monitoring and identification of disturbances around the vehicle are achieved.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation and vehicle safety technology, and in particular to an intelligent vehicle perimeter safety fiber optic sensing early warning method and system. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] With the rapid development of intelligent transportation and autonomous driving technologies, vehicle perimeter environmental perception and safety protection have become crucial components of intelligent vehicle systems. Existing intelligent driving systems primarily rely on external sensors such as millimeter-wave radar, ultrasonic radar, lidar, and cameras for environmental perception. These sensors identify the distance, speed, and shape of objects around the vehicle, providing basic information for automatic parking, collision warning, and driver assistance. However, most of these sensing technologies are point-based or field-of-view limited monitoring modes, with intermittent and directional limitations in the perception area, making it difficult to achieve continuous spatial coverage around the vehicle. Furthermore, sensor signals are susceptible to factors such as weather, lighting, dust, and electromagnetic interference, leading to unstable detection accuracy and a high false alarm rate, thus limiting the reliability of intelligent vehicles in complex environments.

[0004] Traditional millimeter-wave radar offers high stability for mid-to-long-range target detection, but its ability to identify small obstacles, slow-moving objects, and non-metallic objects at close range (less than 0.3 m) is poor. Ultrasonic radar, while providing short-range ranging capabilities in low-speed parking environments, suffers from low resolution, weak angle perception, and susceptibility to reflection angles and noise. Cameras and lidar rely on visual features for target recognition, making them extremely sensitive to lighting conditions, with performance significantly degraded in rain, fog, at night, or in backlight. Furthermore, while multi-sensor fusion solutions can improve recognition accuracy to some extent, they also increase system complexity, cost, and energy consumption, and still cannot completely eliminate blind spots. These issues mean that traditional perception systems still face significant technical bottlenecks in near-field vehicle safety protection and all-weather operation.

[0005] In recent years, distributed optical fiber sensing technology has been widely used in pipeline leak detection, structural health monitoring, and boundary intrusion monitoring due to its characteristics such as resistance to electromagnetic interference, high sensitivity, and ability to perform continuous detection over long distances. The development of phase-sensitive optical time-domain reflectometry (φ-OTDR) and Brillouin scattering techniques has enabled optical fibers to achieve continuous acoustic and strain detection over their entire length, providing new approaches for vehicle perimeter monitoring.

[0006] However, applying fiber optic sensing technology to dynamic mobile platforms (such as automobiles) still faces several challenges: First, mechanical vibrations and environmental noise during vehicle operation can cause complex interference signals; second, the fiber optic deployment structure is significantly affected by the vehicle's shape, resulting in spatial non-uniformity of signal characteristics; and third, traditional signal processing algorithms struggle to separate effective disturbances from background noise in real time. How to achieve dynamic modeling, disturbance identification, and multimodal fusion of fiber optic signals has become a key scientific and engineering problem in this field.

[0007] In recent years, the introduction of artificial intelligence technology has enabled vehicle environmental perception to gradually develop towards intelligence and data-driven directions. Deep learning models (such as convolutional neural networks, recurrent neural networks, and Transformers) have achieved remarkable results in image and temporal signal recognition, but these models often rely on a large amount of labeled data, making it difficult to interpret complex physical processes, and they are prone to performance degradation in dynamic vehicle environments. Summary of the Invention

[0008] To address the aforementioned issues, this invention proposes an intelligent vehicle perimeter safety fiber optic sensing and early warning method and system. By constructing a continuous fiber optic sensing network around the vehicle, and through optical signal demodulation and intelligent algorithms, it achieves real-time monitoring and identification of pedestrians, obstacles, collisions, airflow, and environmental disturbances around the vehicle.

[0009] To achieve the above objectives, the present invention adopts the following technical solution:

[0010] In a first aspect, the present invention provides a fiber optic sensing early warning method for perimeter safety of intelligent vehicles, comprising:

[0011] Acquire the fiber optic scattering signal and determine the spatial location of the disturbance;

[0012] The multi-point fiber scattering signal is continuously spatiotemporally reconstructed according to the spatial location of the disturbance and the sampling time to obtain the disturbance energy density field;

[0013] Using the optical fibers deployed around the vehicle as nodes, the local disturbance energy as the node state, and the spatial propagation intensity between nodes as the edge weight, an optical fiber topology graph is constructed. Physical constraints are introduced based on the disturbance energy density field, which serve as the optimization objective of the optical fiber topology graph, thereby updating the node state.

[0014] After fusing the disturbance energy density field, vehicle operation information, and driver physiological signals, the disturbance event type identification result is obtained by combining the local disturbance energy of the optical fiber.

[0015] As an alternative implementation method, the perturbation energy density field is ;in: For mapping functions; To disturb spatial position The encoding; Sampling time The encoding.

[0016] As an alternative implementation method, the physical constraints are:

[0017] ;

[0018] Where: α is the signal attenuation coefficient; β is the disturbance propagation coefficient; η(z,t) is the external disturbance excitation; To perturb the energy density field, To disturb the spatial position, Sampling time.

[0019] As an alternative implementation, the process of updating the node state includes:

[0020] Let node Connecting edge sets The state of each node Represents local perturbation energy, edge weight Indicates spatial propagation intensity;

[0021] Define adaptive edge update: ;

[0022] The node status has been updated to: ;

[0023] The training objective for node state updates is: ;

[0024] in, For activation functions; For state update functions, The parameter is used to calculate the state change based on the weighted energy difference between the current node state and its neighboring nodes. For nodes With nodes Spatial distance between; for Time Node With nodes Spatial propagation intensity; for Time Node Local disturbance energy; For time intervals; for Time Node The local disturbance energy.

[0025] As an alternative implementation, the fusion process involves aligning and fusing the energy space of the fiber optic perturbation energy density field, including vehicle operation information such as IMU acceleration and vehicle acoustic waves, as well as the driver's physiological signals, using an energy alignment mechanism. ;in: For modal energy estimation operators; To unify energy standards; This represents the original signal of the m-th sensing mode.

[0026] As an alternative implementation method, based on the fusion result Local disturbance energy of optical fiber and the rate of change of local disturbance energy Construct the perturbation semantic vector as ;in, The classification weight matrix maps the fused feature vectors to the category space, and then uses an activation function. Output the probability distribution of each disturbance event category;

[0027] Output categories include Furthermore, an alarm is triggered when the location and direction of the disturbance meet the set danger conditions.

[0028] Simultaneously, through a self-evolutionary learning mechanism, perturbation samples are optimized online. The total loss function includes physical constraints, perturbation category cross-entropy, topological smoothing regularization, and multimodal energy consistency constraints.

[0029] Secondly, the present invention provides an intelligent vehicle perimeter safety fiber optic sensing and early warning system, comprising:

[0030] The acquisition module is configured to acquire fiber optic scattering signals and determine the spatial location of the disturbance;

[0031] The neural field modeling module is configured to continuously reconstruct the perturbation energy density field by taking the multi-point fiber scattering signal and performing spatiotemporal reconstruction based on the perturbation spatial location and sampling time.

[0032] The topology adaptive module is configured to construct an optical fiber topology graph using optical fibers deployed around the vehicle as nodes, local disturbance energy as node states, and spatial propagation intensity between nodes as edge weights. It introduces physical constraint terms based on the disturbance energy density field, which serve as the optimization objective of the optical fiber topology graph, thereby updating the node states.

[0033] The identification module is configured to fuse the disturbance energy density field, vehicle operation information and driver physiological signals, and then combine the local disturbance energy of the optical fiber to obtain the disturbance event type identification result.

[0034] Thirdly, the present invention provides an electronic device including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the method described in the first aspect.

[0035] Fourthly, the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in the first aspect.

[0036] Fifthly, the present invention provides a computer program product, including a computer program that, when executed by a processor, implements the method described in the first aspect.

[0037] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0038] 1. Achieving continuous spatial perception and high-precision early warning of vehicle perimeter. This invention utilizes distributed fiber optic sensing technology to construct a continuous perception loop at the outer edge of the vehicle. Through phase-sensitive light temporal reflectance, it achieves dynamic monitoring with millimeter-level spatial resolution, enabling real-time perception of disturbances such as pedestrian approach, object contact, and minor collisions within a range of 0–3 meters. This achieves blind-spot-free safety perception and high-precision risk positioning of the vehicle perimeter.

[0039] 2. A physically constrained intelligent self-learning algorithm enhances system stability and interpretability. A neural field-driven physical fusion self-evolutionary network is introduced, combined with optical-acoustic coupled partial differential equations for physical constraint modeling, enabling the algorithm to possess continuous spatiotemporal representation and physical consistency. Simultaneously, the system possesses online self-evolutionary learning capabilities, automatically adjusting topology and parameters under different vehicle structures and environmental conditions, thereby maintaining long-term robustness and model interpretability.

[0040] 3. Superior Multimodal Collaborative Recognition and Anti-interference Performance. This invention integrates fiber optic signals, inertial data, radar signals, and environmental sensor information through a multimodal energy alignment mechanism to achieve disturbance feature alignment and classification in a unified energy space, effectively distinguishing complex disturbance types such as wind noise, raindrops, and collisions. This mechanism enables the system to maintain a recognition accuracy rate of over 98% even in rain, fog, nighttime, and high-noise environments, significantly improving the active safety protection capabilities of intelligent vehicles under extreme conditions.

[0041] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0043] Figure 1 This is a flowchart of the intelligent vehicle perimeter safety fiber optic sensing early warning method provided in Embodiment 1 of the present invention;

[0044] Figure 2 This is a diagram of the optical fiber deployment scheme provided in Embodiment 1 of the present invention;

[0045] Figure 3 This is a schematic diagram of the intelligent vehicle perimeter safety fiber optic sensing early warning process provided in Embodiment 1 of the present invention;

[0046] Figure 4 This is a flowchart of data noise reduction provided in Embodiment 1 of the present invention. Detailed Implementation

[0047] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0048] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0049] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form as well. Furthermore, it should be understood that the terms “comprising” and “including”, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0050] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0051] Terminology Explanation:

[0052] 1. Distributed Fiber-Optic Sensing (DFOS).

[0053] Distributed fiber optic sensing refers to a sensing method that uses optical fiber as a continuous sensing medium to continuously measure and locate external disturbances (including vibration, strain, temperature, and sound waves) along the length of the fiber using optical time-domain reflectometry (OTDR) or phase-sensitive optical time-domain reflectometry. Unlike traditional point sensors, distributed fiber optic sensing can form hundreds or thousands of sensing units on a single fiber, offering advantages such as continuous spatial resolution, resistance to electromagnetic interference, flexible deployment, and all-weather monitoring.

[0054] 2. Neural Field (NF).

[0055] Neural fields are a deep learning modeling method based on implicit function representations, used to represent complex physical field or signal distributions in continuous spatial and temporal domains. Unlike traditional discrete networks (such as convolutional neural networks and recurrent neural networks), neural fields achieve a mapping from coordinate input to signal output through a continuously differentiable function, thereby predicting perturbation energy or physical quantities at arbitrary locations.

[0056] 3. Physics-Aware Neural Field Evolution Network (PNFEN).

[0057] The Physically Fusion Self-Evolving Network (PNFEN) is a spatiotemporal learning framework that integrates the physical properties of optical fibers with a deep neural model to achieve dynamic prediction and semantic recognition of perturbation signals. This network introduces acoustic-thermal coupled partial differential equations as physical constraints on the neural field, ensuring the physical consistency of the model output. Simultaneously, PNFEN possesses a self-evolving learning mechanism, which can continuously optimize performance through online incremental updates under different vehicle structures and environmental conditions. This model achieves an innovative architecture that integrates continuous spatial modeling, real-time dynamic evolution, and physical interpretability.

[0058] 4. Modal Energy Alignment (MEA).

[0059] The multimodal energy alignment mechanism is a feature matching method for information fusion between different sensing modes. This mechanism maps the outputs from fiber optic sensors, millimeter-wave radar, inertial measurement units (IMUs), and other sensors to an energy space. Through energy consistency constraints, it achieves alignment and fusion of different modal signals under the same energy reference. This mechanism enables the system to maintain consistency and high robustness in multi-source sensing even under complex interference environments.

[0060] Example 1

[0061] Traditional millimeter-wave radar, ultrasonic, and camera systems suffer from performance degradation in rain, fog, nighttime, or strong light conditions, and have limited detection accuracy for close-range targets, making it difficult to achieve all-weather, blind-spot-free safety monitoring. To address this, this embodiment proposes an intelligent vehicle perimeter safety fiber optic sensing and early warning method. By constructing a flexible fiber optic sensing loop and a continuous spatiotemporal modeling network, it achieves high-precision monitoring, identification, and early warning of disturbances in the vehicle's surrounding environment, significantly improving the vehicle's safety perception capability in complex traffic environments. This solves the problems of low recognition rate, large blind spots, susceptibility to interference, and lack of continuous spatial perception in existing vehicle perimeter monitoring systems under complex environments.

[0062] like Figure 1 As shown, it specifically includes:

[0063] Acquire the fiber optic scattering signal and determine the spatial location of the disturbance;

[0064] The multi-point fiber scattering signal is continuously spatiotemporally reconstructed according to the spatial location of the disturbance and the sampling time to obtain the disturbance energy density field;

[0065] Using the optical fibers deployed around the vehicle as nodes, the local disturbance energy as the node state, and the spatial propagation intensity between nodes as the edge weight, an optical fiber topology graph is constructed. Physical constraints are introduced based on the disturbance energy density field, which serve as the optimization objective of the optical fiber topology graph, thereby updating the node state.

[0066] After fusing the disturbance energy density field, vehicle operation information, and driver physiological signals, the disturbance event type identification result is obtained by combining the local disturbance energy of the optical fiber.

[0067] This embodiment, based on distributed fiber optic sensing technology and neural field modeling theory, constructs an intelligent perception system that can continuously characterize the energy field of vehicle perimeter disturbances. It possesses continuous spatial perception, physical constraint reasoning, and self-evolutionary learning capabilities. By constructing a continuous fiber optic sensing network around the vehicle, and through optical signal demodulation and intelligent algorithms, it achieves real-time monitoring and identification of pedestrians, obstacles, collisions, airflow, and environmental disturbances around the vehicle. This enables high-precision, low-latency, all-weather safety perception of the vehicle in complex environments, providing a new technical path for intelligent transportation and autonomous driving.

[0068] Specifically:

[0069] Fiber optic sensing units are deployed in a ring around the outer edge of the vehicle. Specifically, single or multiple fiber optic sensing links are deployed around the vehicle, a closed-loop fiber optic main chain is deployed at the outer edge of the vehicle, and built-in branches are deployed in key monitoring areas.

[0070] By utilizing optical time-domain reflectometry or phase-sensitive optical time-domain reflectometry, the distribution of continuous physical quantities along the length of the optical fiber can be monitored, enabling real-time sensing of external disturbances.

[0071] When external disturbances occur (such as pedestrians approaching, objects colliding, airflow disturbances, etc.), these disturbances will produce minute strain, temperature or sound pressure changes in the optical fiber, causing phase drift, frequency shift or scattering intensity changes in the optical signal in a local area. The Rayleigh scattering phase changes are collected and converted into a spatiotemporal signal sequence through the optical signal demodulation module.

[0072] Traditional deep learning relies on a large number of parameters and static feature mappings, making it difficult to adapt to the real-time decision-making needs of vehicles in dynamic environments, under multimodal disturbances, and with non-stationary characteristics of fiber optic signals. Therefore, after filtering and synchronization processing, the spatiotemporal signal sequence is processed using a constructed neural field-driven physical fusion self-evolving network system. This system integrates Implicit Neural Representation (INR) and Physically Informed Graph Neural Dynamics (PGND) to continuously model the fiber optic disturbance signal in spatiotemporal time and reconstruct the energy field, enabling accurate monitoring and early warning of the vehicle's surrounding environment. Unlike traditional point-based sensing or image detection, this embodiment constructs a distributed energy sensing layer through continuous optical fibers, endowing the vehicle with environmental perception capabilities.

[0073] Furthermore, an adaptive topological graph layer is used to dynamically adjust the connection relationships between fiber optic nodes to adapt to changes in vehicle attitude and installation differences. A multimodal energy alignment mechanism is used to fuse radar, IMU, and other sensor signals to achieve energy domain consistency and joint identification of multi-source information.

[0074] Ultimately, when a specific type of disturbance is detected and a potential risk is identified, the vehicle control unit can automatically trigger audible and visual alarms, braking assistance, or avoidance strategies to achieve active safety protection.

[0075] The following is a detailed explanation.

[0076] In this embodiment, flexible distributed fiber optic sensing links are deployed around the vehicle body (including the front bumper, side skirts, rear bumper, and roof edge) to construct a continuous fiber optic sensing loop. Combined with redundancy mechanisms, dual-loop deployment, multi-core fiber optic spare cores, and distributed OTDR monitoring technology, an integrated perimeter safety perception system encompassing "vehicle body-fiber optics-algorithm-early warning" is formed. Backscattered light provides continuous spatial detection capabilities with a positioning accuracy of 1 meter. Figure 2 As shown.

[0077] It mainly includes the following core modules:

[0078] 1. Fiber optic deployment scheme, including a ring main chain fiber deployed on the outer edge of the vehicle, auxiliary branch fibers deployed in key areas of the vehicle, a multi-level installation structure for fixing the fiber, and demodulation equipment installed inside the vehicle.

[0079] The fiber optic vehicle perimeter deployment employs a topology and coverage strategy of a ring main chain plus auxiliary branches. Specifically, a ring main chain (Outer Loop) is deployed along the outer edge of the vehicle body (front bumper → left side skirt → rear bumper → right side skirt → back to the front) to achieve continuous coverage around the vehicle. Built-in auxiliary branches are deployed in key monitoring areas (door handles, trunk edge, roof edge, rearview mirrors) to enhance localized sensitivity. Figure 2 As shown.

[0080] Specifically: Pre-defined attachment areas are established on the vehicle body structure (front bumper, side skirts, rear bumper, etc.) corresponding to the fiber optic deployment path. These areas are coated with automotive-grade weather-resistant two-component adhesive or pressure-sensitive adhesive fiber optic fixing tape. The fiber optic cable can be directly attached to the inner surface of the exterior trim to form a continuous bonding interface, ensuring stability even under high-speed driving, wind loads, and minor impacts. In areas such as rearview mirrors, door handles, door trim panels, and tailgate trim panels, the fiber optic deployment can be integrated with the existing component assembly structure (such as internal slots, the pressing surface on the back of trim strips, and pre-reserved fiber optic channels within the mold) for concealed fixing, keeping the wiring invisible. Fiber optic cables deployed from the vehicle's outer edge are introduced into the equipment compartment through a sealed sleeve, entering the fiber optic splice tray or fiber optic management module inside the chassis, and connecting to the optical interface of the demodulation equipment. Through this installation method, the demodulation equipment obtains a stable operating environment inside the vehicle and can make low-loss connections with the ring main chain and auxiliary branch fibers, thereby improving the system's demodulation accuracy and overall reliability.

[0081] 2. The distributed fiber optic sensing unit is responsible for converting external physical disturbances (vibration, sound waves, temperature, contact, airflow, etc.) into scattered signals distributed along the optical fiber. The optical fiber adopts phase-sensitive optical time-domain reflectometry technology, which has high spatial resolution and high sensitivity characteristics, and can realize real-time perception of disturbances within a range of 0–3 meters around the vehicle perimeter.

[0082] 3. The optical signal demodulation module consists of a laser source, an optical splitter, a photodetector, and an interferometric demodulator. It is housed in an equipment compartment within the vehicle's trunk, fixed to the vehicle body via elastic vibration damping supports, and connected to a ring main chain / auxiliary branch chain via fiber optic patch cords. The equipment compartment has a dustproof and waterproof structure and is equipped with EMI (Electromagnetic Interference) shielding and grounding.

[0083] Understandably, the installation location and fixing method can be adjusted according to different vehicle models (cars, commercial vehicles, buses, or special vehicles). For models with limited onboard space, the demodulator can be split into an optical submodule (near the optical fiber) and an electronic submodule (near the power / data link). The two are connected via short jumpers or opto-isolated interfaces. The optical submodule is placed in a prominent position near the outer ring to shorten the optical fiber length, while the electronic submodule is placed in a location that facilitates heat dissipation and grounding.

[0084] This module reconstructs the spatiotemporal perturbation distribution along the optical fiber by emitting laser pulses and acquiring echo signals, measuring the phase change of Rayleigh scattering. This process realizes the conversion from physical perturbation to digital signal and serves as the sensing entry point of the system.

[0085] 4. Data acquisition and synchronization module, responsible for time synchronization, range correction and filtering of fiber optic demodulated data, IMU data and environmental information (wind speed, temperature, etc.).

[0086] A unified clock signal (Time Trigger) ensures that all sensor channels are synchronized at the microsecond level, providing a precise time base for subsequent spatiotemporal modeling.

[0087] 5. Data preprocessing module (signal denoising).

[0088] The signals acquired by optical fibers on the vehicle platform consist of three main components: target disturbance signals (pedestrians, collisions, scratches, etc.), structural / dynamic noise generated by the vehicle itself (engine, drive, tire / road noise, aerodynamic noise, etc.), and environmental background noise (wind, rain, road vibration, electromagnetic or temperature drift, etc.). A high-performance denoising strategy, combined with physical priors and data-driven methods, is employed for hierarchical processing. This involves first performing engineered preprocessing, then introducing blind source separation and model-based adaptive denoising, and finally incorporating event detection and self-learning calibration.

[0089] 6. Neural field spatiotemporal modeling module (PNFEN core algorithm).

[0090] This module is the core intelligent computing unit, employing a neural field-driven physical fusion self-evolving network. It continuously models the spatiotemporal distribution of fiber optic signals based on implicit neural representations, while introducing optical-acoustic coupling equations as physical constraints to construct an energy field for vehicle perimeter disturbances. Through dynamic topology graph mechanisms and neural differential equations (Neural ODEs), it models disturbance propagation, achieving continuous spatial-real-time evolution of full-field perception.

[0091] 7. Multimodal fusion and semantic recognition module.

[0092] To enhance the system's environmental understanding capabilities, this module performs energy alignment and feature fusion between fiber optic signals and vehicle millimeter-wave radar, IMU, and even driver physiological signals (heart rate, skin conductance).

[0093] By employing an energy space alignment mechanism, multimodal fusion is achieved in a unified energy domain, thereby identifying different types of external events (such as pedestrian approach, collision, scraping, wind noise, raindrop impact, etc.).

[0094] 8. Early warning decision and control module.

[0095] When a potential dangerous disturbance (such as an approaching obstacle or an external collision) is detected and its location and orientation meet the dangerous conditions, the vehicle control unit (ECU) will automatically trigger an audible and visual alarm, brake assist, or automatic parking protection mechanism.

[0096] 9. System Communication and Interface Module: Communicates with the vehicle control system via CAN bus, Ethernet, or in-vehicle Ethernet interface, supporting data upload, event marking, and remote diagnostics. It also retains an OTA (Over-the-Air) interface, enabling online algorithm updates and self-evolving model training.

[0097] In this embodiment, the specific workflow is as follows: Figure 3 As shown, it mainly includes:

[0098] 1. Signal acquisition: The laser pulse propagates along the optical fiber, and external disturbances cause phase shifts, thereby collecting the scattered signals to form a time series.

[0099] Specifically:

[0100] When a laser pulse propagates along an optical fiber, the inhomogeneity of the fiber's internal microstructure causes some light to be backscattered. Different scattering mechanisms correspond to different measurable physical quantities: for Rayleigh scattering, the response quantities are vibration and strain, characterized by fast high-frequency response and suitability for acoustic detection; for Brillouin scattering, the response quantities are temperature and strain, characterized by the ability to achieve joint detection of two parameters; for Raman scattering, the response quantity is temperature, characterized by suitability for detecting environmental thermal changes.

[0101] In this embodiment, the φ-OTDR mechanism is adopted, which utilizes the phase-sensitive characteristics of Rayleigh scattering to detect minute strain changes along the fiber length direction by interferometry.

[0102] When an external disturbance occurs, the phase change of the scattered light satisfies:

[0103] (1);

[0104] Where n is the refractive index of the optical fiber; λ is the operating wavelength; and ΔL(z,t) is the local length variation.

[0105] The spatial location z of the disturbance is determined by measuring the time delay Δt of the echo signal.

[0106] (2);

[0107] Where c is the speed of light and n is the refractive index of the optical fiber.

[0108] By mapping the echo signals from different time periods to specific coordinates outside the vehicle, a mapping from the time domain to the spatial domain is achieved. For example, the fiber optic segment on the left side of the vehicle body corresponds to a coordinate interval. The right side corresponds When a sudden increase in signal energy is detected, the direction and location of the disturbance can be determined, achieving centimeter-level positioning.

[0109] 2. Signal preprocessing. For example... Figure 4 As shown, it specifically includes:

[0110] (1) Baseline and temperature compensation (de-temperature drift and slow-varying background noise).

[0111] The baseline for the Exponential Moving Average (EMA) is:

[0112] (3);

[0113] Then perform baseline deduction:

[0114] (4);

[0115] constant Related to the baseline update cycle, for example , It is the window length (number of frames).

[0116] If the vehicle temperature T[n] is measured (or the temperature is obtained from the Brillouin / Raman co-calculation), regression compensation can be performed: (5).

[0117] in, The baseline for the exponential moving average; The original fiber optic signal (or the signal after preliminary processing) is the input signal to be compensated. The standard deviation of noise; To remove the actual disturbance signal (i.e., the target signal) after temperature drift. This is the temperature compensation coefficient, representing the weight of the effect of temperature changes on the signal; To measure noise or other unmodeled random disturbances.

[0118] (2) Time-frequency transformation (STFT) and local normalization.

[0119] Short-time Fourier transform (window w[m]):

[0120] (6);

[0121] Where: H is the step size; This is the time frame index, representing the k-th time window of the current analysis; ω represents the angular frequency, corresponding to the frequency components, and is expressed in radians per second. The sampling point index within the window function, ranging from 00 to M. 1; The window function length, i.e., the number of sampling points contained in each time window, determines the frequency resolution.

[0122] Local normalization of time-frequency energy: can be achieved by subtracting the local average spectrum from the short-time energy ratio or the logarithmic spectrum.

[0123] (3) Bandpass time-frequency filtering (to remove vehicle noise).

[0124] Parallel bandpass channels are set up for common disturbance frequency bands, such as: pedestrians / footsteps 5–50 Hz; minor collisions and scratches 100–2000 Hz; mechanical vibrations (engine shaft) are usually in the 10–200 Hz range. Vehicle noise is filtered out by utilizing its concentration in the low-frequency or structural resonance regions, using finite impulse response (FIR) or zero-phase bidirectional filtering to eliminate phase distortion.

[0125] FIR bandpass filter convolution:

[0126] (7);

[0127] in: represents the bandpass filter coefficients; L represents the filter length.

[0128] Alternatively, to eliminate the phase distortion introduced by conventional filtering, a zero-phase bidirectional filter can be used. ; The input is the original optical fiber time-domain signal sequence; Let L be the coefficients of an FIR bandpass filter, satisfying (n<0 or n≥L); The time-reversed sequence of the filter; This is the discrete convolution operator; The output signal after zero-phase bidirectional filtering has a phase response of zero.

[0129] (4) Wavelet threshold denoising (removing high-frequency noise).

[0130] Perform Discrete Wavelet Transform (DWT) to obtain detail coefficients Apply a soft threshold to the detail coefficients:

[0131] (8).

[0132] VisuShrink threshold (common empirical values):

[0133] (9).

[0134] in, The median of the first-level detail coefficients can be used for estimation:

[0135] (10).

[0136] Denoising and reconstruction yield ;

[0137] Where d represents the wavelet detail coefficients, obtained from wavelet decomposition; These are the detail coefficients after thresholding, used to reconstruct the net signal; The threshold size; This is an estimate of the noise standard deviation.

[0138] 3. Neural field modeling: Perform continuous spatiotemporal reconstruction to obtain the energy field distribution.

[0139] The fiber optic signal is treated as a spatiotemporal perturbation field around the vehicle and modeled using implicit neural representation:

[0140] (11);

[0141] in: For the MLP-NEF (Neural Energy Field) model, the neural field parameter mapping function; High-frequency encoding for position and time; This represents the perturbation energy density field.

[0142] This model reconstructs the disturbance response at any point on the vehicle perimeter in a continuous manner, avoiding the resolution limitations caused by discrete sampling.

[0143] 4. Physical constraints and topology adaptation: Dynamically update the model structure to achieve physical consistency.

[0144] Combining Rayleigh scattering and Brillouin strain principles in distributed optical fibers, a physical constraint term is introduced:

[0145] (12);

[0146] Where: α is the signal attenuation coefficient (related to fiber loss); β is the disturbance propagation coefficient; η(z,t) is the external disturbance excitation (such as human / animal / vehicle vibration).

[0147] This equation is an acoustic-thermal coupled partial differential model, constraining the dynamic evolution of the network. By introducing a physical residual term, the training objective is:

[0148] (13).

[0149] Topology-adaptive layers capture dynamic topological changes in spatial perturbation propagation and boundary relationships. The fiber optic network surrounding the vehicle is discretized into nodes. Connecting edge sets The state of each node Represents local perturbation energy, edge weight Indicates the intensity of spatial propagation.

[0150] To capture dynamic topology changes, an adaptive edge update is defined:

[0151] (14);

[0152] Node status update:

[0153] (15);

[0154] in, The activation function (such as Sigmoid or Tanh) is used to map the original values ​​of the edge weights to a reasonable range (e.g., [0,1] or [-1,1]) to ensure the stability and interpretability of the topology. For learnable state update functions (such as neural networks), the parameters are: It is used to calculate the change in state based on the weighted energy difference between the current node state and the neighboring nodes; The spatial distance between node i and node j (the difference in length along the fiber optic path) reflects the influence of physical location on disturbance propagation.

[0155] This layer can adjust adjacent structures in real time to adapt to changes in vehicle posture, fiber optic distribution bending, or installation differences.

[0156] 5. Feature extraction and semantic recognition.

[0157] The temporal evolution of the perturbation is described using neural differential equations (Neural ODEs):

[0158] (16).

[0159] Achieving continuous-time inference using the ODE solver:

[0160] (17);

[0161] Where E represents environmental variables (vehicle speed, temperature, wind speed, etc.); Let be the system state vector at time t (usually the perturbation energy state of the fiber optic node or its high-dimensional characteristic representation).

[0162] This step first integrates the discrete data collected from the optical fiber into a continuous dynamic prediction, and then performs energy alignment on the collected multi-dimensional data to ensure that the optical fiber vibration responds to different energy spaces. This avoids discrete-time sampling errors and achieves continuous dynamic prediction.

[0163] Considering that a vehicle may simultaneously collect: fiber optic signal F(z,t); IMU acceleration a(t); vehicle body acoustic wave s(t); and driver physiological signal p(t); to fuse these signals, an energy alignment mechanism is proposed to achieve energy spatial alignment and fusion of fiber optic vibration energy with vehicle motion and physiological signals.

[0164] (18);

[0165] in: The modal energy estimation operator (calculated from the instantaneous power spectral density); To unify energy standards; This represents the original signal of the m-th sensing mode; The modal index represents the different sensor types or signal sources involved in the fusion.

[0166] This layer ensures the consistency of different modal signals in the energy space, enabling high-precision event judgment.

[0167] 6. Early warning output: Determine the event type and distance threshold to trigger an alarm signal or auxiliary control.

[0168] The output perturbation semantic vector is:

[0169] (19);

[0170] in, The rate of change of local disturbance energy at node i (i.e., the first derivative with respect to time) reflects the instantaneous trend of disturbance energy change and is used to capture the transient characteristics of dynamic events. The classification weight matrix maps the fused feature vectors to the category space, and then uses an activation function. Output the probability distribution for each category.

[0171] Output categories include:

[0172] (20).

[0173] Event semantic visualization is achieved through Energy-Topology Embedding (ETE), which has stronger interpretability and physical consistency compared to traditional classification networks.

[0174] 7. Self-learning and updating: Through the self-evolutionary learning (SERL) mechanism, new perturbation samples are optimized online to improve long-term robustness.

[0175] The total loss function consists of four parts:

[0176] (twenty one);

[0177] in: For physical constraints; Cross-entropy of the perturbation category; For topological smoothing regularization; This is a constraint on the consistency of multimodal energy.

[0178] The training strategy employs a self-evolutionary learning mechanism, enabling the system to continuously learn new disturbance samples during vehicle operation and achieve zero-sample adaptation.

[0179] In the above method, the sensing layer uses distributed optical fiber as the core to realize the perception of physical field distribution without blind spots; the perception layer constructs a continuous energy field through a neural field spatiotemporal network, integrating physical equations and data science; the decision layer performs real-time early warning, vehicle control linkage and self-evolution model update based on the energy semantic recognition results.

[0180] Example 1: Pedestrian approach recognition and early warning.

[0181] When pedestrians approach the perimeter of a vehicle, the movement of their footsteps and clothing creates low-frequency disturbances (5–50 Hz) in the air. These micro-vibrations act on the vehicle's outer shell and trigger phase changes via fiber optic sensing links.

[0182] The detected fiber optic disturbance signal is represented as follows:

[0183] ;

[0184] in: The fiber optic length is slightly disturbed due to pedestrians approaching.

[0185] Phase change rate obtained by optical signal demodulation The local energy density field is obtained through energy calculation: ;

[0186] The energy field is reconstructed into a continuous distribution using a neural field model: Automatically identify the spatial propagation pattern and energy distribution characteristics of the disturbance.

[0187] When the energy distribution is continuous, the spectrum is concentrated, and it gradually approaches the vehicle, the output is based on classification: ,determination =Pedestrians approach;

[0188] If the location of the disturbance center is detected The corresponding distance is less than the safety threshold. If this occurs, an audible and visual alarm or a braking assist signal will be triggered immediately.

[0189] This example demonstrates the application in low-frequency dynamic disturbance identification, achieving highly sensitive detection of pedestrian approach through continuous energy field reconstruction.

[0190] Example 2: Minor vehicle collision recognition and location.

[0191] When an external object makes slight contact or scratches the vehicle surface, a high-frequency transient strain wave is generated in a local fiber optic segment. Its dynamic propagation satisfies the fiber optic acoustic-thermal coupling partial differential equation:

[0192] ;

[0193] in, Indicates the collision excitation source, These are the signal attenuation and diffusion coefficients, respectively.

[0194] The neural network is trained by applying physical residual constraints to equation (5). This maintains the physical consistency of the model.

[0195] Real-time updates of fiber optic node status: This process reflects the propagation and energy attenuation of disturbances in the optical fiber.

[0196] When a high-frequency transient energy peak is detected and the states of neighboring nodes satisfy spatial consistency, the collision category is automatically output: Simultaneously through peak propagation time difference Determine the collision location: .

[0197] This example demonstrates its application in high-frequency transient event recognition and spatial localization, enabling the determination of collision point locations with centimeter-level accuracy.

[0198] Example 3: Identification of wind noise and raindrop disturbance (distinguishing complex backgrounds).

[0199] When a vehicle is in motion or during heavy rain, external air disturbances and raindrop impacts generate complex mixed signals on the optical fiber. These signals are characterized by a wide spectrum, dispersed energy, and short duration.

[0200] First, the fiber optic signal is aligned using a multimode energy alignment mechanism. With vehicle-mounted IMU acceleration Microphone sound pressure Wait for energy domain fusion After alignment, a unified energy description benchmark is formed. .

[0201] Output perturbation semantic vector: ;

[0202] Based on the stability coefficient of the disturbance energy distribution Determine the event type:

[0203] set up If the energy distribution exhibits a random short-pulse pattern, it is determined to be a raindrop disturbance;

[0204] set up If the spectrum is concentrated in the 500–2000 Hz range, it is identified as wind noise disturbance.

[0205] This process demonstrates the application of multimodal energy alignment and perturbation classification in complex backgrounds, achieving stable identification of the system under strong interference environments.

[0206] Example 4: Fiber optic deployment scheme in vehicles (installation, protection and redundancy design).

[0207] (a) Deployment principles.

[0208] Distributed fiber optic sensing units are deployed around the vehicle perimeter to couple with vehicle body structural vibrations via optical fibers, enabling real-time sensing of events such as approach, contact, scratches, and impacts. The vehicle environment differs from fixed environments, exhibiting characteristics such as complex structure, severe vibrations, limited space, and high maintenance frequency. Therefore, fiber optic deployment must adhere to the following general principles:

[0209] 1. Continuous coverage principle: Fiber optic cabling should ensure that a closed or quasi-closed detection loop is formed around the vehicle to avoid monitoring blind spots.

[0210] 2. Structural Adaptation Principle: The layout path should be combined with the vehicle body material, interior and exterior structure, opening and closing mechanism and vehicle wiring harness routing to reduce stress concentration and fiber breakage risk.

[0211] 3. Engineering maintainability principle: Redundant optical fibers are reserved in parts that require maintenance, such as doors, trunk, and engine compartment, to ensure that maintenance can be completed without disassembling large components.

[0212] 4. Redundancy and reliability principle: Critical links should be equipped with backup channels or parallel fiber cores to improve reliability under conditions of vibration, external impact and environmental aging.

[0213] (ii) Fiber optic installation and protection.

[0214] Based on the main link, local branches can be added for specific parts of the vehicle to enhance the monitoring precision of key areas. Typical branches include areas where frequent contact events occur, such as door edges and door handles. Deploying short secondary fiber optic paths at these locations can improve local spatial resolution and signal sensitivity. In high-incidence areas such as door panels, two parallel fibers can be used to form local redundancy, enhancing identification, localization, and anti-interference capabilities, improving positioning accuracy, and reducing the probability of false positives.

[0215] During installation, the optical fiber should utilize the vehicle's existing wiring harness channels, and be fixed and protected in accordance with the vehicle's structural characteristics. Bending-resistant single-mode optical fiber with a flexible sheath should be used, taking into account the fiber's bending radius limitations to avoid micro-bending loss or mechanical fatigue. The internal components of the vehicle body are fixed using adhesive points and clips, while external areas are protected with flexible metal hoses or wear-resistant conduits.

[0216] In opening and closing structures such as car doors and trunks, the reliability of fiber optic deployment is crucial. Stress relief structures should be installed near hinges, and fiber optic connection points between opening and closing components should be located in areas of minimal stress variation. Regarding environmental adaptability, protection designs incorporating IP67-rated connectors and other features should be implemented to ensure long-term stable operation of the fiber optics in various environments.

[0217] (III) Redundancy design and health monitoring.

[0218] To enhance the long-term stability of the system, this solution recommends employing redundancy mechanisms in the fiber optic network, including dual-loop deployment and spare fiber cores in multi-core optical cables. Dual-loop redundancy allows the backup path to take over when the main loop is unavailable, thus preventing overall system failure. Spare fiber cores in multi-core optical cables can quickly restore operation through re-splicing in the event of localized damage. Furthermore, the system can utilize distributed OTDR technology for fiber health monitoring during operation, including attenuation changes, splice loss, and breakpoint location detection, providing maintenance personnel with timely warnings and location information. Continuous monitoring of fiber health status allows for early identification of potential risks, reducing monitoring gaps caused by fiber breaks.

[0219] The method described in this embodiment has the following advantages: the fiber optic sensing structure enables continuous monitoring around the vehicle, with a detection range covering 0–3 meters and a spatial resolution down to the centimeter level; the denoising algorithm is based on three types of noise denoising, analyzing the spectral characteristics of the target disturbance signal, extracting the main frequency components and comparing them with the spectra of the other two types of noise, and adjusting parameters using adaptive filtering technology to retain the target signal and suppress interference; the neural field algorithm introduces physical constraints and dynamic topology mechanisms, giving the model high robustness and physical interpretability; multimodal energy fusion improves the collaborative accuracy between different sensors, enabling the system to maintain stable recognition even in rain, fog, nighttime, and complex traffic scenarios. A novel fiber optic deployment design is proposed, which, while meeting the requirements of sensitivity, spatial resolution, and mechanical reliability, also considers maintainability and vehicle engineering constraints. Compared with traditional point-based sensing schemes, this invention has a simple structure, low cost, strong anti-interference capability, and can be updated online through self-learning, effectively constructing a vehicle's "perception skin" and providing new environmental perception and risk warning capabilities for intelligent driving and active safety systems.

[0220] Example 2

[0221] This embodiment provides an intelligent vehicle perimeter safety fiber optic sensing and early warning system, including:

[0222] The acquisition module is configured to acquire fiber optic scattering signals and determine the spatial location of the disturbance;

[0223] The neural field modeling module is configured to continuously reconstruct the perturbation energy density field by taking the multi-point fiber scattering signal and performing spatiotemporal reconstruction based on the perturbation spatial location and sampling time.

[0224] The topology adaptive module is configured to construct an optical fiber topology graph using optical fibers deployed around the vehicle as nodes, local disturbance energy as node states, and spatial propagation intensity between nodes as edge weights. It introduces physical constraint terms based on the disturbance energy density field, which serve as the optimization objective of the optical fiber topology graph, thereby updating the node states.

[0225] The identification module is configured to fuse the disturbance energy density field, vehicle operation information and driver physiological signals, and then combine the local disturbance energy of the optical fiber to obtain the disturbance event type identification result.

[0226] It should be noted that the above modules correspond to the steps described in Embodiment 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should also be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0227] In further embodiments, the following is also provided:

[0228] An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the method described in Embodiment 1. For brevity, further details are omitted here.

[0229] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0230] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0231] A computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the method described in Embodiment 1.

[0232] The method in Example 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0233] A computer program product includes a computer program that, when executed by a processor, implements the method described in Embodiment 1.

[0234] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer-executable instructions, such as instructions included in program modules, which execute in a device on a target real or virtual processor to perform the processes / methods described above. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc., that perform specific tasks or implement specific abstract data types. In various embodiments, the functionality of program modules can be combined or divided among program modules as needed. The machine-executable instructions for the program modules can execute within a local or distributed device. In a distributed device, the program modules can reside in both local and remote storage media.

[0235] The computer program code used to implement the methods of the present invention may be written in one or more programming languages. This computer program code may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the computer or other programmable data processing device, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a computer, partially on a computer, as a stand-alone software package, partially on a computer and partially on a remote computer, or entirely on a remote computer or server.

[0236] In the context of this invention, computer program code or related data may be carried by any suitable carrier to enable a device, apparatus, or processor to perform the various processes and operations described above. Examples of carriers include signals, computer-readable media, and the like. Examples of signals may include electrical, optical, radio, sound, or other forms of propagation signals, such as carrier waves, infrared signals, etc.

[0237] Those skilled in the art will recognize that the units and algorithm steps described in connection with the various examples of this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0238] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A fiber optic sensing and early warning method for perimeter safety of intelligent vehicles, characterized in that, include: Acquire the fiber optic scattering signal and determine the spatial location of the disturbance; The multi-point fiber scattering signal is continuously spatiotemporally reconstructed according to the spatial location of the disturbance and the sampling time to obtain the disturbance energy density field; Using the optical fibers deployed around the vehicle as nodes, the local disturbance energy as the node state, and the spatial propagation intensity between nodes as the edge weight, an optical fiber topology graph is constructed. Physical constraints are introduced based on the disturbance energy density field, and these physical constraints are used as the optimization objective of the optical fiber topology graph to update the node state. The physical constraint terms are: ； Where: α is the signal attenuation coefficient; β is the disturbance propagation coefficient; η(z,t) is the external disturbance excitation; To perturb the energy density field, To disturb the spatial position, Sampling time; The process of updating the node status includes: Let node Connecting edge sets The state of each node Represents local perturbation energy, edge weight Indicates spatial propagation intensity; Define adaptive edge update: ; The node status has been updated to: ; The training objective for node state updates is: ; in, For activation functions; For state update functions, The parameter is used to calculate the state change based on the weighted energy difference between the current node state and its neighboring nodes. For nodes With nodes Spatial distance between; for Time Node With nodes Spatial propagation intensity; for Time Node Local disturbance energy; For time intervals; for Time Node Local disturbance energy; After fusing the disturbance energy density field, vehicle operation information, and driver physiological signals, the disturbance event type identification result is obtained by combining the local disturbance energy of the optical fiber.

2. The intelligent vehicle perimeter safety fiber optic sensing early warning method as described in claim 1, characterized in that, The perturbation energy density field is ;in: For mapping functions; To disturb spatial position The encoding; Sampling time The encoding.

3. The intelligent vehicle perimeter safety fiber optic sensing early warning method as described in claim 1, characterized in that, The fusion process involves using an energy alignment mechanism to align and fuse the energy space of the fiber optic perturbation energy density field, including vehicle operation information such as IMU acceleration and vehicle acoustic waves, as well as the driver's physiological signals. ;in: For modal energy estimation operators; To unify energy standards; This represents the original signal of the m-th sensing mode.

4. The intelligent vehicle perimeter safety fiber optic sensing early warning method as described in claim 1, characterized in that, Based on the fusion results Local disturbance energy of optical fiber and the rate of change of local disturbance energy Construct the perturbation semantic vector as ;in, The classification weight matrix maps the fused feature vectors to the category space, and then uses an activation function. Output the probability distribution of each disturbance event category; Output categories include Furthermore, an alarm is triggered when the location and direction of the disturbance meet the set danger conditions. Simultaneously, through a self-evolutionary learning mechanism, perturbation samples are optimized online. The total loss function includes physical constraints, perturbation category cross-entropy, topological smoothing regularization, and multimodal energy consistency constraints.

5. A smart vehicle perimeter safety fiber optic sensing and early warning system, characterized in that, include: The acquisition module is configured to acquire fiber optic scattering signals and determine the spatial location of the disturbance; The neural field modeling module is configured to continuously reconstruct the perturbation energy density field by taking the multi-point fiber scattering signal and performing spatiotemporal reconstruction based on the perturbation spatial location and sampling time. The topology adaptive module is configured to construct an optical fiber topology graph using optical fibers deployed around the vehicle as nodes, local disturbance energy as node states, and spatial propagation intensity between nodes as edge weights. It introduces physical constraint terms based on the disturbance energy density field and uses these physical constraint terms as the optimization objective of the optical fiber topology graph to update the node states. The physical constraint terms are: ； Where: α is the signal attenuation coefficient; β is the disturbance propagation coefficient; η(z,t) is the external disturbance excitation; To perturb the energy density field, To disturb the spatial position, Sampling time; The process of updating the node status includes: Let node Connecting edge sets The state of each node Represents local perturbation energy, edge weight Indicates spatial propagation intensity; Define adaptive edge update: ; The node status has been updated to: ; The training objective for node state updates is: ; in, For activation functions; For state update functions, The parameter is used to calculate the state change based on the weighted energy difference between the current node state and its neighboring nodes. For nodes With nodes Spatial distance between; for Time Node With nodes Spatial propagation intensity; for Time Node Local disturbance energy; For time intervals; for Time Node Local disturbance energy; The identification module is configured to fuse the disturbance energy density field, vehicle operation information and driver physiological signals, and then combine the local disturbance energy of the optical fiber to obtain the disturbance event type identification result.

6. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the method according to any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, perform the method described in any one of claims 1-4.

8. A computer program product, characterized in that, Includes a computer program, which, when executed by a processor, implements the method described in any one of claims 1-4.

Images