Anti-interference method of mobile fire detection system

Abstract

The invention discloses an anti-interference method of a mobile fire detection system, and belongs to the technical field of fire detection. A comprehensive solution integrating active vibration compensation, dynamic noise cancellation and sub-pixel accurate positioning is provided. Through vibration detection and real-time reverse driving compensation, the offset of the optical system is controlled within 0.02 mm, and the over-distance positioning precision of 0.3 m outside 150 m is achieved; by constructing a noise feature library capable of being dynamically updated and combining real-time noise matching and a reverse counteracting algorithm, vibration and electromagnetic interference are effectively suppressed, the noise reduction effect is improved by more than 30dB, and signal fluctuation is reduced to 0.05 V; through narrowband filtering and a temperature and humidity compensation mechanism, a real fire source is distinguished from vehicle high temperature, illumination and other interferences, and the monthly average false alarm frequency is reduced from more than 5 times to less than 1 times. The comprehensive performance of the flame detection system under severe working conditions is improved, and reliable technical support is provided for early accurate detection and positioning of highway tunnel fires.

Description

Technical Field

[0001] This invention belongs to the field of fire detection technology, specifically an anti-interference method for a mobile fire detection system. Background Technology

[0002] Highway tunnels, as key nodes in transportation infrastructure, are characterized by their long and narrow spaces (single tunnel length often exceeds 1km), enclosed environment (slow smoke and heat diffusion, low visibility), dense traffic flow (daily traffic volume exceeding 10,000 vehicles), and complex sources of interference (vehicle vibration / exhaust heat / lighting lights), which place stringent requirements on flame detection systems. (1) Precise positioning is required at a distance of more than 100m: Tunnel fires spread rapidly (5-10m per minute in a closed environment), and the specific station number and lane location of the fire source must be determined within a distance of more than 100m. Otherwise, rescuers will have to check each section and delay the golden opportunity to extinguish the fire (for every minute of delay, the fire loss increases by 30%). (2) Strong seismic resistance is required: the passage of heavy trucks generates low-frequency vibrations of 10Hz-50Hz (amplitude 0.1mm-0.5mm), and the resonance of the tunnel structure generates vibrations of 20Hz-30Hz (amplitude up to 0.8mm), which can easily lead to optical offset of the detector, increased signal noise, and affect the detection accuracy. (3) Low false alarm rate is required: Interference sources such as vehicle exhaust pipes (800℃-1000℃, strong infrared radiation), brake pad overheating (2μm-2.5μm infrared wavelength), sodium lamp / LED lighting (visible light-near infrared band) in the tunnel are prone to causing false alarms of traditional detectors (average false alarms ≥5 times per month), which leads to frequent tunnel closures and affects traffic efficiency.

[0003] In existing technologies, traditional point-type infrared flame detectors have three major drawbacks: (1) Lack of positioning capability: It can only output an alarm signal of "fire in a certain area", but cannot determine the fire source pile number and lane, and cannot meet the precise guidance needs of tunnel rescue; (2) Weak seismic performance: There is no special vibration reduction design. Under vibration environment, the optical coaxiality deviation exceeds 0.15mm, the positioning error expands to more than 1m, and the signal-to-noise ratio drops to below 22dB; (3) Poor anti-interference ability: It relies on a single infrared radiation intensity to determine the source of fire and cannot distinguish between mobile fire sources and vehicle high temperature / lighting interference. The false alarm rate is as high as 5 times / month, and there is no adaptation design for tunnel temperature and humidity and electromagnetic environment. Summary of the Invention

[0004] To solve the above problems, the technical solution provided by the present invention is as follows: An anti-interference method for a mobile fire detection system according to the present invention includes: The detection matrix module collects noise feature data and establishes a noise feature library, and configures a dynamic update strategy for the noise feature library; Noise reduction is achieved based on a noise feature library and a real-time noise cancellation algorithm. The real-time noise cancellation algorithm includes the following steps: Noise matching; Inverse noise generation; Superimposed cancellation noise reduction; The noise-reduced signal is sent to the data processing module, and the data processing flow of the data processing module is as follows: S100, Raw data preprocessing; S200 and FFT transforms are used to extract vibration features; S300, optical offset calculation; S400, Generate compensation and correction strategy; S500, drive adjustment reverse compensation; The compensated and offset signals are then used for fire source identification and lane location, specifically... S1100, Signal preprocessing, performs mean filtering on the signal to eliminate high-frequency noise; S1200, peak detection, identifies 2-3 adjacent units (continuous units with signal values ​​higher than the threshold) covered by fire source imaging. S1300, Polynomial fitting, using the least squares method to solve for the coefficients a, b, c, d of the cubic polynomial; S1400, extreme value calculation, solving the derivative equation to obtain sub-pixel level center coordinates; S1500, coordinate transformation, outputs actual location information according to the preset "array unit coordinates-tunnel station number-lane position" mapping table.

[0005] Preferably, the noise feature data collected is specifically: Under typical tunnel conditions, vibration data of 10Hz-100Hz are synchronously collected by the ADXL345 sensor in the detection matrix module. The initial sample set is established by continuous sampling for 100 hours. The typical tunnel conditions include empty / full-load truck passage, tunnel ventilation fan start-up, and traffic accident impact.

[0006] Preferably, establishing the noise feature database specifically involves: The samples were classified and labeled as follows: 10-30Hz low-frequency vibration, 30-60Hz mid-frequency vibration, and 60-100Hz high-frequency vibration. For each vibration type, five core features are extracted: frequency f, amplitude A, waveform factor type, duration T, and phase φ, and stored as a feature vector [f, A, type, T, φ], thereby establishing a noise feature library; The dynamic update strategy for configuring the noise feature library is specifically as follows: Every 24 hours, newly acquired vibration data are automatically clustered. If more than 3 unmatched features are found, a new feature entry is added. The requirement for unmatched features is a deviation >30%.

[0007] Preferably, the noise matching specifically involves performing an FFT transform on the currently acquired signal, extracting the real-time feature vector [f_real, A_real, type_real, T_real, φ_real], calculating the similarity with the feature library entries using Euclidean distance, and selecting the entry with the highest matching degree as the noise source.

[0008] Preferably, the reverse noise generation specifically involves generating a cancellation signal with the same amplitude but opposite phase as the original noise based on the noise feature with the highest matching degree, and converting the digital signal into an analog voltage signal through a 16-bit DAC chip.

[0009] Preferably, the reverse noise generation specifically involves using a differential amplifier to superimpose the original signal and the cancellation signal to achieve noise reduction; it also includes temperature and humidity compensation, collecting temperature data T and humidity data H under the current environment, and comparing them with set thresholds of 25℃ and 50% humidity. The compensation coefficient calculation process is as follows: K_T = 1 + (T - 25) × 0.005; K_H = 1 + (H - 50) × 0.0015; The signal strength generated by the reverse noise is adjusted according to the compensation coefficient.

[0010] Preferably, step S100 includes S110, Data Acquisition: The vibration detection component (190) acquires X / Y / Z axis acceleration data at a sampling rate of 1kHz, acquiring 1024 points at a time to obtain acceleration sequences a_x[n], a_y[n], and a_z[n]. S120, Unit Conversion: Receive acceleration data sent by the vibration detection component (190) and convert the acceleration value from g to m / s²; S130, Remove DC component: Calculate the mean of the sequence and subtract it; S140, Filtering: A Butterworth low-pass filter is applied with a cutoff frequency of 100Hz, retaining the effective vibration frequency band of 10-100Hz.

[0011] Preferably, step S200 includes FFT calculation: Perform a Fast Fourier Transform on the preprocessed acceleration sequence: X(k) = FFT(a_x_dc[n]), which yields the frequency domain amplitude spectrum |X(k)| and phase spectrum ∠X(k); Frequency identification: Find the frequency f_x corresponding to the peak amplitude in the range of 10-100Hz, and similarly obtain f_y; Amplitude calculation: converting frequency domain amplitude into actual vibration displacement amplitude; Displacement amplitude formula: A_x = (a_x_peak) / (4π²f_x² × 1000), Where a_x_peak is the peak acceleration (m / s²), and ×1000 converts meters to millimeters; Phase extraction: Record the phases φ_x and φ_y corresponding to the main vibration frequency to determine the vibration direction.

[0012] Preferably, step S300 includes The vibration displacement is converted into optical system offset by a pre-calibrated transfer function: Δx = A_x × K_x, Δy = A_y × K_y; where K_x and K_y are calibration coefficients, which are obtained by actual measurement using a laser interferometer. Calculate the real-time offset output: Combine the phase information to determine the offset direction (positive / negative), and finally output Δx and Δy containing positive and negative information; Specifically, step S400 involves converting the positive and negative information Δx and Δy into the corresponding compensation amount voltage of the corresponding adjustment driver (250) and sending it to the corresponding adjustment driver (250).

[0013] Preferably, step S1300 specifically involves: Let the coordinates of the detection units be x0, x1, x2 (distributed along the X-axis), and the corresponding signal values ​​be y0, y1, y2. 2， The expression for a cubic polynomial is: y = ax³ + bx² + cx + d; Use curve_fit for least squares fitting, where p0 is the initial parameter guess value; p0 = [0, -1, 0, max(y)] Assume the curve opens downwards and the vertex is close to the maximum signal value; popt, _ = curve_fit(self.cubic_polynomial, x, y, p0=[0, -1, 0, max(y)]); Returns the fitted coefficients [a, b, c, d]; The specific steps S1400 are as follows: S1410. Fit a cubic polynomial and obtain the coefficients a, b, c, d. a, b, c, d = self.fit_polynomial(x_units, signals); S1420, Solve for the point where the derivative is 0. First derivative: y' = 3ax² + 2bx + c Let y' = 0, solve the quadratic equation: 3ax² + 2bx + c = 0 discriminant = (2 b) twenty four (3 a) c, the discriminant Δ = (2b)² - 4 (3a) c If the discriminant is less than 0, there are no real roots, indicating that the curve has no obvious peak. In this case, the average coordinates of the covering unit are used as the center. S1430. Calculate the extreme points and filter out valid solutions; S1440. Select the extreme point with the largest signal value, which corresponds to the sub-pixel center coordinates.

[0014] Compared with the prior art, the technical solution provided by this invention has the following advantages: This invention presents an anti-interference method for a mobile fire detection system. Addressing the three major shortcomings of traditional point-type infrared flame detectors in tunnel environments—namely, lack of positioning capability, weak anti-vibration performance, and high false alarm rate—this invention proposes a comprehensive solution integrating active vibration compensation, dynamic noise cancellation, and sub-pixel precise positioning. Through vibration detection and real-time reverse drive compensation, the optical system offset is controlled within 0.02mm, achieving a long-range positioning accuracy of 0.3 meters at a distance of 150 meters. By constructing a dynamically updated noise feature library, combined with real-time noise matching and reverse cancellation algorithms, vibration and electromagnetic interference are effectively suppressed, improving noise reduction by more than 30dB and reducing signal fluctuation to 0.05V. Through narrowband filtering and temperature and humidity compensation mechanisms, the invention significantly distinguishes between real fire sources and interference from vehicle temperatures and lighting, reducing the average monthly false alarm frequency from more than 5 times to less than 1 time. This invention fundamentally improves the comprehensive performance of flame detection systems under harsh conditions such as vibration, high temperature, and high humidity, providing reliable technical support for the early and accurate detection and positioning of fires in highway tunnels. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the lens module of the present invention; Figure 2This is a cross-sectional view of the detection matrix module of the present invention; Figure 3 This is a top view of the detection matrix module of the present invention; Figure 4 This is a schematic diagram of the detector housing of the present invention in its installed state; Figure 5 This is a schematic diagram of the detector housing of the present invention.

[0016] Explanation of the labels in the diagram: 100. Lens module; 110. Housing; 120. Miniature heating element; 130. Narrowband filter; 140. Temperature sensor; 150. Laser positioning calibrator; 160. Dustproof glass; 170. Auxiliary collimating lens; 180. Main lens; 190. Vibration detection assembly; 200, Detection matrix module; 210, Buffer layer; 220, Sensing unit; 230, Ceramic substrate; 240, Signal output layer; 250, Adjustment driver; 310. Detector housing; 320. Shock absorption mechanism; 330. Shock absorption buffer layer; 340. Bracket; 350. Fixing bolts. Detailed Implementation

[0017] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0018] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this application described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises 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.

[0019] In this application, the terms "upper," "lower," "left," "right," "front," "rear," "top," "bottom," "inner," "outer," "middle," "vertical," "horizontal," "lateral," and "longitudinal" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. These terms are primarily for the purpose of better describing this application and its embodiments, and are not intended to limit the indicated device, element, or component to having a specific orientation, or to be constructed and operated in a specific orientation.

[0020] Furthermore, in addition to indicating location or positional relationship, some of the aforementioned terms may also have other meanings. For example, the term "above" may also be used in some cases to indicate a certain dependency or connection relationship. Those skilled in the art can understand the specific meaning of these terms in this application based on the specific circumstances.

[0021] Furthermore, the terms "installation," "setup," "equipped with," "connection," "linking," and "socketing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral structure; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium, or an internal connection between two devices, components, or parts. Those skilled in the art can understand the specific meaning of these terms in this application based on the specific circumstances.

[0022] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments. Example

[0023] See attached document Figures 1-5 An anti-interference method for a mobile fire detection system, the method being... The detection matrix module 200 collects noise feature data and establishes a noise feature library, and configures a dynamic update strategy for the noise feature library; Noise reduction is achieved based on a noise feature library and a real-time noise cancellation algorithm. The real-time noise cancellation algorithm includes the following steps: Noise matching; Inverse noise generation; Superimposed cancellation noise reduction; The noise-reduced signal is sent to the data processing module, and the data processing flow of the data processing module is as follows: S100, Raw data preprocessing; S200 and FFT transforms are used to extract vibration features; S300, optical offset calculation; S400, Generate compensation and correction strategy; S500, drive adjustment reverse compensation; The compensated and offset signals are then used for fire source identification and lane location, specifically... S1100, Signal preprocessing, performs mean filtering on the signal to eliminate high-frequency noise; S1200, peak detection, identifies 2-3 adjacent units covered by fire source imaging; S1300, Polynomial fitting, using the least squares method to solve for the coefficients a, b, c, d of the cubic polynomial; S1400, extreme value calculation, solving the derivative equation to obtain sub-pixel level center coordinates; S1500, coordinate transformation, outputs actual location information according to the preset "array unit coordinates-tunnel station number-lane position" mapping table.

[0024] This embodiment abandons the traditional passive vibration reduction approach and innovatively introduces a dynamic noise feature library and a real-time noise cancellation algorithm. The principle is as follows: the system pre-learns and establishes a multi-dimensional noise feature library covering frequency, amplitude, waveform, duration, and phase under typical operating conditions; during actual operation, it extracts signal features in real time through Fast Fourier Transform (FFT) and performs high-precision matching with the feature library, thereby generating an inverse cancellation signal with equal amplitude and opposite phase. This mechanism can "cancel" vibration noise, achieving a substantial improvement in the signal-to-noise ratio (noise reduction rate ≥30dB), significantly suppressing signal fluctuations from 0.2V to 0.05V, laying a solid foundation for the subsequent accurate extraction of weak fire source signals.

[0025] The collected noise feature data specifically includes Under typical tunnel conditions, vibration data of 10Hz-100Hz are synchronously collected by the ADXL345 sensor of the detection matrix module 200, and the initial sample set is established by continuous sampling for 100 hours. The typical tunnel conditions include empty / full-load truck passage, tunnel ventilation fan start-up, and traffic accident impact.

[0026] The establishment of the noise feature database specifically involves... The samples were classified and labeled as follows: 10-30Hz low-frequency vibration, 30-60Hz mid-frequency vibration, and 60-100Hz high-frequency vibration. For each vibration type, five core features are extracted: frequency f, amplitude A, waveform factor type, duration T, and phase φ, and stored as a feature vector [f, A, type, T, φ], thereby establishing a noise feature library.

[0027] Example feature library entry: noise_database = [ {"type": "low_freq", "f": 15, "A": 0.12, "wave": "sine", "T": 500, "φ": 0}, # 15Hz low-frequency sinusoidal vibration {"type": "high_impact", "f": 80, "A": 0.35, "wave": "pulse", "T": 10,"φ": 180} # 80Hz impact pulse.

[0028] The dynamic update strategy for configuring the noise feature library is specifically as follows: Every 24 hours, newly acquired vibration data are automatically clustered. If more than 3 unmatched features are found, a new feature entry is added. The requirement for unmatched features is a deviation >30%.

[0029] The dynamic update strategy of the noise feature library transforms the system from a static, pre-set parameter black box. The principle is to periodically cluster and analyze newly acquired vibration data, automatically identifying and learning newly added, unmatched interference features. This mechanism endows the system with online self-learning and adaptive evolution capabilities, enabling it to continuously adapt to new and unknown interference sources that arise during tunnel operation, thereby ensuring long-term reliability and a low false alarm rate throughout the system's entire lifecycle.

[0030] The noise matching process specifically involves performing an FFT transformation on the currently acquired signal, extracting the real-time feature vector [f_real, A_real, type_real, T_real, φ_real], calculating the similarity with the feature library entries using Euclidean distance, and selecting the entry with the highest matching degree as the noise source, specifically, a similarity > 85%.

[0031] The reverse noise generation specifically involves generating a cancellation signal with the same amplitude but opposite phase as the original noise based on the noise feature with the highest matching degree, and then converting the digital signal into an analog voltage signal through a 16-bit DAC chip.

[0032] The sinusoidal noise is: noise_cancel(t) = -A × sin(2πft + φ). Pulse noise: Rectangular pulse inverted.

[0033] Overlay: signal_clean = signal_raw + noise_cancel The reverse noise generation specifically involves using a differential amplifier to superimpose the original signal and the cancellation signal to achieve noise reduction; it also includes temperature and humidity compensation, which involves collecting temperature data T and humidity data H under the current environment and comparing them with set thresholds of 25℃ and 50% humidity. The compensation coefficient calculation process is as follows: K_T = 1 + (T - 25) × 0.005; K_H = 1 + (H - 50) × 0.0015; The signal strength generated by the reverse noise is adjusted according to the compensation coefficient.

[0034] The temperature and humidity sensor is located near the detection matrix module 200 and is configured to collect data (T, H) every 100ms.

[0035] It also includes enhanced moving average filtering. Implementation logic: For the noise-cancelled signal, take 5 consecutive frames (50ms each) of data and perform an arithmetic average. def moving_average(signal, window=5): filtered = [] for i in range(len(signal) - window + 1): filtered.append(sum(signal[i:i+window]) / window) return filtered.

[0036] The system includes a lens module 100 and a detection matrix module 200 that are configured to cooperate with each other. The lens module 100 is equipped with a vibration detection component 190, and the detection matrix module 200 is equipped with an adjustment driver 250 for driving the detection matrix module 200 and performing reverse movement compensation. The vibration detection component 190 and the adjustment driver 250 are both communicatively connected to a data processing module. The data processing module receives vibration data collected by the vibration detection component 190 and generates a corresponding drive voltage to control the adjustment driver 250 to perform reverse movement compensation.

[0037] The lens module 100 includes a housing 110 and a miniature heating element 120 attached to the outside of the housing 110. The housing 110 contains an auxiliary collimating lens 170, a main lens 180, a narrowband filter 130, and a temperature sensor 140 arranged sequentially. A laser positioning calibrator 150 is also provided on the outside of the housing 110. The main lens 180 is made of single-crystal germanium (Ge), with a focal length of 300mm, an aperture of 60mm, and a field of view of 8°. The surface of the main lens is coated with a Ge-ZnSe-Ge three-layer dielectric anti-reflection film. The auxiliary collimating lens 170 is a front-mounted concave collimating lens with a focal length of -50mm. The centers of the optical elements of the auxiliary collimating lens 170, the main lens 180, and the narrowband filter 130 are on the same straight line (i.e., the optical axis).

[0038] During the operation of lens module 100, the main lens 180, made of single-crystal germanium, is highly sensitive to temperature. Its refractive index changes with temperature, causing focal length drift (thermal defocusing) and affecting image sharpness. In cold or high-humidity environments, water vapor or frost can condense on the surface of optical components, completely blocking infrared signals. The heating element raises the temperature of the housing and internal lenses above the ambient dew point, thus preventing fogging / frost and ensuring the system operates normally in all weather conditions. Therefore, the miniature heating element 120 allows the module to maintain a relatively constant temperature, stabilizing the performance of the optical system and reducing performance drift caused by temperature fluctuations.

[0039] The auxiliary collimating lens 170 is a front-mounted concave collimating lens that can "pre-converge" incident rays from a large field of view (8°), allowing them to enter the main lens 180 at a smaller angle. This achieves an 8° field of view without increasing the size and complexity of the main lens. Furthermore, in conjunction with the main lens 180, it can effectively correct aberrations such as spherical aberration and coma throughout the optical system, improving the imaging quality of the peripheral fields of view.

[0040] The 60mm aperture of the main lens 180 ensures that the intensity of the collected target signal (infrared light) is within the preset range, and the 300mm focal length ensures that the system's magnification and resolution meet the requirements.

[0041] Material Selection: Single-crystal germanium: Germanium is a core material in infrared optics, exhibiting extremely high transmittance in the 3-12μm infrared band while remaining opaque in the visible light band. This inherently provides a natural filtering effect. Its high refractive index also allows for the use of fewer lenses and more curved surfaces to achieve long focal lengths, simplifying the system structure. Ge-ZnSe-Ge three-layer dielectric antireflection coating: This is a key process for improving performance. Due to germanium's high refractive index, approximately 36% of incident light is reflected from its surface. The antireflection coating, through the principle of destructive interference of light, significantly reduces this reflection loss, thereby significantly improving the overall system transmittance (from 64% to over 95%).

[0042] In this embodiment, the narrowband filter 130 is configured to filter out light except for 4.3μm, so that stray infrared radiation such as sunlight and lamplight in the background will be filtered out, and only the laser signal carrying information can reach the detector. This allows the extremely weak effective signal to be extracted from strong background noise, which greatly improves the signal-to-noise ratio and detection range of the system.

[0043] The detection matrix module 200 includes a ceramic substrate 230 and a plurality of sensing units 220 uniformly disposed on the ceramic substrate 230. A buffer layer 210 is provided on the side surface of the ceramic substrate 230 where the sensing units 220 are located, and a signal lead-out layer 240 is provided on the side surface of the ceramic substrate 230 away from the sensing units 220. The detection matrix module 200 is attached to the temperature sensor 140. In this embodiment, the detection matrix module 200 includes a ceramic substrate 230 and a plurality of sensing units 220 uniformly disposed on the ceramic substrate 230. A buffer layer 210 is provided on the side surface of the ceramic substrate 230 where the sensing units 220 are located, and a signal lead-out layer 240 is provided on the side surface of the ceramic substrate 230 away from the sensing units 220. The detection matrix module 200 is attached to the temperature sensor 140. The sensing unit 220 employs a ternary solid solution sensing layer of LiTaO3-Nb2O5-ZrO2 (Nb2O5 doping 5 mol%, ZrO2 doping 3 mol%), with a thermoelectric coefficient ≥500 μC / (m²). K); The sensor unit 220 has dimensions of 0.8mm × 0.8mm, voltage response ≥ 1.5V / W (@4.3μm, 1kHz modulation frequency), and noise equivalent power (NEP) ≤ 2 × 10⁻⁶. -11 W / √Hz.

[0044] The ceramic substrate 230 provides a robust and flat mounting base for the entire detection matrix. All sensing units 220 are arranged with high precision and uniformity using it as a reference plane, directly determining the spatial consistency of the imaging. Furthermore, the ceramic substrate 230 integrates complex metal wires, connecting the weak electrical signals generated by each sensing unit 220 to the signal extraction layer 240. Its high insulation ensures that the signals between thousands of sensing units 220 do not interfere with each other. The ceramic substrate 230 also possesses excellent thermal stability and thermal conductivity. It helps to homogenize the temperature of the entire detector array and dissipate excess heat, reducing measurement errors caused by local thermal gradients.

[0045] Sensing unit 220 is used to directly convert infrared thermal radiation into a voltage signal. The selected material is a LiTaO3-Nb2O5-ZrO2 ternary solid solution sensitive layer, in which... LiTaO3 (lithium tantalate): It is an excellent thermoelectric material substrate with a high thermoelectric coefficient and good physicochemical stability.

[0046] Nb2O5 (niobium pentoxide) doping (5 mol%): This doping is usually used to optimize the dielectric and thermoelectric properties of the material, and may be used to improve the thermoelectric coefficient or adjust the Curie temperature so that the material performs optimally within the operating temperature range.

[0047] ZrO2 (zirconia) doping (3 mol%): The purpose is to enhance the mechanical strength and thermal stability of the material, while potentially helping to suppress grain growth and improve the microstructure of the material, thereby improving the reliability and consistency of the device.

[0048] Furthermore, the detection matrix module 200 is positioned in conjunction with the temperature sensor 140, enabling the detection matrix module 200 to accurately and promptly measure real-time data.

[0049] Both the lens module 100 and the detection matrix module 200 are housed inside the detector housing 310. One side of the detector housing 310 is connected to the bracket 340 via a shock-absorbing mechanism 320. The bracket 340 is fixed inside the tunnel by fixing bolts 350. A shock-absorbing buffer layer 330 is provided on the side of the bracket 340 that is in contact with the shock-absorbing mechanism 320.

[0050] Four adjustment drivers 250 are evenly distributed on one side of the ceramic substrate 230 away from the sensing unit 220. The four adjustment drivers 250 are respectively configured to adjust the movement compensation in the positive and negative directions of the X-axis and the positive and negative directions of the Y-axis, wherein the X-axis and Y-axis are the X-axis and Y-axis on the adjustment plane formed by the four adjustment drivers 250.

[0051] The step S100 includes S110, Data Acquisition: The vibration detection component 190 acquires X / Y / Z axis acceleration data at a sampling rate of 1kHz, acquiring 1024 points at a time to obtain acceleration sequences a_x[n], a_y[n], and a_z[n] (unit: g); S120, Unit Conversion: Receives acceleration data sent by vibration detection component 190 and converts the acceleration value from g to m / s²; S130, Remove DC component: Calculate the mean of the sequence and subtract it; S140, Filtering: A Butterworth low-pass filter is applied with a cutoff frequency of 100Hz, retaining the effective vibration frequency band of 10-100Hz.

[0052] The step S200 includes FFT calculation: Perform a Fast Fourier Transform on the preprocessed acceleration sequence: X(k) = FFT(a_x_dc[n]), which yields the frequency domain amplitude spectrum |X(k)| and phase spectrum ∠X(k); Frequency identification: Find the frequency f_x (X-axis principal vibration frequency) corresponding to the amplitude peak in the range of 10-100Hz, and similarly obtain f_y (Y-axis). Amplitude calculation: Convert the frequency domain amplitude into the actual vibration displacement amplitude (unit: mm). Displacement amplitude formula: A_x = (a_x_peak) / (4π²f_x² × 1000), Where a_x_peak is the peak acceleration (m / s²), and ×1000 converts meters to millimeters; Phase extraction: Record the phase φ_x (X-axis) and φ_y (Y-axis) corresponding to the main vibration frequency to determine the vibration direction.

[0053] The step S300 includes The vibration displacement is converted into optical system offset by a pre-calibrated transfer function: Δx = A_x × K_x, Δy = A_y × K_y; where K_x and K_y are calibration coefficients, which are obtained by actual measurement using a laser interferometer. Calculate the real-time offset output: Combine the phase information to determine the offset direction (positive / negative), and finally output Δx and Δy containing positive and negative information; Specifically, step S400 involves converting the positive and negative information Δx and Δy into the corresponding compensation voltage of the corresponding adjustment driver 250 and sending it to the corresponding adjustment driver 250.

[0054] Specifically, this also includes the rules for calculating compensation. Basic compensation formula: Driver target displacement = -Optical offset (negative sign indicates reverse compensation) For example, if the optical system is offset by +0.08mm on the X-axis (in the positive direction along the tunnel length), the detection array needs to be moved 0.08mm in the negative X-axis direction.

[0055] Voltage conversion: Based on the characteristics of PZT-5H piezoelectric ceramic (displacement accuracy 10μm / V), calculate the driving voltage: U = target displacement (mm) × 1000 (μm / mm) ÷ 10 (μm / V) = target displacement × 100 (V / mm); Example: A compensation of 0.08mm corresponds to a voltage U = 0.08 × 100 = 8V.

[0056] Driver division of labor: X+ direction driver: responsible for positive X-axis compensation (voltage 0-100V). X-direction driver: responsible for X-axis negative direction compensation (voltage 0-100V). Y+ direction driver: responsible for positive Y-axis compensation (voltage 0-100V). Y-direction driver: responsible for Y-axis negative direction compensation (voltage 0-100V). 2. Closed-loop feedback control logic Feedback signal acquisition: The coordinates (x_img, y_img) of the imaging points on the detector array are monitored in real time using a sub-pixel interpolation algorithm.

[0057] Deviation calculation: Compare with the target coordinates (x0, y0) (the preset imaging point corresponding to the lane center) to obtain the residual deviation: e_x = x_img - x0, e_y = y_img - y0 Compensation correction: If |e_x|>0.01mm or |e_y|>0.01mm, then adjust the drive voltage. U_new=U_old+e×100(V / mm) (proportional coefficient 100) Precision locking: Repeat steps 1-3 until |e_x|≤0.01mm and |e_y|≤0.01mm, ensuring the total offset is ≤0.02mm.

[0058] Complete compensation calculation case Scenario: Vibration caused by heavy trucks passing through the tunnel; testing revealed: X-axis (tunnel length direction): main vibration frequency 30Hz, peak acceleration 1.2g, phase 0° (positive offset) Y-axis (tunnel width direction): main vibration frequency 25Hz, peak acceleration 0.8g, phase 180° (negative offset) 1. Data Analysis Unit Calculation Process X-axis acceleration conversion: a_x_peak = 1.2g × 9.81 m / s² / g = 11.77 m / s² X-axis displacement amplitude: A_x = 11.77 / (4π²×30²×1000) ≈ 11.77 / (4×9.87×900×1000) ≈ 0.094mm X-axis optical offset: Δx = 0.094mm × 0.85 (K_x) ≈ +0.08mm (+ indicates positive direction) Y-axis acceleration conversion: a_y_peak = 0.8g × 9.81 = 7.85 m / s² Y-axis displacement amplitude: A_y = 7.85 / (4π²×25²×1000) ≈ 7.85 / (4×9.87×625×1000) ≈ 0.059mm Y-axis optical offset: Δy = 0.059mm × 0.85 (K_y) ≈ -0.05mm (- indicates negative direction, due to 180° phase) 2. Execution process of the active compensation unit X-axis compensation: Requires a reverse movement of -0.08mm, performed by the X-direction driver. U_x-=0.08mm×100V / mm=8V Y-axis compensation: Requires a reverse movement of +0.05mm (to offset -0.05mm offset), performed by the Y+ direction driver. U_y+=0.05mm×100V / mm=5V 3. Closed-loop feedback correction After the initial compensation, the residual deviations of the imaging points are e_x = +0.012 mm and e_y = -0.008 mm. Secondary correction voltage: U_x-=8V+0.012×100=9.2V, U_y+=5V-0.008×100=4.2V Final deviations: e_x=0.005mm, e_y=0.003mm, which meets the requirement that the total offset of the imaging point is ≤0.02mm.

[0059] This solution represents a fundamental breakthrough in earthquake resistance technology. Instead of adopting the traditional approach of single passive vibration reduction, it constructs an active-passive collaborative earthquake resistance system that integrates real-time perception, intelligent decision-making, and precise execution, specifically designed to address the challenges of continuous and complex mechanical vibrations in tunnel environments.

[0060] Firstly, in terms of passive vibration damping, the solution incorporates significant mechanical structural innovation. Traditional detector installation methods often rely on rigid connections, easily transmitting vibrations from the support frame and tunnel structure directly to the core optical and detection units. This solution introduces a multi-stage buffer design between the detector housing 310 and the mounting bracket 340: it not only includes a dedicated damping mechanism 320 capable of absorbing multi-directional impacts, but also adds a high-performance damping buffer layer 330 to the contact surface. This passive system effectively filters out most of the energy from low-to-medium frequency, high-amplitude vibrations generated by heavy truck traffic and tunnel structure resonance, providing the first robust line of defense for the entire system and significantly reducing the vibration intensity transmitted to the core module at the physical level.

[0061] However, passive vibration damping cannot completely eliminate all vibrations, especially tiny displacements that have a significant impact on imaging accuracy. This is precisely where the core of this solution—the active real-time compensation system—comes in. The system integrates a high-precision vibration detection component 190 on the lens module 100, which samples the acceleration of the X, Y, and Z axes at a high frequency of 1kHz. The acquired raw data undergoes precise calculations by the data processing module: first, the main vibration frequency, amplitude, and key direction and phase information are accurately extracted from the time-domain signal using FFT (Fast Fourier Transform); then, the physical vibration displacement is converted into a specific pixel offset of the optical system on the imaging plane (e.g., Δx = +0.08mm) using a transfer function pre-calibrated by a laser interferometer.

[0062] Finally, the system performs a precise "reverse cancellation" operation. Four high-precision piezoelectric ceramic adjustment actuators 250 integrated on the back of the detection matrix module 200 are responsible for driving the X and Y axes in their positive and negative directions, respectively. The data processing module instantly converts the calculated optical offset into a corresponding driving voltage (e.g., 8V is applied to compensate for a 0.08mm offset), controlling the actuators to drive the entire detection matrix, producing a precise movement of equal magnitude and opposite direction to the vibration displacement. This process forms a high-speed closed-loop control with a response speed far faster than the tunnel's vibration cycle, ensuring that the position of the fire source image on the detector array remains stable, fundamentally suppressing image blurring and positioning drift caused by vibration. Testing shows that this active compensation system can stably control the total optical offset within 0.02mm, which plays a decisive role in achieving a long-range positioning accuracy of 0.3 meters at a distance of 150 meters.

[0063] In summary, this solution, through a collaborative design of "passively buffering and isolating macroscopic vibrations and actively compensating for and eliminating microscopic offsets," constructs a three-dimensional seismic-resistant system that far surpasses traditional solutions. This ensures that the fire detection system can maintain extremely high imaging quality and positioning accuracy even under extremely harsh vibration environments, achieving a fundamental leap from "barely usable" to "precise and reliable."

[0064] It also includes a lens module 100 that collects light data and transmits it to a detection matrix module 200 for photoelectric conversion. The converted signal is then sent to a data processing module, and the data processing flow of the data processing module is as follows: S1100, Signal preprocessing: Perform mean filtering (window size 3) on the original signal to eliminate high-frequency noise; S1200, peak detection, identifies 2-3 adjacent units (continuous units with signal values ​​higher than the threshold) covered by fire source imaging. S1300, Polynomial fitting, using the least squares method to solve for the coefficients a, b, c, d of the cubic polynomial; S1400, extreme value calculation, solving the derivative equation to obtain sub-pixel level center coordinates; S1500, coordinate transformation, outputs actual location information according to the preset "array unit coordinates-tunnel station number-lane position" mapping table.

[0065] The specific steps S1300 are as follows: Let the coordinates of the detection units be x0, x1, x2 (distributed along the X-axis), and the corresponding signal values ​​be y0, y1, y2. The cubic polynomial expression is: y = ax³ + bx² + cx + d; Use curve_fit for least squares fitting, where p0 is the initial parameter guess value; p0 = [0, -1, 0, max(y)] Assume the curve opens downwards and the vertex is close to the maximum signal value; popt, _ = curve_fit(self.cubic_polynomial, x, y, p0=[0, -1, 0, max(y)]); Returns the fitted coefficients [a, b, c, d]; The specific steps S1400 are as follows: S1410. Fit a cubic polynomial and obtain the coefficients a, b, c, d. a, b, c, d = self.fit_polynomial(x_units, signals); S1420. Find the points where the derivative is 0 (extreme points). First derivative: y' = 3ax² + 2bx + c Let y' = 0, solve the quadratic equation: 3ax² + 2bx + c = 0 discriminant = (2 b) twenty four (3 a) c, the discriminant Δ = (2b)² - 4 (3a) c If the discriminant is less than 0, there are no real roots, indicating that the curve has no obvious peak. In this case, the average coordinates of the covering unit are used as the center. S1430. Calculate the extreme points and filter the valid solutions (within the cell coordinate range). x1 = (-2 b + np.sqrt(discriminant)) / (6 a); x2 = (-2 b - np.sqrt(discriminant)) / (6 a); candidates = [x for x in [x1, x2] if min(x_units)<= x<= max(x_units)]; S1440. Select the extreme point with the largest signal value. This extreme point corresponds to the sub-pixel center coordinates (cell level). if not candidates; return np.mean(x_units); return max(candidates, key=lambda x: self.cubic_polynomial(x, a, b,c, d)); def convert_to_tunnel_position(self, subpixel_x, subpixel_y) Furthermore, step S1300 in this embodiment is specifically as follows: Let the coordinates of the detection units be x0, x1, x2 (distributed along the X-axis), and the corresponding signal values ​​be y0, y1, y2. The cubic polynomial expression is: y = ax³ + bx² + cx + d; Use curve_fit for least squares fitting, where p0 is the initial parameter guess value; p0 = [0, -1, 0, max(y)] Assume the curve opens downwards and the vertex is close to the maximum signal value; popt, _ = curve_fit(self.cubic_polynomial, x, y, p0=[0, -1, 0, max(y)]); Returns the fitted coefficients [a, b, c, d].

[0066] The specific step S400 is as follows: S1410. Fit a cubic polynomial and obtain the coefficients a, b, c, d. a, b, c, d = self.fit_polynomial(x_units, signals); S1420. Find the points where the derivative is 0 (extreme points). First derivative: y' = 3ax² + 2bx + c Let y' = 0, solve the quadratic equation: 3ax² + 2bx + c = 0 discriminant = (2 b) twenty four (3 a) c # Discriminant Δ = (2b)² - 4 (3a) c # If the discriminant is less than 0, there are no real roots, indicating that the curve has no obvious peak. In this case, the average coordinates of the covering unit are used as the center. S1430. Calculate the extreme points and filter the valid solutions (within the cell coordinate range). x1 = (-2 b + np.sqrt(discriminant)) / (6 a); x2 = (-2 b - np.sqrt(discriminant)) / (6 a); candidates = [x for x in [x1, x2] if min(x_units)<= x<= max(x_units)]; S1440. Select the extreme point with the largest signal value. This extreme point corresponds to the sub-pixel center coordinates (cell level). if not candidates; return np.mean(x_units); return max(candidates, key=lambda x: self.cubic_polynomial(x, a, b,c, d)); def convert_to_tunnel_position(self, subpixel_x, subpixel_y).

[0067] The coordinate transformation in step S500 is specifically as follows: Subpixel coordinates are converted into tunnel station numbers and lane information. Subpixel coordinates include X-axis subpixel coordinates (cell level) and Y-axis subpixel coordinates (cell level). The station number offset in the X-axis direction (cell coordinate difference × cell spacing × scale coefficient) is calculated for the X-axis subpixel coordinates (cell level) and compared with the reference station number to obtain the actual station number. Lane information is calculated for the Y-axis subpixel coordinates (cell level).

[0068] Example as follows: Scenario: A moving fire source is located in the second lane of the tunnel at a distance of 150m. The imaging covers the X-axis unit [16, 17, 18] and the Y-axis unit [31, 32, 33]. The signal values ​​are shown in the table below: direction Element coordinates Signal value (V) X-axis 16 0.8 X-axis 17 1.5 X-axis 18 1.2 Y-axis 31 1.0 Y-axis 32 1.6 Y-axis 33 0.9 Calculation process: X-axis sub-pixel center calculation: Fitting a cubic polynomial: y = -0.1x³ + 5.1x² - 52.3x + 174.5 Differentiating, we get: y' = -0.3x² + 10.2x - 52.3 Solving the equation -0.3x² + 10.2x - 52.3 = 0, we obtain an effective solution x = 17.23 (at the unit level). Y-axis sub-pixel center calculation: Fitting a cubic polynomial: y = -0.2x³ + 12.8x² - 258.5x + 1710.3 Taking the derivative, we get: y' = -0.6x² + 25.6x - 258.5 Solving the equation yields an effective solution y = 32.15 (at the unit level). Convert to tunnel location: X-axis offset: (17.23 - 32) × 0.8mm × 0.01m / mm = -0.118m (base station K1+200); Actual station number: K1+200 - 0.118m = K1+199.9; Y-axis unit 32.15 corresponds to lane 2; Output result: "subpixel_x": 17.23, "subpixel_y": 32.15, "stake_number": "K1+199.9", "lane": "Second lane", "position_error": 0.04mm (spatial error at 150m ≈ 0.3m).

[0069] In summary, this invention achieves a fundamental breakthrough in long-range precise positioning and strong anti-vibration interference capability by introducing an active-passive collaborative seismic resistance system and a sub-pixel-level fire source localization algorithm. Traditional point-type infrared flame detectors can only output area alarm signals and cannot determine the specific location of the fire source. Furthermore, in vibration environments, the optical coaxiality deviation often exceeds 0.15mm, causing the positioning error to expand to more than 1 meter. This invention, on the other hand, collects acceleration data in real time through a vibration detection component, extracts vibration frequency, amplitude, and phase using FFT transformation, and then converts mechanical vibration into optical offset through a preset transfer function, driving the adjustment actuator for reverse compensation. This system has a fast response speed and high control precision, and can stably control the total optical offset within 0.02mm, thereby achieving a spatial positioning accuracy of 0.3 meters at a distance of 150 meters. Meanwhile, by combining polynomial fitting and sub-pixel extreme value calculation, the system can improve the accuracy of fire source center positioning to "sub-pixel level" without increasing hardware resolution. Then, through a preset coordinate-station-lane mapping table, it can directly output the specific tunnel station and lane position where the fire source is located, completely solving the dilemma of "alarm without positioning" in traditional systems.

[0070] In terms of resisting complex interference and reducing false alarm rate, this invention constructs a triple anti-interference mechanism of dynamic noise feature library + real-time noise cancellation + temperature and humidity compensation, which is significantly better than the traditional method that relies on a single infrared intensity judgment. Traditional detectors are susceptible to interference from non-fire source infrared radiation such as high temperatures from vehicle exhaust pipes and lighting sources, with an average of more than 5 false alarms per month. This invention first establishes a noise feature library containing frequency, amplitude, waveform, duration, and phase under typical tunnel conditions, and updates it dynamically every 24 hours, enabling it to continuously learn and adapt to new interference sources. In actual operation, signal features are extracted in real time through FFT and matched with the feature library to generate inverse noise signals with equal amplitude and opposite phase. These signals are then converted by DAC and superimposed by differential amplifiers to achieve noise cancellation, with a noise reduction effect of more than 30dB and signal fluctuations suppressed from 0.2V to 0.05V. In addition, the system also integrates a temperature and humidity sensor to dynamically adjust the intensity of the inverse noise according to environmental conditions, further improving the system's adaptability and reliability in complex tunnel environments. This mechanism suppresses interference at the signal source and, combined with the selective transmission of the 4.3μm flame characteristic wavelength by a narrowband filter, effectively distinguishes real fire sources from interference sources such as vehicle high temperature and lights, reducing the monthly false alarm rate to less than once.

[0071] In terms of system reliability, environmental adaptability, and long-term stability, this invention achieves all-weather, highly reliable operation in extreme tunnel environments through multi-sensor fusion, closed-loop feedback control, and adaptive learning mechanisms. Traditional detectors lack real-time sensing and compensation capabilities for environmental parameters (such as temperature and humidity), and their optical systems are susceptible to thermal defocusing and fogging. This invention integrates a micro-heating element and a temperature sensor in the lens module 100, effectively preventing lens fogging and stabilizing optical performance; the detection matrix uses a ceramic substrate and a ternary solid solution sensitive layer, combining high thermoelectric coefficients with excellent stability; the vibration compensation system has a closed-loop feedback mechanism, continuously correcting the driving voltage by real-time monitoring of the imaging point coordinates and comparing them with the target position, ensuring continuously controllable compensation accuracy. Furthermore, the dynamic update strategy of the noise feature library enables the system to become "smarter with use," actively identifying and learning new interference patterns, fundamentally improving the system's robustness and maintainability throughout its entire lifecycle. In summary, this invention not only solves the three major pain points of traditional flame detectors in tunnel applications—inaccurate positioning, susceptibility to vibration, and susceptibility to false alarms—but also constructs a mobile fire detection system with high precision, strong anti-interference capabilities, and strong self-adaptive capabilities through a systematic hardware and software co-design, providing reliable technical support for fire prevention and control in highway tunnels.

[0072] The above-described embodiments are merely illustrative of certain implementations of the present invention, and are described in a relatively specific and detailed manner. However, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements are all within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the appended claims.

Claims

1. An anti-interference method for a mobile fire detection system, characterized in that: Noise feature data is collected and a noise feature library is established through the probe matrix module (200), and a dynamic update strategy for the noise feature library is configured. Noise reduction is achieved based on a noise feature library and a real-time noise cancellation algorithm. The real-time noise cancellation algorithm includes the following steps: Noise matching; Inverse noise generation; Superimposed cancellation noise reduction; The noise-reduced signal is sent to the data processing module, and the data processing flow of the data processing module is as follows: S100, Raw data preprocessing; S200 and FFT transforms are used to extract vibration features; S300, optical offset calculation; S400, Generate compensation and correction strategy; S500, drive adjustment reverse compensation; The compensated and offset signals are then used for fire source identification and lane location, specifically... S1100, Signal preprocessing, performs mean filtering on the signal to eliminate high-frequency noise; S1200, peak detection, identifies 2-3 adjacent units covered by fire source imaging; S1300, Polynomial fitting, using the least squares method to solve for the coefficients a, b, c, d of the cubic polynomial; S1400, extreme value calculation, solving the derivative equation to obtain sub-pixel level center coordinates; S1500, coordinate transformation, outputs actual location information according to the preset "array unit coordinates-tunnel station number-lane position" mapping table.

2. The anti-interference method for a mobile fire detection system according to claim 1, characterized in that: The collected noise feature data specifically includes Under typical tunnel conditions, vibration data of 10Hz-100Hz are synchronously collected by the ADXL345 sensor of the detection matrix module (200), and the initial sample set is established by continuous sampling for 100 hours. The typical tunnel conditions include empty / full-load truck passage, tunnel ventilation fan start-up, and traffic accident impact.

3. The anti-interference method for a mobile fire detection system according to claim 2, characterized in that: The specific steps to establish a noise feature database are as follows: The samples were classified and labeled as follows: 10-30Hz low-frequency vibration, 30-60Hz mid-frequency vibration, and 60-100Hz high-frequency vibration. For each vibration type, five core features are extracted: frequency f, amplitude A, waveform factor type, duration T, and phase φ, and stored as a feature vector [f, A, type, T, φ], thereby establishing a noise feature library; The specific dynamic update strategy for the configured noise feature library is as follows: Every 24 hours of operation, newly acquired vibration data are automatically clustered. If more than 3 unmatched features are found, a new feature entry is added. The requirement for unmatched features is a deviation > 30%.

4. The anti-interference method for a mobile fire detection system according to claim 3, characterized in that: The noise matching process specifically involves performing an FFT transformation on the currently acquired signal, extracting the real-time feature vector [f_real, A_real, type_real, T_real, φ_real], calculating the similarity with the feature library entries using Euclidean distance, and selecting the entry with the highest matching degree as the noise source.

5. The anti-interference method for a mobile fire detection system according to claim 4, characterized in that: The reverse noise generation specifically involves generating a cancellation signal with the same amplitude but opposite phase as the original noise based on the noise feature with the highest matching degree, and then converting the digital signal into an analog voltage signal through a 16-bit DAC chip.

6. The anti-interference method for a mobile fire detection system according to claim 5, characterized in that: The reverse noise generation specifically involves using a differential amplifier to superimpose the original signal and the cancellation signal to achieve noise reduction; it also includes temperature and humidity compensation, which involves collecting temperature data T and humidity data H under the current environment and comparing them with set thresholds of 25℃ and 50% humidity. The compensation coefficient calculation process is as follows: K_T = 1 + (T - 25) × 0.005; K_H = 1 + (H - 50) × 0.0015; The signal strength generated by the reverse noise is adjusted according to the compensation coefficient.

7. The anti-interference method for a mobile fire detection system according to claim 1, characterized in that: The step S100 includes S110, Data Acquisition: The vibration detection component (190) acquires X / Y / Z axis acceleration data at a sampling rate of 1kHz, acquiring 1024 points at a time to obtain acceleration sequences a_x[n], a_y[n], and a_z[n]. S120, Unit Conversion: Receive acceleration data sent by the vibration detection component (190) and convert the acceleration value from g to m / s²; S130, Remove DC component: Calculate the mean of the sequence and subtract it; S140, Filtering: A Butterworth low-pass filter is applied with a cutoff frequency of 100Hz, retaining the effective vibration frequency band of 10-100Hz.

8. The anti-interference method for a mobile fire detection system according to claim 7, characterized in that: The step S200 includes FFT calculation: Performing a Fast Fourier Transform on the preprocessed acceleration sequence: X(k) = FFT(a_x_dc[n]), which yields the frequency domain amplitude spectrum |X(k)| and phase spectrum ∠X(k); Frequency identification: Find the frequency f_x corresponding to the peak amplitude in the range of 10-100Hz, and similarly obtain f_y; Amplitude calculation: converting frequency domain amplitude into actual vibration displacement amplitude; Displacement amplitude formula: A_x = (a_x_peak) / (4π²f_x² × 1000), Where a_x_peak is the peak acceleration (m / s²), and ×1000 converts meters to millimeters; Phase extraction: Record the phases φ_x and φ_y corresponding to the main vibration frequency to determine the vibration direction.

9. The anti-interference method for a mobile fire detection system according to claim 8, characterized in that: The step S300 includes The vibration displacement is converted into optical system offset by a pre-calibrated transfer function: Δx = A_x × K_x, Δy = A_y × K_y; where K_x and K_y are calibration coefficients, which are obtained by actual measurement using a laser interferometer. Calculate the real-time offset output: Combine the phase information to determine the offset direction (positive / negative), and finally output Δx and Δy containing positive and negative information; Specifically, step S400 involves converting the positive and negative information Δx and Δy into the corresponding compensation amount voltage of the corresponding adjustment driver (250) and sending it to the corresponding adjustment driver (250).

10. The anti-interference method for a mobile fire detection system according to claim 9, characterized in that: The specific steps S1300 are as follows: Let the coordinates of the detection units be x0, x1, x2 (distributed along the X-axis), and the corresponding signal values ​​be y0, y1, y2. 2， The expression for a cubic polynomial is: y = ax³ + bx² + cx + d; Use curve_fit for least squares fitting, where p0 is the initial parameter guess value; p0 = [0, -1, 0, max(y)] Assume the curve opens downwards and the vertex is close to the maximum signal value; popt, _ = curve_fit(self.cubic_polynomial, x, y, p0=[0, -1, 0, max(y)]); Returns the fitted coefficients [a, b, c, d]; The specific steps S1400 are as follows: S1410. Fit a cubic polynomial and obtain the coefficients a, b, c, d. a, b, c, d = self.fit_polynomial(x_units, signals); S1420, Solve for the point where the derivative is 0. First derivative: y' = 3ax² + 2bx + c Let y' = 0, solve the quadratic equation: 3ax² + 2bx + c = 0 discriminant = (2 b) twenty four (3 a) c, the discriminant Δ = (2b)² - 4 (3a) c If the discriminant is less than 0, there are no real roots, indicating that the curve has no obvious peak. In this case, the average coordinates of the covering unit are used as the center. S1430. Calculate the extreme points and filter out valid solutions; S1440. Select the extreme point with the largest signal value, which corresponds to the sub-pixel center coordinates.

Images