A drift self-calibration method for multimodal sensors in complex industrial environments

Abstract

The present invention provides a multimodal sensor drift self-calibration method in a complex industrial environment, which relates to the field of sensor calibration technology. It includes: deploying an environmental reference sensor group to collect temperature gradient, equipment operation time and environmental humidity change data in real time, building a multivariate autoregressive model to dynamically predict the temperature deviation of the thermal imaging sensor, introducing a compensation coefficient to correct the temperature value after standardization processing and recursive parameter optimization, and using a two-dimensional discrete cosine transform to calculate the high-frequency energy of the image block to generate a clarity score; when the score is lower than the preset threshold, the golden section method is used to search for the optimal focus position within the preset focal length range to achieve adaptive calibration. The calibration is dynamically triggered by a self-supervised closed-loop mechanism, combined with sliding window data update and outlier filtering to ensure that the model adapts to dynamic changes in the environment. This method solves the problems of temperature drift of thermal imaging sensors and image blur of visible light sensors in high temperature and high dust scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of sensor calibration, and in particular to a multi-modal sensor drift self-calibration method in a complex industrial environment. Background Art

[0002] In complex industrial environments like cotton and linen textiles and chemical fiber processing, where high temperatures, high dust levels, and continuous equipment operation are common, multimodal sensors (such as long-wave thermal imaging sensors and visible light cameras) have become core components of fire warning systems due to their ability to collaboratively monitor temperature fields and smoke characteristics. However, these sensors face multiple technical challenges in long-term operation: Due to high temperatures, the temperature measurement accuracy of thermal imaging sensors gradually decreases over time, significantly increasing the risk of missed fire reports. Furthermore, dust accumulation and equipment vibration can cause visible light lens focus shifts or image blur, reducing the accuracy of smoke and fire detection models for low-contrast smoke targets, seriously affecting the effective recognition of early fire signals.

[0003] Existing research relies on manual, periodic downtime for calibration, such as thermal imaging blackbody calibration and manual focusing of visible light lenses. A single calibration can take hours, making it difficult to adapt to the 24 / 7 continuous operation requirements of industrial production. Furthermore, these manually set fixed calibration parameters struggle to respond in real time to dynamic disturbances such as ambient temperature and humidity fluctuations and equipment layout changes. This results in lag errors in sensor performance after calibration and fails to fundamentally address drift issues in complex environments. These shortcomings lead to a gradual decrease in the detection accuracy of multimodal sensors over long-term use, posing a significant challenge to the reliability of fire warning systems. Summary of the Invention

[0004] In order to overcome the deficiencies of the prior art, the purpose of the present invention is to provide a multimodal sensor drift self-calibration method in a complex industrial environment, which solves the problem of time-consuming and low-accuracy sensor calibration in the prior art.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A multimodal sensor drift self-calibration method in a complex industrial environment, comprising:

[0007] By deploying an environmental reference sensor group, real-time data on temperature gradients, equipment operation time, and environmental humidity changes in industrial scenarios can be collected.

[0008] The multivariate autoregressive model is used to construct the initial temperature deviation autoregressive model;

[0009] The initial temperature deviation autoregressive model is optimized based on the temperature gradient data, the equipment operation time data and the ambient humidity change data to obtain the final temperature deviation autoregressive model;

[0010] According to the final temperature deviation autoregressive model and the introduction of compensation coefficient, the corrected temperature data is obtained;

[0011] Visible light image data is acquired and the image clarity score is calculated based on the high-frequency energy of the two-dimensional discrete cosine transform. When the clarity is lower than the preset threshold, the golden section method is used to adaptively adjust the lens focal length to obtain the calibrated image clarity.

[0012] Preferably, it also includes:

[0013] Collecting the corrected temperature data within a preset time and the image clarity data of a preset number of frames;

[0014] Comparing the corrected temperature data within a preset time with the environmental reference data to obtain a first comparison result;

[0015] Comparing the image definition data of a preset number of frames with the image definition of a reference image to obtain a second comparison result;

[0016] Retrain the parameters of the final temperature deviation autoregressive model according to the first comparison result;

[0017] The focal length of the lens is readjusted according to the second comparison result.

[0018] Preferably, the initial temperature deviation autoregressive model is optimized according to the temperature gradient data, the equipment operation time data and the ambient humidity change data to obtain the final temperature deviation autoregressive model, including:

[0019] Normalize the temperature gradient data, equipment operation time data, and ambient humidity change data to obtain normalized data;

[0020] The initial temperature deviation autoregressive model is optimized according to the normalized data to obtain the final temperature deviation autoregressive model, wherein the expression of the final temperature deviation autoregressive model is:

[0021]

[0022] Where, ΔT t is the temperature deviation, p is the autoregressive order, α i , β, γ, δ are the first model parameter, the second model parameter, the third model parameter and the fourth model parameter respectively, ΔH t,scaled 、D t,norm 、 are respectively the normalized environmental humidity change data, equipment operation time data and temperature gradient data, ∈ t is zero-mean Gaussian noise, and i is a natural number.

[0023] Preferably, the expression of the corrected temperature data is:

[0024] T c,t =T sensor,t -K t ΔT t ;

[0025] Among them, T c,t is the corrected temperature data, T sensor,t is the measured temperature data, K t is the output compensation coefficient.

[0026] Preferably, obtaining visible light image data and calculating an image clarity score based on high frequency energy of a two-dimensional discrete cosine transform comprises:

[0027] Performing image segmentation on the visible light image data to obtain segmented data;

[0028] Perform two-dimensional DCT transformation on the block data to obtain the frequency domain coefficient matrix;

[0029] Determine the energy of high-frequency components based on the frequency domain coefficient matrix;

[0030] The image clarity score is determined based on the energy of the high-frequency components.

[0031] Preferably, the frequency domain coefficient matrix is ​​expressed as:

[0032]

[0033] Among them, C(u,v) is the frequency domain coefficient matrix, α(u) is the DC component coefficient, α(v) is the AC component coefficient; u is the frequency index of the image block in the horizontal direction; v is the frequency index of the image block in the vertical direction; I(x,y) is the pixel value of the image block at (x,y).

[0034] Preferably, the energy of the high-frequency component is expressed as:

[0035]

[0036] Among them, E b is the energy of the high frequency component.

[0037] Preferably, the lens focal length is adaptively adjusted using the golden section method to obtain calibrated image clarity, including:

[0038] Set the initial split point within the preset focal length range and calculate the image clarity at the corresponding position;

[0039] Narrow the search interval according to the clarity comparison result until the interval accuracy meets the preset threshold;

[0040] The endpoint with the highest clarity is selected as the optimal focal length position, and the clarity of the image after calibration is verified to obtain the clarity of the calibrated image.

[0041] The present invention discloses the following technical effects:

[0042] The present invention provides a multimodal sensor drift self-calibration method in a complex industrial environment, comprising:

[0043] By deploying an environmental reference sensor group, temperature gradient data, equipment operating time data, and environmental humidity change data in industrial scenarios are collected in real time; an initial temperature deviation autoregressive model is constructed using a multivariate autoregressive model; the initial temperature deviation autoregressive model is optimized based on the temperature gradient data, equipment operating time data, and environmental humidity change data to obtain a final temperature deviation autoregressive model; based on the final temperature deviation autoregressive model and introducing a compensation coefficient, corrected temperature data is obtained; visible light image data is obtained and the image clarity score is calculated based on the high-frequency energy of the two-dimensional discrete cosine transform. When the clarity is lower than a preset threshold, the golden section method is used to adaptively adjust the lens focal length to obtain the calibrated image clarity. The present invention realizes real-time correction and adaptive optimization of sensor performance by constructing a temperature deviation autoregressive model and an image clarity measurement method based on DCT high-frequency energy, combined with golden section search and dynamic parameter update mechanism. The method adopts a self-supervised closed-loop design, which can dynamically respond to environmental changes and equipment status fluctuations, significantly improving the long-term stability and detection accuracy of the sensor. In addition, the system does not require manual intervention or shutdown operations, and is fully adapted to the continuous production needs of industrial scenarios. It provides efficient and reliable technical support for the reliable operation of fire warning systems and multimodal sensor monitoring in complex industrial environments, and has broad application potential and promotion value. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 An embodiment of the present invention provides a multimodal sensor drift self-calibration method in a complex industrial environment. DETAILED DESCRIPTION

[0046] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0047] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0048] like Figure 1 As shown, the present invention provides a multimodal sensor drift self-calibration method in a complex industrial environment, comprising:

[0049] Step 100: Deploy an environmental reference sensor group to collect real-time temperature gradient data, equipment operation time data, and environmental humidity change data in the industrial scene;

[0050] Specifically, the environmental reference sensor group is fixedly installed in an unobstructed area of ​​the workshop and includes high-precision thermocouples (accuracy of ±0.5°C) and temperature and humidity sensors (humidity accuracy of ±2% RH), with a sampling frequency of 1Hz. The sensor group adopts a rectangular grid layout with 2 meters of horizontal spacing (east-west) and 1.5 meters of vertical spacing (north-south). A total of 3×3 high-precision thermocouples are deployed to form a temperature field monitoring network.

[0051] Step 200: constructing an initial temperature deviation autoregressive model using a multivariate autoregressive model;

[0052] Step 300: Optimizing the initial temperature deviation autoregressive model based on the temperature gradient data, the equipment operation time data, and the ambient humidity change data to obtain a final temperature deviation autoregressive model;

[0053] Step 400: According to the final temperature deviation autoregressive model and introducing the compensation coefficient, the corrected temperature data is obtained;

[0054] Step 500: Obtain visible light image data and calculate the image clarity score based on the high-frequency energy of the two-dimensional discrete cosine transform. When the clarity is lower than a preset threshold, the golden section method is used to adaptively adjust the lens focal length to obtain the calibrated image clarity.

[0055] Furthermore, it also includes:

[0056] Collecting the corrected temperature data within a preset time and the image clarity data of a preset number of frames;

[0057] Comparing the corrected temperature data within a preset time with the environmental reference data to obtain a first comparison result;

[0058] Comparing the image definition data of a preset number of frames with the image definition of a reference image to obtain a second comparison result;

[0059] Retrain the parameters of the final temperature deviation autoregressive model according to the first comparison result;

[0060] The focal length of the lens is readjusted according to the second comparison result.

[0061] Furthermore, the initial temperature deviation autoregressive model is optimized based on the temperature gradient data, the equipment operation time data, and the ambient humidity change data to obtain the final temperature deviation autoregressive model, including:

[0062] Normalize the temperature gradient data, equipment operation time data, and ambient humidity change data to obtain normalized data;

[0063] The initial temperature deviation autoregressive model is optimized according to the normalized data to obtain the final temperature deviation autoregressive model;

[0064] Specifically, environmental benchmark data collection and preprocessing:

[0065] The temperature deviation of the thermal imaging sensor is defined as: the measured temperature data T sensor (t) and the reference temperature T provided by the environmental reference sensor ref The difference between:

[0066] ΔT t =T sensor (t)-T ref (t);

[0067] According to the above formula, the historical temperature deviation can be obtained: ΔT t-1 ,ΔT t-2 ,…

[0068] In order to statistically analyze the impact of ambient humidity fluctuations on sensor drift, the change in ambient humidity is calculated:

[0069] ΔH t =H t -H t-1 ;

[0070] Record the cumulative running time of the device: D t ;

[0071] The difference in temperature distribution in different areas of the environment is described by the ambient reference temperature gradient:

[0072]

[0073] in, is the rate of change of temperature in the horizontal direction (east-west); is the rate of change of temperature in the vertical direction (north-south direction), which is calculated as follows:

[0074] For a certain sensor point (x i ,y i ), select the horizontally adjacent sensors (x i-1 ,y i )、(x i+1 ,y i ); Select the vertically adjacent sensors (x i ,y i-1 )、(x i ,y i+1 ),calculate:

[0075]

[0076] Using a multivariate autoregressive model (AR), the initial temperature deviation autoregressive model expression is:

[0077]

[0078] Where p is the autoregressive order, determined by the Akaike Information Criterion (AIC); ∈ t is zero-mean Gaussian noise.

[0079] Variables of different dimensions may have significant differences in their numerical ranges (for example, the time span is much larger than the temperature deviation), which may lead to model parameters (α i , β, γ, δ) in the estimation process, the variable with larger value dominates the loss function optimization, and the contribution of other variables is diluted. ΔH t 、D t Normalize to eliminate the effect of dimension. Do as follows:

[0080] right Use z-score standardization:

[0081]

[0082] Among them, μ x and σ x They are The mean and standard deviation of .

[0083] ΔH t Using Min-Max normalization:

[0084]

[0085] Where ΔH max and ΔHmin They are the minimum and maximum values ​​of historical data respectively.

[0086] To D t Use z-score standardization:

[0087]

[0088] Among them, μ D and σ d D t The mean and standard deviation of .

[0089] The final expression of the temperature deviation autoregressive model is:

[0090]

[0091] Where, ΔT t is the temperature deviation, p is the autoregressive order, α i , β, γ, δ are the first model parameter, the second model parameter, the third model parameter and the fourth model parameter respectively, ΔH t,scaled 、D t,norm 、 are respectively the normalized environmental humidity change data, equipment operation time data and temperature gradient data, ∈ t is zero-mean Gaussian noise, and i is a natural number.

[0092] Furthermore, using the historical 30-day benchmark data and sensor raw data, the model parameters are solved by the least squares method to minimize the residual sum of squares:

[0093] R is the total number of days.

[0094] New data is collected every 5 minutes, and the parameters are dynamically updated using the recursive least squares method to adapt to fluctuations in ambient temperature and humidity and equipment aging.

[0095] Model output compensation coefficient K t , the corrected temperature value is:

[0096] T c,t =T sensor,t -K t ΔT t ;

[0097] Among them, K t Deviation from the prediction of the AR model Calculation yields:

[0098]

[0099] The corrected temperature T is set every 5 minutes. c,tCompared with the environmental benchmark data, if the absolute error exceeds ±2°C, the model parameters will be retrained.

[0100] Furthermore, obtaining visible light image data and calculating an image clarity score based on high-frequency energy of a two-dimensional discrete cosine transform includes:

[0101] Performing image segmentation on the visible light image data to obtain segmented data;

[0102] Perform two-dimensional DCT transformation on the block data to obtain the frequency domain coefficient matrix;

[0103] Determine the energy of high-frequency components based on the frequency domain coefficient matrix;

[0104] The image clarity score is determined based on the energy of the high-frequency components.

[0105] Specifically, image clarity, as a core indicator for measuring image quality, has a significant correlation between its numerical value and the subjective perception of the human eye. Low clarity usually manifests as a blurred image. To achieve automatic calibration of the visible light sensor, the system first calculates the clarity value of the current image. This value must be compared with the clarity threshold established based on the reference image. When the clarity of the real-time image is detected to be lower than the preset threshold, the system will automatically trigger the self-calibration procedure. It should be emphasized that the establishment of this threshold is based on the clarity value of the reference image obtained in advance through manual focusing. This setting method based on a standard reference object ensures the objectivity and reliability of the calibration system.

[0106] Image energy is primarily concentrated in the mid- and low-frequency regions of the amplitude spectrum, while the richness of image detail and sharpness of contours are determined by the high-frequency components. When an image is out of focus, its high-frequency information is significantly attenuated, so the degree of defocus can be determined by the amount of high-frequency information. While image clarity evaluation functions based on Fourier transforms are effective, the transformation process involves imaginary coefficients, resulting in high computational complexity.

[0107] The two-dimensional discrete cosine transform (DCT) is widely used in the field of image compression. This transform converts from the spatial domain to the frequency domain, decomposing the image signal into a weighted sum of cosine function basis signals of different frequencies. High-frequency components correspond to image details and edge features, while low-frequency components correspond to smooth areas. When an image is blurred, its high-frequency components are correspondingly reduced. Therefore, the amplitude of the high-frequency portion of the DCT coefficients can be used as a measure of image clarity. Furthermore, the transform coefficients of the DCT algorithm are all real numbers, significantly reducing the computational complexity of image clarity assessment algorithms compared to the Fourier transform.

[0108] Define image I with a resolution of W×H. The steps for implementing its two-dimensional DCT are as follows:

[0109] Image segmentation:

[0110] Split I(x,y) into multiple non-overlapping N×N blocks, where N = 8. Each block is denoted as I(x,y), where x,y∈[0,N-1].

[0111] Two-dimensional DCT transform:

[0112] Perform a two-dimensional DCT transformation on each block to obtain the frequency domain coefficient matrix C(u,v), which is calculated as follows:

[0113]

[0114] in:

[0115]

[0116] In the above formula, when u=0, the DC component coefficient When u>0, the DC component coefficient The same applies to v and α(v).

[0117] Each segmented N×N matrix is ​​transformed through the DCT to yield another N×N matrix, consisting of direct current coefficients (DC components) and alternating current coefficients (AC components). The DC component represents the low-frequency components of the image, containing the primary information about image brightness; the AC component represents the mid- and high-frequency components of the image, corresponding to edges and details. Therefore, it is only necessary to exclude the DC component, define the high-frequency region, and calculate its energy, which serves as a measure of the sharpness of the segmented N×N matrix.

[0118] N is the size of the segmented image block (N×N), and N=8 is taken in this paper; u is the frequency index of the image block in the horizontal direction; v is the frequency index of the image block in the vertical direction; satisfying: u, v∈p0,N-1; I(x,y) is the pixel value of the image block at (x,y), x,y∈[0,N-1].

[0119] High frequency energy calculation:

[0120]

[0121] Among them, E b is the energy of the high-frequency component; C(0,0) is the DC component.

[0122] Image clarity rating:

[0123] The high-frequency energy of all blocks (N×N) is averaged to obtain the clarity of the entire image:

[0124]

[0125] Among them, K is the total number of N×N small blocks that the entire image is divided into; Corresponding to the clarity (high-frequency energy) of each N×N small block.

[0126] The larger the S value, the more significant the high-frequency energy, the richer the details, and the clearer the image. Conversely, the smaller the S value, the higher-frequency components are suppressed, and the image becomes more blurred.

[0127] Furthermore, the golden section method is used to adaptively adjust the lens focal length to obtain the calibrated image clarity, including:

[0128] Set the initial split point within the preset focal length range and calculate the image clarity at the corresponding position;

[0129] Narrow the search interval according to the clarity comparison result until the interval accuracy meets the preset threshold;

[0130] The endpoint with the highest clarity is selected as the optimal focal length position, and the clarity of the image after calibration is verified to obtain the clarity of the calibrated image.

[0131] Specifically, an adaptive focus method based on image clarity measurement was designed. By adjusting the focal length of the lens, the image clarity at each position was calculated, and the focal length position corresponding to the maximum clarity was found.

[0132] Preset an image with good clarity as a reference image I r , and calculate its clarity, using S(I r ) indicates that when the clarity of the real-time image is less than 80% of the clarity of the reference image, that is: The lens is judged to have focus offset or surface contamination, triggering the adaptive calibration process:

[0133] Focus adjustment range based on historical presets Adjustments are made with the goal of maximizing image clarity, and the golden section method is used to search for the optimal focal length position.

[0134] Initial interval Take the initial segmentation point Satisfy respectively:

[0135]

[0136] Calculate separately and Image clarity under and

[0137] Update the search interval according to the following rules:

[0138] when When , the maximum value is in the interval Update the split point:

[0139]

[0140] calculate Image clarity under and and compare.

[0141] On the contrary, when The maximum value is located in the interval Update the split point:

[0142]

[0143] calculate Image clarity under and and compare.

[0144] The loop executes until the stop condition is met: in,

[0145] When the termination condition is met, the midpoint of the interval or the endpoint with higher image clarity is taken as the optimal focal length position:

[0146]

[0147] in,

[0148] After calibration, recalculate f best The image clarity under the condition is compared with the image clarity of the reference image. Then start the secondary focusing.

[0149] Furthermore, the trigger conditions are calibrated:

[0150] Thermal imaging sensor calibration trigger conditions:

[0151] Temperature deviation |ΔT for 3 consecutive sampling periods (15 minutes) t |>2℃, or the cumulative operating time of the equipment D t When the time is ≥720 hours, the model will be forced to retrain (using the data from the last 30 days) and perform calibration.

[0152] Visible light sensor calibration trigger conditions:

[0153] When the clarity drops rapidly, the clarity of 10 consecutive frames is less than 80% of the reference image clarity, that is, S(I)<0.8·S(I r ), triggering the visible light sensor self-calibration.

[0154] Historical data update:

[0155] A sliding window is used to store the calibration data for the last 30 days. Data outside the window is removed in chronological order to ensure that the model training data reflects the current environmental characteristics.

[0156] Environmental benchmark sensor data (such as thermocouple temperature, temperature and humidity) is filtered for outliers using the IQR method to prevent impulsive noise from contaminating the training set. Specifically, the data body range is determined by calculating the difference between the first quartile (Q1, 25% quantile) and the third quartile (Q3, 75% quantile) (IQR = Q3 - Q1). This is used as the basis for defining the outlier boundaries. Data points below Q1 - 1.5 × IQR or above Q3 + 1.5 × IQR are considered outliers.

[0157] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0158] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A multimodal sensor drift self-calibration method in a complex industrial environment, characterized by: include: By deploying an environmental reference sensor group, real-time data on temperature gradients, equipment operation time, and environmental humidity changes in industrial scenarios can be collected. The multivariate autoregressive model is used to construct the initial temperature deviation autoregressive model; The initial temperature deviation autoregressive model is optimized based on the temperature gradient data, the equipment operation time data and the ambient humidity change data to obtain the final temperature deviation autoregressive model; According to the final temperature deviation autoregressive model and the introduction of compensation coefficient, the corrected temperature data is obtained; Obtain visible light image data and calculate the image clarity score based on the high-frequency energy of the two-dimensional discrete cosine transform. When the clarity is lower than the preset threshold, the golden section method is used to adaptively adjust the lens focal length to obtain the calibrated image clarity. Also includes: Collecting the corrected temperature data within a preset time and the image clarity data of a preset number of frames; Comparing the corrected temperature data within a preset time with the environmental reference data to obtain a first comparison result; Comparing the image definition data of a preset number of frames with the image definition of a reference image to obtain a second comparison result; Retrain the parameters of the final temperature deviation autoregressive model according to the first comparison result; re-adjusting the focal length of the lens according to the second comparison result; The final temperature deviation autoregressive model is obtained by optimizing the initial temperature deviation autoregressive model based on the temperature gradient data, equipment operation time data, and ambient humidity change data, including: Normalize the temperature gradient data, equipment operation time data, and ambient humidity change data to obtain normalized data; The initial temperature deviation autoregressive model is optimized according to the normalized data to obtain the final temperature deviation autoregressive model, wherein the expression of the final temperature deviation autoregressive model is: ； in, is the temperature deviation, is the autoregressive order, , , , are the first model parameter, the second model parameter, the third model parameter and the fourth model parameter, respectively. 、 、 They are respectively the normalized environmental humidity change data, equipment operation time data and temperature gradient data. is zero-mean Gaussian noise, is the historical temperature deviation, is a natural number; The expression of the corrected temperature data is: ； in, is the corrected temperature data, is the measured temperature data, is the output compensation coefficient; Acquire visible light image data and calculate image clarity scores based on high-frequency energy from a two-dimensional discrete cosine transform, including: Performing image segmentation on the visible light image data to obtain segmented data; Perform two-dimensional DCT transformation on the block data to obtain the frequency domain coefficient matrix; Determine the energy of high-frequency components based on the frequency domain coefficient matrix; Determine the image clarity score based on the energy of the high-frequency components; Adopting the golden section method to adaptively adjust the lens focal length, we can obtain the calibrated image clarity, including: Set the initial split point within the preset focal length range and calculate the image clarity at the corresponding position; Narrow the search interval according to the clarity comparison result until the interval accuracy meets the preset threshold; The endpoint with the highest clarity is selected as the optimal focal length position, and the clarity of the image after calibration is verified to obtain the clarity of the calibrated image.

2. A multimodal sensor drift self-calibration method in a complex industrial environment according to claim 1, characterized in that: The expression of the frequency domain coefficient matrix is: ； in, is the frequency domain coefficient matrix, is the DC component coefficient, is the AC component coefficient; is the frequency index of the image block in the horizontal direction; v is the frequency index of the image block in the vertical direction For image blocks in The pixel value at , N is the number of pixels of the side length of each sub-block.

3. A multimodal sensor drift self-calibration method in a complex industrial environment according to claim 1, characterized in that: The energy of the high-frequency component is expressed as: ； in, is the energy of the high-frequency component, is the DC component.

Images