Infrared temperature measurement error compensation method and system in dust environment

Abstract

The invention discloses an infrared temperature measurement error compensation method and system in a dust environment, and the method comprises the following steps: collecting an infrared image of power transmission and transformation equipment in the dust environment through a thermal infrared imager, and synchronously recording environment parameters, including environment temperature, humidity, reflectivity / measurement distance and dust particle physical characteristics; preprocessing the infrared image, including graying, noise filtering and dust coverage area segmentation, classifying according to dust interference types, and extracting statistical characteristics of the infrared image; the method has the beneficial effects that the precision advantage is as follows: experimental data shows that when the dust concentration is 500mg / m < 3 >, MAE is reduced to + / -1.5 DEG C from + / -4.5 DEG C of a traditional method and is increased by 66.7%; dynamic adaptability: model online updating and threshold dynamic adjustment ensure stability (error fluctuation is less than + / -0.8 DEG C) under sudden dust or load fluctuation; and the operation and maintenance efficiency is as follows: the defect positioning time is shortened from 15 minutes to 3 minutes by AR assistance, and the inspection efficiency is improved by 80%.

Description

Technical Field

[0001] The present invention relates to the technical field of infrared temperature measurement error compensation, and in particular to an infrared temperature measurement error compensation method and system in a dusty environment. Background Art

[0002] Existing infrared temperature measurement technology is susceptible to multiple interferences in dusty environments: surface dust adhesion causes changes in thermal radiation reflectivity, lens contamination causes image blur, and dust in the optical path attenuates infrared radiation intensity. Existing compensation solutions often rely on single models (such as linear regression) or static thresholds, making them difficult to handle fluctuations in dust concentration, differences in particle characteristics, and complex interference scenarios. Furthermore, defect assessment lacks integrated analysis of dynamic environmental parameters (such as load factor and temperature change rate), leading to missed detections or misjudgments. Summary of the Invention

[0003] The main technical problem solved by the present invention is to provide an infrared temperature measurement error compensation method and system in a dusty environment, so as to solve one or more of the above-mentioned existing technical problems.

[0004] To solve the above technical problems, the present invention adopts a technical solution: a method for compensating infrared temperature measurement errors in a dusty environment, the innovation of which is that it includes the following steps:

[0005] (1) Collect infrared images of power transmission and transformation equipment in dusty environments using an infrared thermal imager and simultaneously record environmental parameters;

[0006] (2) Preprocessing of infrared images, including grayscale conversion, noise filtering, and dust-covered area segmentation;

[0007] (3) Classification according to dust interference type;

[0008] (4) Extracting statistical features of infrared images, including mean Standard deviation σ, skewness S k , kurtosis K, combined with multispectral data for feature fusion, and constructed a nonlinear polynomial regression compensation model.

[0009] (5) Correct the infrared temperature measurement value based on the compensation model and output the calibrated temperature T calibrated =T measured +ΔT;

[0010] (6) Combined with the timing analysis module, the LSTM network is used to predict the temperature change trend of the equipment, and the defect judgment threshold is dynamically adjusted based on the dual criteria.

[0011] In some embodiments, the environmental parameters in step (1) include ambient temperature, humidity, reflectivity / measurement distance, and physical properties of dust particles; in step (2), the dust coverage area is identified by the following steps: using an adaptive threshold segmentation algorithm to extract abnormal grayscale distribution areas in the image; combining texture feature analysis to distinguish dust areas from equipment surfaces; in step (3), the dust interference type is classified as follows: target surface dust: determining the degree of dust adhesion by grayscale image contrast analysis; lens dust: quantifying the pollution level based on image blur detection; optical path dust: evaluating the attenuation of infrared radiation by dust through a transmittance calculation model; composite dust interference scenario: establishing a coupling compensation model for scenarios where surface dust and optical path attenuation coexist; in step (4), constructing a nonlinear polynomial regression compensation model, the formula is:

[0012]

[0013] Among them, x1, x2, x3, and x4 represent σ, S k , K, coefficient a i 、b i 、c i d i , b0 is determined through dust experimental data training;

[0014] The compensation model further includes a dynamic update module that adjusts coefficients in real time based on an online learning mechanism to adapt to changes in dust concentration and particle characteristics;

[0015] In step (6), the defect judgment threshold is dynamically adjusted based on the dual criteria, using the formula:

[0016]

[0017] When the ambient temperature change rate dT ambient When / dt≥5℃ / min, dynamic threshold recalibration is triggered.

[0018] In some embodiments, the transmittance calculation model in step (3) is:

[0019]

[0020] Where β is the dust extinction coefficient, d is the optical path length, C dust is the dust concentration, measured in real time by a light scattering sensor;

[0021] The model further introduces dust particle composition parameters (conductivity, hygroscopicity) to modify the calculation formula of the extinction coefficient β.

[0022] In some embodiments, the nonlinear polynomial regression model in step (4) is trained independently for different dust disturbance types:

[0023] Target surface dust: Using linear regression model, the compensation formula is:

[0024] ΔT1=0.2421X1+0.1439X2-50.0101X3-87.2195X4+104.446

[0025] Lens dust: Using a quadratic polynomial model, the compensation formula is:

[0026]

[0027] Optical path dust: Using a cubic polynomial model, the compensation formula is:

[0028]

[0029] For complex dust interference scenarios, an integrated learning model (such as random forest) is used to perform weighted fusion of single model outputs.

[0030] In some embodiments, the defect classification criteria in step (6) include:

[0031] Critical defect (Class I): After calibration, the absolute temperature exceeds the national standard limit and the relative temperature difference is ≥80%;

[0032] Serious defect (Class II): The absolute temperature is within the limit but the relative temperature difference is ≥50%;

[0033] General defects (Class III): Relative temperature difference ≥30% and significant temperature field gradient.

[0034] An infrared temperature measurement error compensation system in a dusty environment, comprising

[0035] Data acquisition module: Integrates infrared thermal imager, environmental sensors (temperature and humidity, dust concentration, multispectral imaging unit) and positioning unit (RTK) to synchronously acquire infrared images and environmental parameters;

[0036] Dust classification module: Determines the type of dust interference (surface, lens, optical path, composite interference) based on image analysis and sensor data;

[0037] Compensation model module: stores pre-trained nonlinear polynomial regression models and cloud-based collaborative training interfaces, and calls corresponding models to calculate temperature compensation values based on dust types;

[0038] Dual-criteria analysis module: combines the calibrated absolute temperature with the relative temperature difference and time series prediction results of similar equipment in the same tower to output the defect level judgment result;

[0039] Self-cleaning module: integrated ultrasonic dust removal device, automatically triggering cleaning operation according to the dust pollution level of the lens;

[0040] Result output module: Generates a visual report containing temperature data, defect location and level, and supports AR device annotation and interaction.

[0041] In some embodiments, the data acquisition module further comprises:

[0042] Blackbody radiation reference source: used for on-site calibration of infrared thermal imagers, with a calibration temperature range of -40°C to 200°C and an accuracy of ±0.5°C;

[0043] Fan control unit: simulates optical path dust interference experiments, adjusts the wind speed range from 0 to 5m / s, and supports pulsed wind speed to simulate sudden dust scenarios.

[0044] In some embodiments, the compensation model module is trained by the following steps:

[0045] (1) Simulate three dust interference scenarios in a wind tunnel environment and collect infrared images and temperature data under different dust concentrations, wind speeds, and particle compositions;

[0046] (2) Extract statistical features of each image σ, S k , K and multi-spectral features as input, and the true temperature deviation ΔT as output;

[0047] (3) The Bayesian optimization algorithm is used to fit the polynomial coefficients, and the optimization goal is to minimize the mean absolute error (MAE) and the root mean square error (RMSE).

[0048] In some embodiments, the dual-criteria analysis module further comprises:

[0049] Equipment matching function on the same tower: by associating the GPS coordinates of the image with the power grid GIS system, similar equipment on the same tower can be automatically grouped together;

[0050] Dynamic threshold adjustment function: according to the ambient temperature change rate dT ambient / dt and equipment load rate fluctuations, and adaptively correct the relative temperature difference judgment threshold.

[0051] In some embodiments, the system further includes an extreme environment protection module, which seals the infrared thermal imager and integrates a temperature and humidity feedback circuit to ensure stable operation in high humidity (>90% RH) or corrosive dust environments.

[0052] The beneficial effects of the present invention are: Precision advantage: Experimental data show that at a dust concentration of 500 mg / m 3When the MAE is tested, it is reduced from ±4.5℃ of the traditional method to ±1.5℃, an improvement of 66.7%; dynamic adaptability: online model update and dynamic adjustment of thresholds ensure stability under sudden dust or load fluctuations (error fluctuation <±0.8℃); operation and maintenance efficiency: AR assistance shortens the defect location time from 15 minutes to 3 minutes, and the inspection efficiency is improved by 80%. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive work, among which:

[0054] Figure 1 This is a system overall architecture block diagram of an infrared temperature measurement error compensation method and system in a dusty environment according to the present invention.

[0055] Figure 2 This is a multi-spectral data fusion flow chart of an infrared temperature measurement error compensation method and system in a dust environment according to the present invention.

[0056] Figure 3 The present invention discloses a dust interference classification and compensation logic diagram of an infrared temperature measurement error compensation method and system in a dusty environment.

[0057] Figure 4 This is a schematic diagram of a dynamic model update mechanism of an infrared temperature measurement error compensation method and system in a dusty environment according to the present invention.

[0058] Figure 5 This is an LSTM time series prediction network structure diagram of an infrared temperature measurement error compensation method and system in a dust environment according to the present invention.

[0059] Figure 6 The present invention discloses a multi-dimensional defect criterion fusion logic diagram of an infrared temperature measurement error compensation method and system in a dusty environment.

[0060] Figure 7 This is a working principle diagram of a self-cleaning module of an infrared temperature measurement error compensation method and system in a dusty environment according to the present invention.

[0061] Figure 8 This is a schematic diagram of the extreme environment protection design of an infrared temperature measurement error compensation method and system in a dusty environment according to the present invention.

[0062] Figure 9 This is a schematic diagram of an AR visualization interactive interface of an infrared temperature measurement error compensation method and system in a dusty environment according to the present invention.

[0063] Figure 10 This is a federated learning cloud collaborative framework diagram of an infrared temperature measurement error compensation method and system in a dusty environment according to the present invention. DETAILED DESCRIPTION

[0064] The following is a clear and complete description of the technical solutions in the embodiments of the present invention. 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.

[0065] like Figures 1 to 10 As shown, the embodiment of the present invention includes:

[0066] 1. Method implementation details

[0067] (1) Data collection and preprocessing

[0068] Multispectral data fusion:

[0069] An infrared thermal imager (FLIR T840, resolution 640 × 480, temperature measurement range -20°C to 650°C) was used to synchronously collect infrared images (8–14 μm band), a visible light camera (Sony IMX477, 12 megapixels) was used to capture spectral images in the 400–700 nm range, and a near-infrared sensor (Hamamatsu InGaAs, 900–1700 nm) was used to obtain surface texture enhancement data.

[0070] Multispectral images are registered through affine transformation, and the fusion formula is:

[0071] I fused =0.6I IR +0.3I Vis +0.1I NIR

[0072] The weight coefficient is determined through optimization of dust occlusion experiments.

[0073] Dust particle characteristics detection:

[0074] Light scattering sensor (TSI 3330) measures dust concentration in real time (resolution 1mg / m 3 ) and particle size distribution (0.1-10 μm divided into 6 grades), and the data are transmitted to the main control unit through the RS485 interface.

[0075] An X-ray fluorescence spectrometer (Olympus Delta) was used to analyze the dust composition (such as the proportion of SiO2 and Al2O3), and the results were used to correct the extinction coefficient β.

[0076] (2) Classification and compensation of dust interference

[0077] Modeling of composite interference scenarios:

[0078] When the surface dust adhesion area accounts for more than 30% and the light path transmittance is less than 70%, the composite compensation model is triggered.

[0079] The calculation formula of weight α is:

[0080]

[0081] Among them, A dust is the dust coverage area, C dust is the real-time concentration, k = 0.1, C0 = 200 mg / m 3 .

[0082] Compensation value ΔT combine By weighted fusion of ΔT1 (surface) and ΔT3 (optical path) using a random forest model (10 decision trees), experiments show that the MAE is reduced to ±1.2°C.

[0083] Dynamic model updates:

[0084] The compensation model updates the coefficients every 5 minutes using the incremental least squares method. The update formula is:

[0085] θ t+1 =θ t +η(X T X) -1 X T (y-Xθ t )

[0086] Where η = 0.01 is the learning rate, X is the newly added feature matrix, and y is the measured temperature deviation.

[0087] (3) Timing analysis and defect determination

[0088] LSTM temperature trend prediction:

[0089] The LSTM network input is the device's temperature sequence over the past hour (with a sampling interval of 1 minute). The network structure consists of two LSTM layers (64 nodes per layer) and one fully connected layer. It outputs a temperature forecast for the next 10 minutes with a mean square error (MSE) of less than 0.5°C. When the predicted value deviates from the measured value by more than 5°C, an alert is triggered and the abnormal timestamp is recorded.

[0090] Multi-dimensional criterion fusion:

[0091] In the defect scoring formula, the weight coefficients are determined by the AHP method: absolute temperature (0.6), relative temperature difference (0.3), and load rate (0.1).

[0092] Dynamic threshold adjustment: When the ambient temperature change rate dT ambient When / dt≥5℃ / min, the relative temperature difference threshold is automatically reduced by 20%.

[0093] 2. System implementation details

[0094] (1) Hardware module

[0095] Multispectral imaging unit:

[0096] The infrared thermal imager and visible light camera are rigidly connected by a mechanical bracket to ensure that the field of view overlap is greater than 95%. The FPGA (Xilinx Zynq-7000) realizes synchronous image acquisition and real-time fusion (delay less than 10ms).

[0097] The near-infrared sensor is installed independently and coupled to the main light path through optical fiber to avoid dust obstruction.

[0098] Self-cleaning module:

[0099] An ultrasonic vibrator (piezoelectric ceramic PZT-5H, resonant frequency 40kHz) is integrated into the periphery of the lens. When the blur detection value (Laplace variance) is greater than 0.15, a pulse cleaning lasting 3 seconds is triggered. Experiments show that image clarity (SSIM index) is improved by 60% after cleaning.

[0100] Extreme environment protection:

[0101] The infrared thermal imager housing is made of CNC-machined aluminum alloy, and the seams are filled with silicone seals (IP67 protection); the internal electric heating film (power 10W) is activated at -40℃ to ensure that the sensor operating temperature is greater than -20℃.

[0102] (2) Software Algorithm

[0103] Cloud-based collaborative training:

[0104] The compensation model is uploaded to the cloud (AWS EC2 instance) via the HTTPS protocol, using a federated learning framework (TensorFlow Federated) to aggregate the global model every 24 hours.

[0105] When sharing data among multiple devices, differential privacy (ε=0.5) is used to protect data security, and the model update takes less than 2 minutes per 100 devices.

[0106] AR Visualization:

[0107] The defect location is mapped to the AR device (Microsoft HoloLens2) through RTK positioning (horizontal accuracy of ±0.1m), and the temperature field is displayed in pseudo-color overlay, supporting gesture interaction to zoom in and out of the hot spot area.

[0108] (3) Experimental verification

[0109] Surface dust compensation: In a silicon powder environment with a 50% adhesion rate, the linear model corrected the temperature measurement deviation from +7.8°C to +1.3°C (MAE=0.9°C).

[0110] Optical path attenuation compensation: When β = 0.2 / m, the cubic polynomial model corrects the error from -6.2°C to -0.7°C (RMSE = 1.1°C).

[0111] Critical defect detection: During testing at a substation, the system identified two Class I defects (temperature rise of 14°C + relative temperature difference of 88%), with a false negative rate of 2.5%.

[0112] The working principle of the infrared temperature measurement error compensation method in dust environment is:

[0113] Multimodal data-driven: By using infrared, visible light, near-infrared spectrum and dust characteristic sensors, a multi-dimensional feature space is constructed to improve the robustness of dust interference identification.

[0114] Classification dynamic compensation: For surface, lens, optical path and complex interference, differentiated models (linear / polynomial / ensemble learning) are used to correct temperature measurement values in real time to adapt to complex scenarios.

[0115] Intelligent decision support: Integrates time series prediction, dynamic thresholds, and multi-dimensional criteria to achieve high-precision defect classification and early warning.

[0116] The working principle of the infrared temperature measurement error compensation system in dusty environment is:

[0117] Hardware collaboration: Multispectral imaging ensures data integrity, the self-cleaning module reduces manual maintenance requirements, and the extreme protection design expands applicable scenarios (-40℃ ~ 200℃).

[0118] Software optimization: Cloud-based collaborative training improves model generalization capabilities, and AR visualization accelerates on-site operation and maintenance responses.

[0119] The advantages of this technical solution are:

[0120] Accuracy advantage: Experimental data show that at a dust concentration of 500mg / m 3 When the temperature is adjusted by 0.1°C, the MAE is reduced from ±4.5°C of the traditional method to ±1.5°C, an improvement of 66.7%.

[0121] Dynamic adaptability: Online model updates and dynamic threshold adjustments ensure stability under sudden dust or load fluctuations (error fluctuation <±0.8°C).

[0122] Operation and maintenance efficiency: AR assistance reduces defect location time from 15 minutes to 3 minutes, improving inspection efficiency by 80%.

[0123] Effects of the embodiment

[0124] In a mine power transmission and transformation scenario (PM2.5 concentration 800mg / m 3 )middle:

[0125] Composite interference compensation: When surface dust (adhesion rate 40%) and optical path dust (β = 0.18 / m) coexist, the random forest model corrects the temperature measurement error from -9.1°C to -1.4°C.

[0126] Dynamic threshold triggering: When the ambient temperature suddenly rose by 8°C / min, the system automatically adjusted the relative temperature difference threshold from 50% to 40%, successfully identifying one Class II defect.

[0127] Extreme environment operation: At -35℃ and 95% RH, the system can work continuously for 24 hours without any failure, and the temperature measurement deviation is stable within ±2℃.

[0128] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for compensating infrared temperature measurement errors in a dusty environment, characterized by: The following steps are involved: (1) Collect infrared images of power transmission and transformation equipment in dusty environments using an infrared thermal imager and simultaneously record environmental parameters; (2) Preprocessing of infrared images, including grayscale conversion, noise filtering, and dust-covered area segmentation; (3) Classification according to dust interference type; (4) Extracting statistical features of infrared images, including mean Standard deviation σ, skewness S k , kurtosis K, Combine multispectral data for feature fusion and construct a nonlinear polynomial regression compensation model; (5) Correct the infrared temperature measurement value based on the compensation model and output the calibrated temperature T calibrated =T measured +ΔT; (6) Combined with the timing analysis module, the LSTM network is used to predict the temperature change trend of the equipment, and the defect judgment threshold is dynamically adjusted based on the dual criteria.

2. The infrared temperature measurement error compensation method in a dusty environment according to claim 1, characterized in that: The environmental parameters in step (1) include ambient temperature, humidity, reflectivity / measurement distance, and physical properties of dust particles; the dust coverage area in step (2) is identified by the following steps: using an adaptive threshold segmentation algorithm to extract abnormal grayscale distribution areas in the image; combining texture feature analysis to distinguish dust areas from equipment surfaces; step (3) dust interference type classification: target surface dust: determine the degree of dust adhesion through grayscale image contrast analysis; lens dust: quantify the pollution level based on image blur detection; Optical path dust: Evaluate the attenuation of infrared radiation by dust through the transmittance calculation model; Composite dust interference scenario: For the scenario where surface dust and optical path attenuation coexist, a coupling compensation model is established; in step (4), a nonlinear polynomial regression compensation model is constructed, and the formula is: Among them, x1, x2, x3, and x4 represent σ, S k , K, coefficient a i 、b i 、c i d i , b0 is determined through dust experimental data training; The compensation model further includes a dynamic update module that adjusts coefficients in real time based on an online learning mechanism to adapt to changes in dust concentration and particle characteristics; In step (6), the defect judgment threshold is dynamically adjusted based on the dual criteria, using the formula: When the ambient temperature change rate dT ambient When / dt≥5℃ / min, dynamic threshold recalibration is triggered.

3. The infrared temperature measurement error compensation method in a dusty environment according to claim 2, characterized in that: The transmittance calculation model in step (3) is: Where β is the dust extinction coefficient, d is the optical path length, C dust is the dust concentration, measured in real time by a light scattering sensor; The model further introduces dust particle composition parameters to modify the calculation formula of the extinction coefficient β.

4. The infrared temperature measurement error compensation method in a dusty environment according to claim 2, characterized in that: The nonlinear polynomial regression model in step (4) is trained independently for different dust interference types: Target surface dust: Using linear regression model, the compensation formula is: ΔT1=0.2421X1+0.1439X2-50.0101X3-87.2195X4+104.446 Lens dust: Using a quadratic polynomial model, the compensation formula is: Optical path dust: Using a cubic polynomial model, the compensation formula is: For complex dust interference scenarios, an integrated learning model is used to perform weighted fusion of single model outputs.

5. The infrared temperature measurement error compensation method in a dusty environment according to claim 2, characterized in that: The defect classification criteria in step (6) include: Critical defect (Class I): After calibration, the absolute temperature exceeds the national standard limit and the relative temperature difference is ≥80%; Serious defect (Class II): The absolute temperature is within the limit but the relative temperature difference is ≥50%; General defects (Class III): Relative temperature difference ≥30% and significant temperature field gradient.

6. An infrared temperature measurement error compensation system in a dusty environment, characterized by: include Data acquisition module: Integrates infrared thermal imager, environmental sensor and positioning unit to synchronously acquire infrared images and environmental parameters; Dust classification module: Determines the type of dust interference based on image analysis and sensor data; Compensation model module: stores pre-trained nonlinear polynomial regression models and cloud-based collaborative training interfaces, and calls corresponding models to calculate temperature compensation values based on dust types; Dual-criteria analysis module: combines the calibrated absolute temperature with the relative temperature difference and time series prediction results of similar equipment in the same tower to output the defect level judgment result; Self-cleaning module: integrated ultrasonic dust removal device, automatically triggering cleaning operation according to the dust pollution level of the lens; Result output module: Generates a visual report containing temperature data, defect location and level, and supports AR device annotation and interaction.

7. The infrared temperature measurement error compensation system in a dusty environment according to claim 6, characterized in that: The data acquisition module also includes: Blackbody radiation reference source: used for on-site calibration of infrared thermal imagers, with a calibration temperature range of -40°C to 200°C and an accuracy of ±0.5°C; Fan control unit: simulates optical path dust interference experiments, adjusts the wind speed range from 0 to 5m / s, and supports pulsed wind speed to simulate sudden dust scenarios.

8. The infrared temperature measurement error compensation system in a dusty environment according to claim 6, characterized in that: The compensation model module is trained by the following steps: (1) Simulate three dust interference scenarios in a wind tunnel environment and collect infrared images and temperature data under different dust concentrations, wind speeds, and particle compositions; (2) Extract statistical features of each image σ, S k , K and multi-spectral features as input, and the true temperature deviation ΔT as output; (3) The Bayesian optimization algorithm is used to fit the polynomial coefficients, and the optimization goal is to minimize the mean absolute error and the root mean square error.

9. The infrared temperature measurement error compensation system in a dusty environment according to claim 6, characterized in that: The dual-criteria analysis module also includes: Equipment matching function on the same tower: by associating the GPS coordinates of the image with the power grid GIS system, similar equipment on the same tower can be automatically grouped together; Dynamic threshold adjustment function: according to the ambient temperature change rate dT ambient / dt and equipment load rate fluctuations, and adaptively correct the relative temperature difference judgment threshold.

10. The infrared temperature measurement error compensation system in a dusty environment according to claim 6, characterized in that: The system further includes an extreme environment protection module, which seals the infrared thermal imager and integrates a temperature and humidity feedback circuit to ensure stable operation in high humidity or corrosive dust environments.

Images