A Dual-Color and Tri-Color Combined Colorimetric Temperature Measurement Method Based on Ensemble Learning

Abstract

The present invention discloses a dual-color and triple-color combined colorimetric temperature measurement method based on ensemble learning. Considering the influence of the non-linear response region of the CCD, channel reliability calibration is proposed. Considering that the noise level dominates the influence on the temperature measurement accuracy, noise level calibration is proposed. By collecting the blackbody furnace images with uniform temperature for a period of time, linear functions and quadratic functions are fitted using the relationship between the spatial domain image and the temporal domain image, and the statistical parameters of the noise magnitude introduced under the current measurement conditions are obtained as the prior knowledge of ensemble learning. Finally, a feature domain is constructed by combining channel reliability, noise parameters, and the coefficient of variation of the image spatial domain and the colorimetric domain, and the optimal temperature measurement formula is learned and predicted using random forest. When measuring temperature using the CCD image, first, the trained random forest is used to determine the optimal temperature measurement formula applicable to the current image, and then the temperature is measured based on the colorimetric temperature measurement principle. This method can combine the advantages of dual-color temperature measurement method and triple-color temperature measurement method, ensure the temperature measurement dynamic range while improving the temperature measurement accuracy, and improve the overall performance of the temperature measurement system.

Description

Technical Field

[0001] The present invention belongs to the technical field of radiation thermometry. More specifically, it relates to a dual-color and triple-color combined colorimetric thermometry method based on ensemble learning. Background Art

[0002] The measurement of high-temperature temperature fields is to measure the temperature distribution of the entire field of a high-temperature target object, which has extensive applications and important significance in modern industrial production environments. In the metallurgical industry, accurate measurement of the transient high-temperature furnace temperature field is crucial for ensuring the compositional stability of steel and avoiding material defects. The turbine blades of aeroengines can withstand a maximum temperature of 2000 °C during operation. Monitoring the temperature amplitude and heat flux distribution on their surfaces is a guarantee for the safe operation of the engine. The measurement of high-temperature temperature fields is of great significance for ensuring the safe operation of industrial equipment, improving product quality, and optimizing production processes.

[0003] High-temperature temperature field measurement technologies can be classified into two categories: contact measurement and non-contact measurement. Contact measurement is vulnerable to interference in high-temperature environments and is a single-point measurement, unable to cover the entire high-temperature full-field target. In recent years, to solve the problems of full-field temperature measurement and transient measurement, non-contact measurement has received extensive attention. Common non-contact measurement technologies include infrared thermometry, fiber-optic sensor-based thermometry, CCD-based colorimetric thermometry, etc. Among them, the accuracy of infrared thermometry is affected by surface emissivity and the measurement environment. The fiber-optic thermometry system is relatively complex and costly. The CCD charge-coupled device has advantages such as high sensitivity, small signal distortion, and stable and reliable operation. Therefore, colorimetric thermometry based on CCD has become the optimal choice for high-temperature measurement.

[0004] However, in the process of using a CCD charge-coupled device for temperature measurement, since it can collect the radiation amounts of the red, green, and blue channels, it is possible to calculate the temperature using the information of two channels, namely dual-color colorimetric thermometry, or calculate the temperature using the information of three channels, namely triple-color colorimetric thermometry. In existing colorimetric thermometry, they are considered in an either-or relationship, that is, either two channels are used throughout the entire temperature measurement process, or three channels are used throughout. In fact, dual-color colorimetric thermometry and triple-color colorimetric thermometry each have their corresponding advantages. For example, dual-color colorimetric thermometry has a higher dynamic range, and triple-color colorimetric thermometry has a higher theoretical accuracy. If the two methods can be organically combined and the most suitable thermometry method is selected under appropriate conditions, it is of great significance for improving the measurement accuracy, robustness, automation, and thermometry quality of CCD-based colorimetric thermometry. However, under what conditions which colorimetric method has better thermometry performance is a complex problem coupled with multiple factors, and it is difficult for humans to determine a general criterion. Summary of the Invention

[0005] The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a two-color and three-color combined colorimetric temperature measurement method based on ensemble learning to achieve two-color and three-color combined colorimetric temperature measurement and improve temperature measurement performance.

[0006] To achieve the above-mentioned object of the invention, the present invention provides a two-color and three-color combined colorimetric temperature measurement method based on ensemble learning, which is characterized by comprising the following steps:

[0007] (1) Construction and training of a classifier model for selecting two-color or three-color temperature measurement

[0008] 1.1) Channel reliability calibration: Obtain the channel reliability range (V min ,V max );

[0009] 1.2) Noise level calibration: Calculate M noise statistical parameters under the current temperature measurement conditions;

[0010] 1.3) Training set construction and generation

[0011] 1.3.1) Fix the temperature of the blackbody furnace at T℃ for a period of time until the temperature in the center of the blackbody furnace is uniform. Set up the CCD camera and set the exposure time in increments. Take N images of the center of the blackbody furnace {I 1 ,I 2 ,…,I N};

[0012] 1.3.2) Take the center image of the blackbody furnace {I 1 ,I 2 ,…,I N The image block of the central local area A×B is calculated, and the mean value of the image block of each channel is used as the channel pixel value, that is, for the nth (n=1,…,N) image I n , whose red channel pixel value is:

[0013]

[0014] in, Indicates I n The pixel value of the red channel in row a and column b;

[0015] The green channel pixel values are:

[0016]

[0017] in, Indicates I n The pixel value of the green channel in row a and column b;

[0018] The blue channel pixel values are:

[0019]

[0020] Among them, represents the pixel value of the a-th row and b-th column of the blue channel of I n ;

[0021] 1.3.3), calculate the two-color colorimetric value set V 1 and the three-color colorimetric value set V 2 , …, I N} of N images {I two and three :

[0022]

[0023] 1.3.4), calculate the proportionality coefficient corresponding to each group of two-color colorimetric values

[0024]

[0025] Among them, C2 = 1.4388×10 -2 m·K is the second radiation constant, λ r , λ g , λ b are the wavelengths of red, green, and blue light respectively;

[0026] 1.3.5), change the blackbody furnace temperature T °C, and repeat 1.3.1) to 1.3.4) several times to obtain a predetermined number of pairs of two-color colorimetric values and their proportionality coefficients pairs of three-color colorimetric values and their proportionality coefficients

[0027] Use the least squares method to fit the functional relationship between the two-color colorimetric value and its proportionality coefficient:

[0028]

[0029] Among them: S two represents the functional relationship between the two-color colorimetric value and its proportionality coefficient;

[0030] Use the least squares method to fit the functional relationship between the three-color colorimetric value and its proportionality coefficient:

[0031]

[0032] Among them: S three represents the functional relationship between the three-color colorimetric value and its proportionality coefficient;

[0033] 1.3.6) For each blackbody furnace image collected, construct a feature vector:

[0034] Based on the channel reliability interval (V min , V max ), determine whether the red, green, and blue channels of each image are within the channel reliability interval (V min , V max ). If they are, the result is 1; if not, the result is 0. The judgment results of the three channels serve as the first 3 dimensions of the feature vector.

[0035] The M noise statistical parameters under the current temperature measurement conditions serve as the subsequent M dimensions of the feature vector.

[0036] Calculate the spatial domain local coefficient of variation CV I :

[0037]

[0038] where I is the local area of the blackbody furnace center in the blackbody furnace image, Var(I) represents the variance of the local image, and E(I) is the average value of the local image. The three-channel values of the spatial domain local coefficient of variation CV I serve as the (3 + M + 1)-th, (3 + M + 2)-th, and (3 + M + 3)-th dimensions of the feature vector.

[0039] Calculate the two-color ratio color gamut local coefficient of variation CV two :

[0040]

[0041] where I two is the two-color ratio value local image obtained by calculating the two-color ratio value for each pixel point in the local image I, E(I two ) is the spatial domain mean value of the two-color ratio value local image I two , and Var(I two ) is the spatial domain variance of the two-color ratio value local image I two .

[0042] Take CV two as the (3 + M + 4)-th dimension of the feature vector.

[0043] Calculate the three-color ratio color gamut local coefficient of variation CV three :

[0044]

[0045] where I three is the three-color ratio value local image obtained by calculating the three-color ratio value for each pixel point in the local image I, E(I three) is the spatial domain mean value of the trichromatic colorimetric value local image I three , and Var(I three ) is the spatial domain variance of the trichromatic colorimetric value local image I three ;

[0046] Take CV three as the (3 + M + 5)-th dimension of the feature vector;

[0047] 1.3.7), For each blackbody furnace image I n collected, calculate its dichromatic colorimetric value and trichromatic colorimetric value respectively, and substitute the dichromatic colorimetric value and trichromatic colorimetric value into the function relation obtained by fitting:

[0048]

[0049] to obtain the corresponding proportionality coefficients, denoted as the inversion proportionality coefficients and Then substitute the inversion proportionality coefficients and into the following formula to obtain the inversion temperature value of this image:

[0050]

[0051] Compare the two inversion temperatures T n ′ and T two ′ of the blackbody furnace image I three ′ with the difference between the true temperature T of the blackbody furnace when collecting this blackbody furnace image I n . If the temperature inverted by the trichromatic method is closer to the true temperature, record the temperature measurement label of this image as 1, otherwise the temperature measurement label is 0;

[0052] Form a training sample with the feature vector and the temperature measurement label. Each blackbody furnace image I n can be used as a training sample;

[0053] Under different temperature measurement conditions, repeat steps 1.1) to 1.3) to obtain the input training set of the random forest learner;

[0054] 1.4), Use the ensemble learning method based on random forest to train a classifier model to determine whether dichromatic colorimetric temperature measurement or trichromatic colorimetric temperature measurement should be used for each blackbody furnace image to achieve better temperature measurement accuracy;

[0055] (2), Perform radiation temperature measurement on the newly collected blackbody furnace image based on the trained classifier model

[0056] For each newly acquired blackbody furnace image, a corresponding feature vector is obtained according to the method in 1.3.6), and then substituted into the classifier model to obtain the final prediction result. If it is 1, the blackbody furnace image is thermally measured by the three-color thermometry method; otherwise, the two-color thermometry method is used for thermal measurement.

[0057] The invention object of the present invention is realized as follows:

[0058] The dual-color and three-color combined colorimetric thermometry method based on ensemble learning of the present invention extracts the influencing factors affecting the performance of dual-color thermometry and three-color thermometry as feature vectors based on the colorimetric thermometry principle, and combines the ensemble learning method to train a random forest model as the classifier model to determine the optimal thermometry formula for the current blackbody furnace image and improve the overall thermometry performance. In the first stage of CCD imaging, considering the influence of the non-linear response region of CCD, channel reliability calibration is proposed to calibrate the reliable interval of the channel amplitude, and the channel reliability is used as one of the influencing features. In the second stage of CCD imaging, the noise level dominates the influence on the thermometry accuracy, and noise level calibration is proposed. By collecting blackbody furnace images with a uniform temperature for a period of time, the linear function and quadratic function are fitted using the relationship between the spatial domain image and the temporal domain image to obtain the statistical parameters of the noise magnitude introduced under the current measurement conditions as the prior knowledge of ensemble learning. Finally, the channel reliability, noise parameters, and the coefficient of variation of the image spatial domain and colorimetric domain are combined to form the feature domain, and the random forest is used to learn and predict the optimal thermometry formula. When thermally measuring using the CCD image, first, the trained random forest is used to judge the optimal thermometry formula applicable to the current image, and then the thermometry is performed based on the colorimetric thermometry principle. This method can combine the advantages of the two-color thermometry method and the three-color thermometry method, ensure the thermometry dynamic range while improving the thermometry accuracy, and improve the overall performance of the thermometry system.

[0059] Meanwhile, the dual-color and three-color combined colorimetric thermometry method based on ensemble learning of the present invention also has the following beneficial effects:

[0060] 1. The present invention combines the respective advantages of the two-color thermometry method and the three-color thermometry method, and incorporates the ensemble learning method into the traditional CCD radiation thermometry. The machine learning method is used to quickly judge the most suitable thermometry formula for the current image. The thermometry system can not only retain the advantage of the high dynamic range of the two-color thermometry method but also combine the advantage of the small thermometry error of the three-color thermometry. Compared with the system of the traditional single thermometry method, it has better thermometry performance and robustness.

[0061] 2. In the process of constructing the characteristic factors that affect the optimal temperature measurement formula, the present invention takes into account that in the stage from the CCD pixel receiving spectral radiation to generating photoelectrons, due to the existence of the non-linear response interval of the CCD, the proportionality between photoelectrons and spectral radiation may be destroyed. Therefore, a channel reliability calibration method based on colorimetric values is proposed. By using the fact that the colorimetric values at the same temperature should remain unchanged, the upper and lower bounds of the reliability interval are searched by changing the shooting conditions, so as to judge whether the amplitude of a certain channel is reliable. This improves the effectiveness of the feature vector and the classification accuracy of the learning machine.

[0062] 3. In the process of constructing the characteristic factors that affect the optimal temperature measurement formula, the present invention takes into account that in the process from photoelectrons to generating pixel amplitudes, the influence of CCD noise on pixel amplitudes begins to dominate. A noise level calibration method based on spatio-temporal features is proposed. By collecting a period of images of the uniform temperature at the center of the blackbody furnace, the time-domain and space-domain feature images are calculated. The least squares method is used to fit linear and quadratic functions to derive the statistical parameters of the noise magnitude under the current conditions. This provides a reference for whether to use the information of a certain channel for temperature measurement. It improves the temperature measurement characterization ability of the feature vector. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 is a flowchart of a specific implementation manner of the dual-color and triple-color combined colorimetric temperature measurement method based on ensemble learning of the present invention;

[0064] Figure 2 is a specific flowchart of channel reliability calibration in the present invention;

[0065] Figure 3 is an example result diagram of the dual-color critical value search process in the channel reliability calibration process of the present invention;

[0066] Figure 4 is an example result diagram of the triple-color critical value search process in the channel reliability calibration process of the present invention;

[0067] Figure 5 is a specific flowchart of the noise statistical parameter calibration method in the present invention;

[0068] Figure 6 is a fitting result diagram of the quadratic function in the noise statistical parameter calibration process of the present invention;

[0069] Figure 7 is a fitting result diagram of the linear function in the noise statistical parameter calibration process of the present invention;

[0070] Figure 8 is a specific flowchart of constructing the ensemble learning training set samples in the present invention;

[0071] Figure 9It is a schematic diagram of the spatial domain coefficient of variation of the sample set collected by the present invention;

[0072] Figure 10 It is a schematic diagram of the coefficient of variation of the specific color gamut of the sample set collected by the present invention;

[0073] Figure 11 It is a comparison chart of the temperature measurement results of the two-color and three-color combined colorimetric temperature measurement method based on ensemble learning of the present invention and the temperature measurement results of the traditional pure two-color temperature measurement method and pure three-color temperature measurement method. Specific Embodiments

[0074] The following describes the specific embodiments of the present invention with reference to the accompanying drawings so that those skilled in the art can better understand the present invention. It should be particularly noted that in the following description, when the detailed description of known functions and designs may dilute the main content of the present invention, these descriptions will be omitted here.

[0075] Figure 1 It is a flowchart of a specific embodiment of the two-color and three-color combined colorimetric temperature measurement method based on ensemble learning of the present invention.

[0076] In this embodiment, as Figure 1 shown, the two-color and three-color combined colorimetric temperature measurement method based on ensemble learning of the present invention includes the following steps:

[0077] Step S1: The classifier model construction and training step for selecting two-color colorimetric temperature measurement or three-color colorimetric temperature measurement Step S1.1: Channel reliability calibration

[0078] In this embodiment, as Figure 2 shown, the channel reliability calibration includes the following steps:

[0079] Step S1.1.1: Initialize the two-color critical value set Three-color critical value set Set the change significance threshold μ Two , μ Three ;

[0080] Step S1.1.2: Fix the temperature of the blackbody furnace at T °C for a period of time until the temperature at the center of the blackbody furnace hearth is uniform. Set up the CCD camera and set increasing exposure times Take N central images of the blackbody furnace {I 1 , I 2 , …, I N};

[0081] Step S1.1.3: Take the central images of the blackbody furnace {I 1 , I 2 , …, I N} Image patches of size A×B in the central local area, calculate the mean value of each channel's image patch as the channel pixel value, that is, for the nth (n = 1, …, N) image I n , its red channel pixel value is:

[0082]

[0083] Where, represents the pixel value of the a-th row and b-th column of the red channel of I n ;

[0084] The green channel pixel value is:

[0085]

[0086] Where, represents the pixel value of the a-th row and b-th column of the green channel of I n ;

[0087] The blue channel pixel value is:

[0088]

[0089] Where, represents the pixel value of the a-th row and b-th column of the blue channel of I n ;

[0090] Step S1.1.4: Calculate the two-color colorimetric value set V 1 and the three-color colorimetric value set V 2 of the N images {I N}, respectively: two and the three-color colorimetric value set V three :

[0091]

[0092] Step S1.1.5: Find n1 groups of colorimetric values with relatively stable colorimetric value magnitudes in the two-color colorimetric value set V two , calculate their mean value as the two-color colorimetric value reference value

[0093]

[0094] Where, V two {n1} represents the set of n1 groups of stable colorimetric values in the two-color colorimetric value set V two ;

[0095] Find n2 groups of colorimetric values with relatively stable colorimetric value magnitudes in the three-color colorimetric value set V three , calculate their mean value as the three-color colorimetric value reference value

[0096]

[0097] Among them, V three {n2} represents the set of n2 groups of stable colorimetric values in the trichromatic colorimetric value set V three ;

[0098] Step S1.1.6: Calculate the difference value between each group of colorimetric values in the dichromatic colorimetric value set V two and the dichromatic colorimetric reference value ;

[0099]

[0100] In this embodiment, as Figure 3 shown, find the blackbody furnace center image with the minimum exposure time corresponding to the n1 groups of blackbody furnace center images in the colorimetric value set V two {n1}, and record the corresponding difference value as the difference value set In the difference value set find the difference value greater than the change significance threshold Δμ Two ; the blackbody furnace center image with the maximum exposure time Add the corresponding green channel amplitude to the dichromatic critical value set

[0101] Step S1.1.7: Calculate the difference value between each group of colorimetric values in the dichromatic colorimetric value set V two and the dichromatic colorimetric reference value ;

[0102]

[0103] In this embodiment, as Figure 4 shown, find the blackbody furnace center image with the maximum exposure time corresponding to the n1 groups of blackbody furnace center images in the colorimetric value set V two {n1}, and record the corresponding difference value as the difference value set In the difference value set find the difference value greater than the change significance threshold Δμ Two ; the blackbody furnace center image with the minimum exposure time Add the corresponding red channel amplitude to the dichromatic critical value set

[0104] Step S1.1.8: Calculate the trichromatic colorimetric value set Vthree For each group of colorimetric values and the reference values of trichromatic colorimetric values The difference value between

[0105]

[0106] Find the blackbody furnace center image with the minimum exposure time corresponding to n2 groups of blackbody furnace center images in the colorimetric value set V three {n2}, and record the corresponding difference value as the difference value set In the difference value set Find the difference value Greater than the change significance threshold μ Three Of the blackbody furnace center image with the maximum exposure time Add the corresponding blue channel amplitude To the trichromatic critical value set

[0107] Step S1.1.9: Calculate the difference value between each group of colorimetric values in the trichromatic colorimetric value set V three And the reference values of trichromatic colorimetric values The difference value between

[0108]

[0109] Find the blackbody furnace center image with the maximum exposure time corresponding to n2 groups of blackbody furnace center images in the colorimetric value set V three {n2}, and record the corresponding difference value as the difference value set In the difference value set Find the difference value Greater than the change significance threshold μ Three Of the blackbody furnace center image with the minimum exposure time Add the corresponding red channel amplitude To the trichromatic critical value set

[0110] Step S1.1.10: Change the temperature T of the blackbody furnace, and repeat steps S1.1.2 to S1.1.9 several times to obtain the critical value sets of two colors and three colors

[0111] Step S1.1.11: Calculate the maximum value of the lower critical value Obtain the lower limit of amplitude reliability as

[0112] Step S1.1.12: Calculate the minimum value of the upper critical value Obtain the upper limit of amplitude reliability as

[0113] Step S1.1.13: Obtain the channel reliability interval as (V min , V max )

[0114] When the amplitude of a certain channel in the blackbody furnace center image satisfies: V min < v R , v G , v B < V max (where v R , v G , v B are the pixel amplitudes of the red, green, and blue channels respectively), it can be considered that at this time, the influence of thermal noise and pixel saturation overflow on this pixel value is relatively small. Using this pixel value as the input of the colorimetric temperature measurement formula can obtain a relatively reliable temperature measurement result. On the contrary, if the pixel value does not satisfy the above interval, the colorimetric temperature value obtained based on this pixel value may have a large error.

[0115] Step S1.2: Noise level calibration

[0116] In this embodiment, as Figure 5 shown, the noise level calibration includes the following steps

[0117] Step S1.2.1: Initialize the set X = Φ of the spatial domain mean E(μ(I)) of the mean image μ(I), the set Y1 = Φ of the spatial domain variance Var(μ(I)), and the set Y2 = Φ of the spatial domain mean E(σ 2 (I)) of the variance image σ 2 (I);

[0118] Step S1.2.2: Initialize the CCD image acquisition system, keep the temperature of the blackbody furnace at T for a period of time to ensure uniform temperature inside the blackbody furnace. Continuously acquire a total of F frames of blackbody furnace images, use the circular detection algorithm to obtain the center area of the blackbody furnace, and take a local image of size A × B near the center as the uniform brightness image, denoted as L i (i = 1, 2,..., F), and then calculate the spatial domain mean E(μ(I)), the spatial domain variance Var(μ(I)), and the spatial domain mean E(σ 2 (I)) according to the following formula:

[0119]

[0120] Step S1.2.3: The calculated spatial domain mean E(μ(I)), spatial domain variance Var(μ(I)), and spatial domain mean E(σ 2(I)) are respectively added to the corresponding sets: X = X ∪ E(μ(I)), Y1 = Y1 ∪ Var(μ(I)), Y2 = Y2 ∪ E(σ 2 (I));

[0121] Step S1.2.4: Change the temperature T of the blackbody furnace, and repeat Step S1.2.2 to Step S1.2.3 until the set X, Y1, Y2 to be fitted have the set data volume;

[0122] Step S1.2.5: As Figure 6 shown, use the least squares method to fit the data in Y1 and X into a quadratic function. As Figure 7 shown, fit the data in Y2 and X into a linear function, and obtain the functional relationships y1 = a1x 2 +b1x + c1 and y2 = k2x + d2 between them.

[0123] Step S1.2.6: Take:

[0124]

[0125] as the noise statistical parameters under the current temperature measurement conditions.

[0126] Subsequently, these statistical parameters will be combined as one of the features for judging the optimal temperature measurement formula, and the discriminant accuracy of the optimal formula under different shooting conditions will be enhanced by the learning agent.

[0127] Step S1.3: Training set construction and generation

[0128] In this embodiment, as Figure 8 shown, it includes the following steps:

[0129] Step S1.3.1: Fix the temperature of the blackbody furnace at T °C for a period of time until the temperature at the center of the blackbody furnace hearth is uniform. Set up the CCD camera and set increasing exposure times respectively to take a total of N central images of the blackbody furnace {I 1 , I 2 , …, I N};

[0130] Step S1.3.2: Take the image blocks of the size A × B in the central local area of the central images {I 1 , I 2 , …, I N} of the blackbody furnace, and calculate the mean value of each channel image block as the channel pixel value. That is, for the nth (n = 1, …, N) image I n , its red channel pixel value is:

[0131]

[0132] wherein, represents the pixel value of the a-th row and b-th column of the red channel of I n ;

[0133] The pixel value of the green channel is:

[0134]

[0135] wherein, represents the pixel value of the a-th row and b-th column of the green channel of I n ;

[0136] The pixel value of the blue channel is:

[0137]

[0138] wherein, represents the pixel value of the a-th row and b-th column of the blue channel of I n .

[0139] Step S1.3.3: Calculate the set V 1 of bichromatic colorimetric values and the set V 2 of trichromatic colorimetric values for each of the N images {I N}: two three :

[0140]

[0141] Step S1.3.4: Calculate the proportionality coefficient corresponding to each bichromatic colorimetric value and the proportionality coefficient

[0142]

[0143]

[0144] -2 where C2 = 1.4388×10 r m·K is the second radiation constant, and λ g , λ b are the wavelengths of red, green, and blue light respectively; Step S1.3.5: Change the temperature T °C of the blackbody furnace, and repeat Steps S1.3.1 to S1.3.4 several times to obtain a predetermined number of pairs of bichromatic colorimetric values and their proportionality coefficients and pairs of trichromatic colorimetric values and their proportionality coefficients

[0145]

[0146] Use the least squares method to fit the functional relationship between the bichromatic colorimetric values and their proportionality coefficients:​​

[0147]

[0148] Wherein: S two represents the functional relationship between the two-color colorimetric value and its proportionality coefficient;

[0149] The functional relationship between the three-color colorimetric value and its proportionality coefficient is obtained by fitting using the least squares method:

[0150]

[0151] Wherein: S three represents the functional relationship between the three-color colorimetric value and its proportionality coefficient;

[0152] Step S1.3.6: For each blackbody furnace image collected, construct a feature vector:

[0153] Based on the channel reliability interval (V min , V max ), determine whether the red, green, and blue channels of each image are within the channel reliability interval (V min , V max ). If it is within, it is 1; if not, it is 0. The judgment results of the three channels are used as the first 3 dimensions of the feature vector;

[0154] The M noise statistical parameters under the current temperature measurement conditions are used as the subsequent M dimensions of the feature vector;

[0155] Calculate the spatial domain local coefficient of variation CV I :

[0156]

[0157] Wherein, I is the local area of the blackbody furnace center of the blackbody furnace image, Var(I) represents the variance of the local image, and E(I) is the average value of the local image.

[0158] In this embodiment, the spatial domain local coefficient of variation CV I is as Figure 9 shown.

[0159] The three-channel values of the spatial domain local coefficient of variation CV I are used as the 3+M+1, 3+M+2, and 3+M+3 dimensions of the feature vector;

[0160] Calculate the two-color color gamut local coefficient of variation CV two :

[0161]

[0162] Wherein, I twois the dichromatic colorimetric value local image obtained by calculating the dichromatic colorimetric value for each pixel point in the local image I, E(I two ) is the dichromatic colorimetric value local image I two 's spatial domain mean, Var(I two ) is the spatial domain variance of the dichromatic colorimetric value local image I two .

[0163] In this embodiment, the local coefficient of variation CV of the dichromatic colorimetric gamut two is as Figure 9 shown.

[0164] Take CV two as the (3 + M + 4)-th dimension of the feature vector;

[0165] Calculate the local coefficient of variation CV of the trichromatic colorimetric gamut three :

[0166]

[0167] where I three is the trichromatic colorimetric value local image obtained by calculating the trichromatic colorimetric value for each pixel point in the local image I, E(I three ) is the spatial domain mean of the trichromatic colorimetric value local image I three , Var(I three ) is the spatial domain variance of the trichromatic colorimetric value local image I three

[0168] In this embodiment, the local coefficient of variation CV of the trichromatic colorimetric gamut three is as Figure 10 shown

[0169] Take CV three as the (3 + M + 5)-th dimension of the feature vector;

[0170] Step S1.3.7: For each blackbody furnace image I n acquired, calculate its dichromatic colorimetric value and trichromatic colorimetric value respectively, and substitute the dichromatic colorimetric value and trichromatic colorimetric value into the fitted functional relationship:

[0171]

[0172] to obtain the corresponding proportionality coefficients, denoted as the inversion proportionality coefficients and Then substitute the inversion proportionality coefficients and into the following formula to obtain the inversion temperature value of this image:

[0173]

[0174] ​Compare the two inversion temperatures T n ′ and T two ′ of the blackbody furnace image I three with the true temperature T of the blackbody furnace when collecting this blackbody furnace image I n . If the temperature inverted by the three-color method is closer to the true temperature, record the temperature measurement label of this image as 1; otherwise, the temperature measurement label is 0.

[0175] Form a training sample by combining the feature vector and the temperature measurement label. Each blackbody furnace image I n can be used as a training sample.

[0176] Under different temperature measurement conditions, repeat steps 1.1) to 1.3) to obtain the input training set of the random forest learner.

[0177] Step S1.4: Use the ensemble learning method based on random forest to train a classifier model to determine whether better temperature measurement accuracy can be achieved by using two-color or three-color colorimetric temperature measurement for each blackbody furnace image

[0178] Step S1.4.1: Set the number of decision trees as n tree , the maximum depth as depth max , the minimum number of samples required for node splitting as min sp , and the leaf node merging threshold as min leaf .

[0179] Step S1.4.2: Randomly select |D| samples with the same size as the original sample set from the original training set D with replacement based on bootstrap, and repeat n tree times to obtain the training subsets Build a decision tree based on each training subset. Each decision tree randomly selects m out of M feature vectors (m < M) to form a feature subset for training, forming a complete random forest.

[0180] Step S2: Perform radiation temperature measurement on the newly collected blackbody furnace images based on the trained classifier model

[0181] Step S2.1: For each newly collected blackbody furnace image, obtain a corresponding feature vector according to the method in step S1.3.6, and substitute the feature vector of the new image into the n tree decision trees in the random forest to classify and predict the optimal temperature measurement formula for this image, and finally obtain n tree prediction results where is the output result of the n tree th decision tree.

[0182] Step S2.2: Use the mode of the prediction results of each decision tree as the final prediction result of the image.

[0183]

[0184] Where y = 0 or 1, which is the predicted temperature measurement result, and I(·) is the indicator function with values of 1 or 0.

[0185] Step S2.3: If the predicted temperature measurement result is 1, then measure the temperature of the image using the three-color temperature measurement method; otherwise, use the two-color temperature measurement method.

[0186] Example

[0187] In this example, first, calibrate the reliability of the CCD channel. Take images of the blackbody furnace with different exposures at the same temperature, and use the mean value of the central area as the pixel amplitude. The process of calculating the two-color critical value at 1100°C is as Figure 3 shown. Among them, each point represents a set of two-color colorimetric values, and the dashed line represents the reference value of the two-color colorimetric value.

[0188] The process of calculating the three-color critical value at 1100°C is as Figure 4 shown. Among them, each point represents a set of three-color colorimetric values, and the dashed line represents the reference value of the three-color colorimetric value.

[0189] After that, calibrate the noise level of the current shooting conditions of the CCD. Videos of the blackbody furnace at 800°C, 810°C, 820°C, 830°C, 840°C, 850°C, and 900°C are collected respectively, and each video is F = 100 frames. Image blocks with a size of 51×51 in the central area of the blackbody furnace are collected as uniform brightness images. The quadratic function fitted by the red channel is as Figure 6 shown, and the linear function fitted is as Figure 7 shown.

[0190] Then collect new training set images, and construct feature vectors based on the previous channel reliability calibration and noise level calibration, as well as the spatial domain and color gamut variation coefficients of the captured images. The spatial domain variation coefficient of the training set images is as Figure 9 shown, and the color gamut variation coefficient is as Figure 10 shown.

[0191] Finally, train the random forest to learn and predict the optimal temperature measurement formula. The comparison chart of the predicted results with the pure two-color measurement and pure three-color measurement results is as Figure 11 shown. The performance results of model training are shown in the following table, and the prediction accuracy reaches 93.3%.

[0192] Table 1 Classification performance of the random forest model

[0193] Classification accuracy ACC Precision Recall F1-score Random forest model 0.9162011173184358 0.9047619047619048 0.7755102040816326 0.8351648351648352 Performance on new data 0.9333333333333333 0.9166666666666666 1.0 0.9565217391304348

[0194] Table 1

[0195] Although the above description of the illustrative embodiments of the present invention has been given for the understanding of those skilled in the art of the present technology, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those of ordinary skill in the art of the present technology, as long as various changes are within the spirit and scope of the present invention defined and determined by the appended claims, these changes are obvious, and all inventions made using the concept of the present invention are within the scope of protection.

Claims

1. A dual-color and triple-color combined colorimetric temperature measurement method based on ensemble learning, characterized by comprising the following steps: (1) Construction and training of a classifier model for selecting dual-color or triple-color colorimetric temperature measurement 1.1) Channel reliability calibration: Obtain the channel reliability interval (V min , V max ) under the current temperature measurement conditions; 1.2) Noise level calibration: Calculate M noise statistical parameters under the current temperature measurement conditions; 1.3) Construction and generation of a training set 1.3.1), Fix the temperature of the blackbody furnace at T °C for a period of time until the temperature at the center of the blackbody furnace hearth is uniform. Set up the CCD camera and set increasing exposure times respectively. Take a total of N images of the center of the blackbody furnace {I 1 , I 2 , …, I N}; 1.3.2), extract the central image of the blackbody furnace {I 1 , I 2 , …, I N} the image block of the size of the central local area A×B, and calculate the mean value of the image blocks of each channel as the channel pixel value, that is, for the nth (n = 1, …, N) image I n , its red channel pixel value is: Among them, Indicates I n The pixel value of the a-th row and b-th column of the red channel; The pixel value of the green channel is: Among them, represents the pixel value of the a-th row and b-th column of the green channel of I n ; The pixel value of the blue channel is: Among them, represents the pixel value of the a-th row and b-th column of the blue channel of I n ; 1.3.3), calculate the two-color colorimetric value sets V 1 , I 2 , …, I N} and the three-color colorimetric value sets V two and V three : 1.3.4), Calculate the proportionality coefficients corresponding to each set of two-color colorimetric values and the proportionality coefficients corresponding to the three-color colorimetric values where C2 = 1.4388×10 -2 m·K is the second radiation constant, and λ r , λ g , λ b are the wavelengths of red, green, and blue light, respectively; 1.3.5), change the temperature of the blackbody furnace to T °C, and repeat 1.3.1) to 1.3.4) several times to obtain a predetermined number of two-color colorimetric values and their proportionality coefficient pairs Three-color colorimetric values and their proportionality coefficient pairs Using the least squares method to fit the functional relationship between the two-color colorimetric value and its proportional coefficient: Where: S two represents the functional relationship between the two-color colorimetric value and its proportionality coefficient; Using the least squares method to fit the functional relationship between the three-color colorimetric value and its proportional coefficient: Where: S three represents the functional relationship between the trichromatic colorimetric values and their proportionality coefficients; 1.3.6) For each blackbody furnace image collected, construct a feature vector: Based on the channel reliability interval (V min , V max ), determine whether the red, green, and blue channels of each image are within the channel reliability interval (V min , V max ). If they are, the result is 1; if not, the result is 0. The judgment results of the three channels are used as the first three dimensions of the feature vector; The M noise statistical parameters under the current temperature measurement conditions are used as the subsequent M dimensions of the feature vector; Calculate the local coefficient of variation CV in the spatial domain I : Among them, I is the local central area of the blackbody furnace in the blackbody furnace image, Var(I) represents the variance of the local image, E(I) is the average value of the local image, and the spatial domain local coefficient of variation CV I The three-channel values of are used as the (3 + M + 1)-th, (3 + M + 2)-th, and (3 + M + 3)-th dimensions of the feature vector; Calculate the local coefficient of variation CV of the two-color ratio gamut two : Among them, I two is the local image of the two-color colorimetric value calculated for each pixel point in the local image I. E(I two ) is the spatial domain mean of the two-color colorimetric value local image I two , and Var(I two ) is the spatial domain variance of the two-color colorimetric value local image I two ; Take CV two as the (3 + M + 4)-th dimension of the feature vector; Calculate the local coefficient of variation CV of the three-color ratio gamut three : Among them, I three is the trichromatic colorimetric value local image obtained by calculating the trichromatic colorimetric value for each pixel point in the local image I, and E(I three ) is the spatial domain mean value of the trichromatic colorimetric value local image I three , and Var(I three ) is the spatial domain variance of the trichromatic colorimetric value local image I three ; Take CV three as the (3 + M + 5)-th dimension of the feature vector; 1.3.7) For each blackbody furnace image I collected n , calculate its two-color colorimetric value and three-color colorimetric value respectively, and substitute the two-color colorimetric value and the three-color colorimetric value into the function relation obtained by fitting: Obtain the corresponding proportionality coefficient, denoted as the inversion proportionality coefficient and Then substitute the inversion proportionality coefficient and into the following formula to obtain the inversion temperature value of the image: Compare the two retrieved temperatures T n ' and T two ' of the blackbody furnace image I three with the true temperature T of the blackbody furnace when collecting this blackbody furnace image I n ; if the temperature retrieved by the three-color method is closer to the true temperature, record the temperature measurement label of this image as 1, otherwise the temperature measurement label is 0; Combine the feature vector and the temperature measurement label to form a training sample, and each blackbody furnace image I n can be used as a training sample; Under different temperature measurement conditions, repeat steps 1.1) to 1.3) to obtain the input training set of the random forest learner; 1.4) Use the ensemble learning method based on random forest to train a classifier model to determine whether better temperature measurement accuracy can be achieved by using two-color colorimetric temperature measurement or three-color colorimetric temperature measurement for each blackbody furnace image; (2) Based on the trained classifier model, perform radiation temperature measurement on the newly collected blackbody furnace images For each newly collected blackbody furnace image, obtain a corresponding feature vector according to the method of 1.3.6), and then substitute it into the classifier model to obtain the final prediction result. If it is 1, the blackbody furnace image is measured by the three-color temperature measurement method; otherwise, the two-color temperature measurement method is used.

2. The two-color and three-color combined colorimetric temperature measurement method based on ensemble learning according to claim 1, wherein The channel reliability calibration is: 1.1.1), Initialize the set of bicolor critical values Set of tricolor critical values Set the change significance threshold μ Two , μ Three ; 1.1.2), Fix the temperature of the blackbody furnace at T °C for a period of time until the temperature at the center of the blackbody furnace hearth is uniform. Set up the CCD camera and set increasing exposure times respectively. Take a total of N images of the center of the blackbody furnace {I 1 , I 2 , …, I N}; 1.1.3), take the central image of the blackbody furnace {I 1 , I 2 , …, I N}, and take image blocks of size A×B in the central local area. Calculate the mean value of each channel's image block as the channel pixel value. That is, for the nth (n = 1, …, N) image I n , its red channel pixel value is: Among them, Indicates I n The pixel value of the a-th row and b-th column in the red channel of The pixel value of the green channel is: Among them, represents I n the pixel value of the a-th row and b-th column of the green channel; The pixel value of the blue channel is: Among them, represents the pixel value of the a-th row and b-th column of the blue channel of I n ; 1.1.4), Calculate the two-color colorimetric value sets V 1 , I 2 , …, I N} and the three-color colorimetric value sets V two and V three : 1.1.5), find the two-color colorimetric value set V two Among them, find n1 groups of colorimetric values with relatively stable colorimetric value magnitudes, and calculate their mean value as the two-color colorimetric value reference value at the current temperature Among them, V two {n1} represents the set V of bicolor colorimetric values two and the set of n1 groups of stable colorimetric values in it; Find the three-color colorimetric value set V three For n2 groups of colorimetric values with relatively stable colorimetric value magnitudes in three , calculate their mean value as the reference value of the three-color colorimetric value at the current temperature Among them, V three {n2} represents the set of n2 groups of stable colorimetric value sets in the trichromatic colorimetric value set V three ; 1.1.6), calculate the set V of bicolor colorimetric values two for each set of colorimetric values and bicolor colorimetric reference values in the difference value Find the blackbody furnace center image with the minimum exposure time corresponding to the n1 groups of blackbody furnace center images in the colorimetric value set V two {n1}, and record the corresponding difference value as the difference value set In the difference value set Find the difference value Greater than the change significance threshold Δμ Two The blackbody furnace center image with the maximum exposure time Add the corresponding green channel amplitude To the two-color critical value set 1.1.7), Calculate the difference value between each group of colorimetric values in the two-color colorimetric value set V two and the two-color colorimetric value reference value ​ Find the one greater than the colorimetric value set V two Among {n1}, find the blackbody furnace center image corresponding to the maximum exposure time of the n1 groups of blackbody furnace center images, and record the corresponding difference value as the difference value set Among the difference value set Find the difference value Greater than the change significance threshold Δμ Two The blackbody furnace center image with the smallest exposure time Add the corresponding red channel amplitude To the two-color critical value set 1.1.8), calculate the trichromatic colorimetric value set V three for each set of colorimetric values and the trichromatic colorimetric reference values in between Find the one smaller than the colorimetric value set V three Among {n2}, find the blackbody furnace center image corresponding to the minimum exposure time of the n2 groups of blackbody furnace center images, and record the corresponding difference value as the difference value set Among the difference value set Find the difference value Greater than the change significance threshold μ Three The blackbody furnace center image with the maximum exposure time Add the corresponding blue channel amplitude To the three-color critical value set 1.1.9), calculate the difference value between each group of colorimetric values in the tricolor colorimetric value set V three and the tricolor colorimetric reference value ​ Find the one greater than the colorimetric value set V three Among {n2}, find the blackbody furnace center image corresponding to the maximum exposure time of the n2 groups of blackbody furnace center images, and record its corresponding difference value as the difference value set Among the difference value set Find the difference value Greater than the change significance threshold μ Three The blackbody furnace center image with the minimum exposure time Add the corresponding red channel amplitude To the three-color critical value set 1.1.10), change the temperature T of the blackbody furnace, and repeat steps 1.1.2) to 1.1.9) several times to obtain the critical value sets of two-color and three-color 1.1.11), Calculate the maximum value of the lower limit critical value Obtain the lower limit of amplitude reliability as 1.1.12), Calculate the minimum value of the upper limit critical value Obtain the upper limit of amplitude reliability as 1.1.13), obtain the channel reliability interval as (V min , V max ).

3. The two-color and three-color combined colorimetric temperature measurement method based on ensemble learning according to claim 1, characterized in that, The noise level calibration is: 1.2.1), Initialize the set X = Φ of the spatial domain mean E(μ(I)) of the mean image μ(I), the set Y1 = Φ of the spatial domain variance Var(μ(I)), and the set Y2 = Φ of the spatial domain mean E(σ 2 (I)) of the variance image σ 2 (I)); 1.2.2), Initialize the CCD image acquisition system, keep the temperature of the blackbody furnace at T for a period of time to ensure uniform temperature inside the blackbody furnace, continuously acquire a total of F frames of blackbody furnace images, use the circular detection algorithm to obtain the central area of the blackbody furnace, and take a local image with a size of A×B near the center of the circle as the uniform brightness image, denoted as L i (i = 1, 2,..., F), and then calculate the spatial domain mean E(μ(I)), the spatial domain variance Var(μ(I)), and the spatial domain mean E(σ 2 (I)) according to the following formula: 1.2.3), add the calculated spatial domain mean E(μ(I)), spatial domain variance Var(μ(I)), and spatial domain mean E(σ 2 (I)) to the corresponding sets respectively: X = X ∪ E(μ(I)), Y1 = Y1 ∪ Var(μ(I)), Y2 = Y2 ∪ E(σ 2 (I)); 1.2.4) Change the temperature T of the blackbody furnace and repeat steps 1.2.2) to 1.2.3) until the set data volume is available in the sets X, Y1, and Y2 to be fitted; 1.2.5), Using the least squares method to fit the data in Y1 and X into a quadratic function, and the data in Y2 and X into a linear function, and obtaining the functional relationships between them as y1 = a1x 2 + b1x + c1 and y2 = k2x + d2; 1.2.6) Take: k2, d2, a1, As the noise statistical parameters under the current temperature measurement conditions.

Images