An infrared temperature measurement method for selecting the optimal regression model

Abstract

The present invention discloses an infrared temperature measurement method for selecting an optimal regression model, which includes the following specific steps: collecting thermal imager images and focal plane temperature data; establishing a set of polynomial F(TP) functions to be fitted according to the focal plane temperature TP; establishing a set of polynomial F(T, TP) functions according to the target temperature T and the focal plane temperature TP; respectively processing the F(TP) function set and the F(T, TP) function set to obtain the parameter sets and R values of each function; setting a threshold thr_R1 and a threshold thr_R2; by comparing the R values of the F(TP) function set and the F(T, TP) function set with the set thresholds, determining the optimal function model and the optimal function parameters; downloading the obtained optimal function model and the optimal function parameters into the thermal imager, obtaining the current focal plane temperature TP of the thermal imager and the current target temperature image value Y, and calculating the target temperature T to complete the mapping from the image value to the temperature value. The algorithm of the present invention automatically selects the optimal model for calculation, reducing the algorithm configuration parameters.

Description

Technical Field

[0001] The present invention relates to the technical field of infrared temperature measurement, and particularly relates to an infrared temperature measurement method for selecting an optimal regression model. Background Art

[0002] All objects in nature with a temperature higher than absolute zero radiate infrared rays at all times. Infrared temperature measurement uses the thermal radiation of the objects in the scene itself to perform infrared imaging on the target and display the temperature. With the development of industry, agriculture, national defense, and medicine, the requirements for temperature measurement are getting higher and higher. For example, when the equipment is powered on, temperature measurement of mechanical equipment, electrical equipment, production equipment, etc., and temperature measurement of products during the production process or in the warehouse without causing product pollution or damage are increasingly demanded. Infrared temperature measurement, as a non-contact and non-destructive temperature measurement technology, has received extensive attention.

[0003] When an infrared thermal imager images an object, the object surface radiates infrared rays outward. After passing through the optical system, the radiation energy representing the object surface temperature is concentrated on the detector. The amplitude of the electrical signal output by the detector is proportional to the magnitude of the input radiation energy. After signal processing and calculation, a thermal image corresponding to the object surface temperature distribution is displayed on the display.

[0004] Since the temperature measurement accuracy of infrared temperature measurement is related to various factors such as object emissivity, environmental radiation, infrared detector target surface temperature, infrared thermal imager response non-uniformity and gray drift, calibration and data processing algorithms, etc. It can be seen that it is difficult to obtain the object surface temperature with high precision and high accuracy, and there are also many error terms, which affect the application of infrared temperature measurement in many technical fields. Therefore, how to effectively reduce the influence of various factors and improve the accuracy of infrared imaging temperature measurement is the top priority in solving the application in this field.

[0005] Currently, the research on object emissivity, infrared thermal imager response non-uniformity and gray drift has become increasingly mature, and the methods adopted in aspects such as calibration and data processing algorithms, and the influence of environmental radiation and infrared detector target surface temperature on infrared temperature measurement data are roughly as follows.

[0006] The two-dimensional image captured by the infrared thermal imager represents the thermal radiation distribution on the target surface and also represents the temperature information of the target. However, since the relationship between the target temperature and the radiation energy received by the detector is non-linear rather than a simple linear relationship, and is also affected by factors such as the ambient temperature, the surface emissivity of the target, and the temperature of the detector's target surface, it is impossible to directly obtain the quantitative temperature value of the target from the thermal image output by the infrared imaging system, but only a qualitative representation. In order to obtain the absolute temperature of the target based on the thermal image of the target, it is necessary to establish the corresponding relationship between the image gray level and the target temperature through calibration. Calibration generally uses a standard radiation source (usually a black body with high precision and high emissivity) as the reference source, uses the thermal imaging system to collect radiation images at different temperatures, and then fits the relationship curve between the image gray level and the temperature according to the gray level of the radiation image and the true temperature value of the black body. In actual temperature measurement applications, the absolute temperature of the target can be calculated according to the calibrated relationship curve and the gray level value of the collected target, realizing temperature measurement.

[0007] Nowadays, there are mainly two common calibration methods: the near-distance extended source method and the far-distance small source method. The near-distance extended source method requires that the area of the calibration source must fill the entire field of view of the infrared imaging system. In the far-distance small source method, the calibrated black body source is placed at a sufficient distance from the thermal imager so that it is within the field of view of the thermal imager and can be clearly imaged, but it cannot fill the field of view. In order to reduce the absorption of infrared rays by the atmosphere and the influence of the non-uniformity factor of the infrared thermal imager's response, the near-distance extended source method is generally used.

[0008] And there are mainly two common data processing methods: the fitting curve method and the look-up table method. In the fitting curve method, the least squares method is used to process the collected sample data to obtain a fitting curve of the corresponding relationship between temperature and gray level. The fitting curve method has the advantages of fewer calibration sample points and less calibration workload, but because there will be errors when fitting the curve, it will affect the improvement of temperature measurement accuracy. The look-up table method is to establish a database of calibration sample points, that is, a look-up table. When actually measuring the temperature, the temperature of the target can be obtained by looking up the table according to the gray level value of the obtained target image. The look-up table method must establish enough sample points to achieve high accuracy, so it is difficult to establish a sample database and it takes a long time. Especially when the temperature measurement range is relatively wide, the calibration period will be very long.

[0009] Most of the existing infrared temperature measurements use fixed polynomial model functions, which can basically only meet one type of detector; and when the following situations occur, the function model needs to be continuously adjusted to achieve higher accuracy.

[0010] 1. Due to the limitations of technology and process, different production batches of the same type of detector will show inconsistencies, and in individual cases, the temperature characteristics will vary greatly, and the function model needs to be adjusted.

[0011] 2. Different manufacturers and models of detectors on the market have different materials. For example, amorphous silicon and vanadium oxide have different temperature characteristics, and the function model also needs to be adjusted.

[0012] 3. When the temperature measurement range of the thermal imager needs to be changed, the detector parameters need to be readjusted, which will also affect the detector function model. Summary of the Invention

[0013] The purpose of the present invention is to overcome the deficiencies of the prior art and provide an infrared temperature measurement method for selecting the optimal regression model.

[0014] The purpose of the present invention is achieved through the following technical solutions:

[0015] An infrared temperature measurement method for selecting the optimal regression model includes the following specific steps:

[0016] S1: Collect thermal imager images and focal plane temperature data;

[0017] S2: Establish a set of F(TP) functions of polynomials to be fitted according to the focal plane temperature TP, which is used to describe the law of the thermal imager fixed temperature image value changing with the focal plane temperature;

[0018] S3: Establish a set of F(T, TP) functions of polynomials according to the target temperature T and the focal plane temperature TP, which is used to describe the law of the thermal imager responsivity changing with the target temperature and the focal plane temperature;

[0019] S4: Process the F(TP) function group and the F(T, TP) function group respectively to obtain the parameter set and the correlation coefficient R value of each function. The R value explains the linear correlation degree of each variable, and the value range is [0, +1]. The closer it is to +1, the higher the correlation degree; in this article, it represents the coincidence degree between the function model and the data model;

[0020] S5: Set the threshold thr_R1 and the threshold thr_R2;

[0021] S6: By comparing the R values of the F(TP) function group and the F(T, TP) function group with the set thresholds, judge to obtain the optimal function model and the optimal function parameters;

[0022] S7: Download the obtained optimal function model and the optimal function parameters into the thermal imager, obtain the current focal plane temperature TP and the current target temperature image value Y of the thermal imager, and obtain the target temperature T to complete the mapping from the image value to the temperature value.

[0023] The set of F(TP) functions of polynomials is expressed as:

[0024]

[0025] Where n is a positive integer and serves as the parameter input during parameter tuning; a is a model parameter, tp is the focal plane temperature value corresponding to the current target temperature acquisition; i is the order of the input function.

[0026] The polynomial F(T, TP) function group is expressed as:

[0027]

[0028] Where i and j are positive integers and serve as the parameter input during parameter tuning; p is a model parameter; n is the highest order of the current model; t is the target temperature value; tp is the focal plane temperature value corresponding to the current target temperature acquisition.

[0029] The specific step S4 is: Through the calculation of least squares, obtain the parameter sets A and P of the F(TP) and F(T, TP) function groups; Through the following algorithm, obtain the R values of the F(TP) and F(T, TP) function groups:

[0030]

[0031]

[0032]

[0033] Where SSR is the sum of squared residuals between the sample and the model; SST is the sum of squared model functions; R_square is the correlation coefficient between the sample and the model (abbreviated as R value in this article); yb is a fixed temperature image value sample; m is the number of samples; f(tp) and f(t, tp) are the function values from the F(TP) and F(T, TP) function groups.

[0034] According to the size of n in the function group, it determines the size of the R value set at this time. Compare each element in the R value set with the thresholds thr_R1 and thr_R2, and the function corresponding to the R value closest to the threshold is the best function of the current model.

[0035] The specific step S6 is: Judge the R value of F(TP), and the one closest to thr_R1 is the optimal F(TP) function, and the parameter set A is the optimal function parameter p_yb;

[0036] Judge the R value of F(T, TP), and the one closest to thr_R2 is the optimal F(T, TP) function, and the parameter set P is the optimal function parameter p_rv.

[0037] Step S7 is specifically as follows: Download the obtained function parameters and the model into the thermal imager. The thermal imager calculates the reference blackbody temperature value YB through the function F(TP) and the parameter p_yb, substitutes YB and the parameter p_rv into the function F(T, TP), and obtains the current focal plane temperature TP and the current target temperature image value Y of the thermal imager, and calculates the temperature T, thereby completing the mapping from the image value to the temperature value.

[0038] The thermal imager image and focal plane temperature data specifically include:

[0039] Focal plane temperature TP;

[0040] Blackbody temperature T = {T1, T2, T3...};

[0041] The blackbody temperature image value Y at each focal plane temperature;

[0042] The reference blackbody temperature image value YB at each focal plane temperature;

[0043] Reference blackbody temperature T REF .

[0044] The calculation of the temperature T specifically uses the following formula:

[0045] Advantages of the present invention:

[0046] 1. Use the R value to determine the matching degree of the model, calculate to judge the best matching model of the detector. If the detector types and temperature measurement ranges are different, the function order can be adjusted for calculation, so the algorithm has good compatibility;

[0047] 2. The output (optimal parameters) of the model function is only related to the order of the function, reducing the workload of algorithm parameter adjustment. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on the structures shown in these drawings without creative efforts.

[0049] Figure 1 It is the flowchart of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0051] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the fact that those of ordinary skill in the art can implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the protection scope required by the present invention.

[0052] As Figure 1 shown, an infrared temperature measurement method for selecting an optimal regression model includes the following specific steps:

[0053] S1: Collect thermal imager images and focal plane temperature data;

[0054] S2: By inputting the function order i, determine the function group YB = F(TP) of the polynomial to be fitted,

[0055]

[0056]

[0057] where n is a positive integer, used as a parameter input during parameter adjustment, with a maximum value of the number of TP samples - 1, and a is a model parameter. Through the least squares calculation, the parameter set A of each fitting function is obtained as follows:

[0058] Transform the function f(tp) into matrix form f(tp) = TP * A, where A is a parameter vector of (n + 1) * 1, and TP is an m * (n + 1) sample matrix, and the matrix size is determined by i.

[0059]

[0060] Then the loss function is defined as The A that makes the loss function minimum is the optimal parameter of the fitting function. Regarding A as the independent variable of the function, taking its derivative and setting the derivative to 0, then there is So the parameter calculation formula is A = (TP T * TP) -1 * TP T * YB.

[0061] S3: Set a threshold thr_R1 and a threshold thr_R2 according to the data in step S1, and calculate the R value of the function F(TP, YB);

[0062] S4: Calculate the R = {R1, R2... R n} values for each function in F(TP, YB).

[0063]

[0064] Where:

[0065] The one closest to thr_R1 is the optimal YB = F(TP) function, and A is the optimal function parameter p_yb;

[0066] S5: Calculate the ratio of the difference between the values of different blackbody temperature images and the reference blackbody image value to the temperature difference to obtain the change in the image response rate at different focal plane temperatures;

[0067]

[0068] S6: In the study of the detector's characteristics, it is found that Rv changes with the change of the target temperature T and the focal plane temperature TP. Therefore, a set of polynomial functions

[0069]

[0070]

[0071] is established. Among them, i and j are positive integers, which are used as parameter inputs during parameter adjustment. The maximum value is the number of samples - 1, and p is the model parameter. Through the least squares calculation, the parameter set P of each fitting function is obtained. The process is as follows:

[0072] Similar to step 3, transform the function f(t, tp) into matrix form f(t, tp) = S * P, where P is a (n + 1) * 1 parameter vector, and S is the sample matrix of T and TP, determined by i and j.

[0073]

[0074] Then the loss function is defined as The P that minimizes the loss function is the best parameter of the fitting function. Regarding P as the independent variable of the function, take its derivative and set the derivative to 0, then there is So the parameter calculation formula is P = (s T * s) -1 * S T * Rv;

[0075] S7: Similar to step 4, calculate the R = {R1, R2... R n} values of the function group F(T, TP)

[0076]

[0077] Where:

[0078] The model closest to thr_R2 is the optimal F(T, TP) function, where P is the optimal function parameter p_rv;

[0079] S8: Download the obtained function parameters and the model into the thermal imager. The thermal imager calculates the reference blackbody temperature value YB through the function F(TP) and the parameter p_y. Substitute YB and the parameter p_rv into the following responsivity calculation equation, and obtain the current focal plane temperature TP and the current target temperature image value Y, T REF as input parameters, set to the same temperature as the reference blackbody used during calibration, and finally obtain T in the equation, thus completing the mapping from the image value to the temperature value.

[0080]

[0081] The above are only the preferred embodiments of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein, should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications, and environments, and can be changed within the scope of the concept described herein through the above teachings or the techniques or knowledge in related fields. Any changes and variations made by those skilled in the art without departing from the spirit and scope of the present invention shall fall within the protection scope of the appended claims of the present invention.

Claims

1. An infrared temperature measurement method for selecting an optimal regression model, characterized in that It includes the following specific steps: S1: Collect the thermal imager images and the focal plane temperature data; S2: Establish a polynomial to be fitted according to the focal plane temperature function group; S3: Establish a polynomial function group based on the target temperature and the focal plane temperature; S4: Process separately the function groups and the function groups to obtain the parameter sets and correlation coefficient R values of each function. Specifically: Through the calculation of least squares, obtain and the parameter sets A and P of the function groups, including: Define the loss functions Loss(A) and Loss(P), and the A and P that minimize the above loss functions are the optimal parameters of the fitting function respectively; R is the correlation coefficient between the sample and the model, which is the ratio of the sum of the squared residuals between the sample and the model to the sum of the squared model functions; S5: Set the threshold thr_R1 and the threshold thr_R2; S6: By comparing the function groups and the R values of the function groups with the set threshold, the optimal function and the optimal function parameters are obtained. Specifically: judge the R value of the function group, and the f(tp) that makes this R value closest to thr_R1 is the optimal function of the function group, and the parameter set A is the optimal function parameter p_yb of the optimal function; judge the R value of the function group, and the f(t, tp) that makes this R value closest to thr_R2 is the optimal function of the function group, and the parameter set P is the optimal function parameter p_rv of the optimal function; S7: Download the obtained optimal function and optimal function parameters to the thermal imager, obtain the current focal plane temperature of the thermal imager and the current target temperature image value Y, and calculate the target temperature T, thus completing the mapping from the image value to the temperature value. Specifically: Download the obtained optimal function parameters and the optimal function to the thermal imager. The thermal imager calculates the reference blackbody temperature image value YB through the optimal function f(tp) and the optimal function parameter p_yb. Substitute YB = F(TP) into the left side of the response rate calculation equation, substitute the optimal function f(t, tp) and the optimal function parameter p_rv into the right side of the response rate calculation equation, and obtain the current focal plane temperature of the thermal imager and the current target temperature image value Y, and the reference blackbody temperature T REF is the input parameter, set to the same temperature as the reference blackbody used during calibration. Finally, calculate T in the response rate calculation equation, thereby completing the mapping from the image value to the temperature value. The specific response rate calculation equation is: .

Images