Imaging colorimeter and light measurement method

Abstract

The present invention provides an imaging colorimeter and a light measurement method, the colorimeter comprising: an object to be measured, for emitting measurement light; a light receiving unit, for receiving the measurement light emitted by the object to be measured; a dual optical path system, for dividing the measurement light received by the light receiving unit into a first light beam and a second light beam; an industrial camera system, for receiving the first light beam, and obtaining pixel information of the object to be measured; a fiber coupling system, for receiving the second light beam, and introducing the second light beam into a spectral analysis device; a spectral analysis device, for analyzing and processing the second light beam received by the fiber coupling system, and obtaining spectral information of the object to be measured; an algorithm control system, for receiving and storing pixel information and spectral information, calibrating the pixel information based on the spectral information, obtaining target colorimetric information of the object to be measured that meets preset conditions, and completing calibration. The scheme adopts the form of dual optical paths and algorithm control, calibrates the industrial camera system based on the spectral analysis device, and can quickly and accurately perform light measurement.

Description

Technical Field

[0001] The present invention relates to the technical field of colorimeters, and in particular to an imaging colorimeter and a light measurement method. Background Art

[0002] With the development of society, various display devices need to improve their display performance according to the growing requirements, with a wider color gamut and higher brightness. Therefore, related measurement equipment is further required to perform more accurate imaging measurement of its chromaticity, brightness and other photochromaticity parameters.

[0003] The imaging colorimeter is the most commonly used imaging brightness and colorimetry device on the market. There are two types of imaging colorimeters in principle. One is a colorimeter with a color filter device. The rotary imaging colorimeter using an XYZ filter requires multiple imaging and the XYZ filter has poor accuracy, which will increase the error. The other is to use a single-point spectrometer to measure the luminous spectrum information of the object being measured. This type of imaging colorimeter has many defects. First, this type of imaging colorimeter uses the spectrum of the center point to map and calibrate the remaining areas, but the spectrum of other areas other than the center point cannot be directly measured, resulting in large measurement errors in areas other than the center area. Secondly, the micro-spectrometers used in this type of imaging colorimeter generally have poor long-term stability, which means that this type of imaging colorimeter usually needs frequent calibration and is inconvenient to use. Summary of the invention

[0004] In view of the defects in the prior art, the purpose of the present invention is to provide an imaging colorimeter and a light measurement method, which can quickly and accurately perform light measurement by adopting a dual light path form and algorithm control and calibrating an industrial camera system based on a spectral analysis device.

[0005] The present invention provides an imaging colorimeter, comprising:

[0006] The object to be measured is used to emit the measuring light;

[0007] A light receiving unit, used for receiving the measuring light emitted by the object to be measured;

[0008] A dual-light path system, used for dividing the measuring light received by the light receiving unit into a first light beam and a second light beam;

[0009] An industrial camera system, used for receiving the first light beam and obtaining pixel information of the object to be measured;

[0010] an optical fiber coupling system, used for receiving the second light beam and guiding the second light beam into a spectrum analysis device;

[0011] A spectrum analysis device, used for analyzing and processing the second light beam received by the optical fiber coupling system to obtain spectrum information of the object under test;

[0012] The algorithm control system is used to receive and store the pixel information and the spectral information, calibrate the pixel information based on the spectral information, obtain the target colorimetric information of the object under test that meets the preset conditions, and complete the calibration.

[0013] Furthermore, the algorithm control system includes:

[0014] A storage module, used for receiving and storing the pixel information and the spectrum information;

[0015] A first conversion module, used for converting the pixel information into first chromaticity information;

[0016] A second conversion module, used for converting the spectral information into second colorimetric information;

[0017] A calibration module, configured to perform comparative analysis based on the first colorimetric information and the second colorimetric information to obtain an analysis result;

[0018] A determination module, used to determine whether the analysis result meets the preset condition;

[0019] A determination module, configured to use the analysis result as the target colorimetric information if the analysis result meets the preset condition;

[0020] The transmission module is used to output and visualize the target colorimetric information.

[0021] Furthermore, the comparative analysis of the calibration module adopts a multi-step approximation fitting algorithm.

[0022] Furthermore, the multi-step approximation fitting algorithm includes at least a three-step approximation fitting algorithm; the three-step approximation fitting algorithm includes:

[0023] The first step approximation fitting algorithm is, but not limited to, the double Newton algorithm;

[0024] The second step approximation fitting algorithm is, but not limited to, the generalized linear regression algorithm;

[0025] The third step approximation fitting algorithm is, but not limited to, the gradient boosting tree algorithm.

[0026] Furthermore, the algorithm control system also includes:

[0027] The circulation module is used to restart the calibration module to perform another comparative analysis when the analysis result does not meet the preset condition.

[0028] Furthermore, the light receiving part is also provided with an optical module and an aperture module, and the optical module is fixedly connected to the aperture module, wherein:

[0029] The optical module is used to change the imaging surface position of the object to be measured;

[0030] The aperture module is used to change the field angle and aperture size of the object to be measured.

[0031] The present invention also provides a light measurement method, comprising:

[0032] The light receiving part receives the measuring light emitted by the object to be measured;

[0033] The light receiving unit transmits the measuring light to the dual light path system;

[0034] The dual-light path system divides the measurement light into a first light beam and a second light beam, and transmits the light beams to the industrial camera system and the optical fiber coupling system respectively;

[0035] The industrial camera system receives the first light beam and analyzes to obtain pixel information of the object under test. Meanwhile, the optical fiber coupling system receives the second light beam and guides the second light beam into the spectrum analysis device, so that the spectrum analysis device analyzes the second light beam to obtain spectrum information of the object under test.

[0036] The algorithm control system receives and stores the pixel information and the spectral information, and calibrates the pixel information based on the spectral information to obtain the target colorimetric information of the object under test that meets preset conditions, thereby completing the calibration.

[0037] Further, the algorithm control system receives and stores the pixel information and the spectral information, and calibrates the pixel information based on the spectral information to obtain target colorimetric information of the object under test that meets preset conditions, and completes calibration, including:

[0038] The storage module receives and stores the pixel information and the spectrum information;

[0039] A first conversion module obtains the pixel information from the storage module and converts the pixel information into first chromaticity information;

[0040] A second conversion module obtains the spectral information from the storage module and converts the spectral information into second chromaticity information;

[0041] The calibration module obtains the first colorimetric information and the second colorimetric information from the first conversion module and the second conversion module respectively, and compares and analyzes the first colorimetric information and the second colorimetric information to obtain an analysis result;

[0042] The determination module obtains the analysis result from the calibration module and determines whether the analysis result meets the preset condition;

[0043] The determination module obtains the analysis result and the determination result from the calibration module and the determination module respectively, and when the determination result indicates that the analysis result meets the preset condition, uses the analysis result as the target colorimetric information to complete the calibration;

[0044] The transmission module obtains the target colorimetric information from the determination module, and outputs and visualizes the target colorimetric information.

[0045] Further, the algorithm control system receives and stores the pixel information and the spectral information, and calibrates the pixel information based on the spectral information to obtain the target colorimetric information of the object under test that meets the preset conditions, and completes the calibration, and also includes:

[0046] The determination module obtains the analysis result and the determination result from the calibration module and the determination module respectively, and starts the circulation module when the determination result indicates that the analysis result does not meet the preset condition;

[0047] The circulation module restarts the calibration module to perform another comparative analysis.

[0048] Furthermore, the light receiving unit receives the measuring light emitted by the object to be measured, including:

[0049] Adjusting the optical module and the aperture module on the light receiving part;

[0050] The light receiving unit receives the measuring light emitted by the object to be measured.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] The imaging colorimeter provided by the present invention adopts a dual light path form and algorithm control, and calibrates an industrial camera system based on a spectral analysis device, can quickly and accurately perform light measurement, obtain precise colorimetric information, and does not require multiple measurements, thereby reducing measurement costs.

[0053] This light measurement method is implemented by the above-mentioned imaging colorimeter, and because this light measurement method includes the technical solution of the above-mentioned imaging colorimeter, it has corresponding technical effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0055] Figure 1 A schematic structural diagram of an imaging colorimeter provided by an embodiment of the present invention;

[0056] Figure 2 A schematic flow chart of a light measurement method provided by an embodiment of the present invention;

[0057] Figure 3 A training flow chart of an algorithm control system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0058] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0059] An imaging colorimeter provided in an embodiment of the present application includes:

[0060] The object to be measured is used to emit the measuring light;

[0061] A light receiving unit, used to receive the measuring light emitted by the object to be measured; optionally, the light receiving unit adopts an objective lens, which receives the measuring light and forms an image of it;

[0062] A dual-light path system, used for splitting the measurement light received by the light receiving unit into a first light beam and a second light beam; optionally, the dual-light path system adopts a visible light band beam splitter or a single-piece reflector, preferably a visible light band beam splitter;

[0063] An industrial camera system, used for receiving the first light beam and obtaining pixel information of the object to be measured;

[0064] A fiber coupling system, used for receiving the second light beam and introducing the second light beam into the spectrum analysis device; optionally, the fiber coupling system adopts a fiber coupler;

[0065] A spectrum analysis device, used for analyzing and processing the second light beam received by the optical fiber coupling system to obtain spectrum information of the object under test; optionally, the spectrum analysis device adopts a spectrometer, preferably a micro-spectrometer;

[0066] The algorithm control system is used to receive and store pixel information and spectral information, calibrate the pixel information based on the spectral information, obtain the target colorimetric information of the object under test that meets the preset conditions, and complete the calibration.

[0067] Among them, the algorithm control system includes:

[0068] A storage module, used for receiving and storing pixel information and spectral information;

[0069] A first conversion module, used for converting pixel information into first chromaticity information;

[0070] A second conversion module, used for converting the spectral information into second colorimetric information;

[0071] A calibration module, used for performing comparative analysis based on the first colorimetric information and the second colorimetric information to obtain an analysis result;

[0072] A determination module is used to determine whether the analysis result meets the preset conditions;

[0073] A determination module, used to use the analysis result as target colorimetric information when the analysis result meets the preset conditions;

[0074] The loop module is used to restart the calibration module for another comparative analysis if the analysis result does not meet the preset conditions;

[0075] The transmission module is used to output and visualize the target colorimetric information.

[0076] In an optional embodiment, the comparative analysis of the calibration module adopts a multi-step approximation fitting algorithm.

[0077] Preferably, the multi-step approximation fitting algorithm includes at least a three-step approximation fitting algorithm; the three-step approximation fitting algorithm may include:

[0078] The first step approximation fitting algorithm is, but not limited to, the double Newton algorithm;

[0079] The second step approximation fitting algorithm is, but not limited to, the generalized linear regression algorithm;

[0080] The third step approximation fitting algorithm is, but not limited to, the gradient boosting tree algorithm.

[0081] In an optional embodiment, an optical module and an aperture module are further provided on the light receiving part, and the optical module is fixedly connected to the aperture module, wherein:

[0082] An optical module, used to change the imaging surface position of the object to be measured; optionally, the optical module adopts an optical module with variable focal length;

[0083] The aperture module is used to change the field of view angle and aperture size of the object being measured.

[0084] See also Figure 1 In a specific embodiment, the imaging colorimeter comprises:

[0085] The object to be measured 1 is used to emit measuring light;

[0086] The light receiving unit 2 is used to receive the measuring light emitted by the object to be measured 1;

[0087] A dual optical path system 3, used for dividing the measuring light received by the light receiving unit 2 into a first light beam and a second light beam;

[0088] The industrial camera system 4 is used to receive the first light beam and obtain pixel information of the object 1;

[0089] The optical fiber coupling system 5 is used to receive the second light beam and guide the second light beam into the micro-spectral analysis device 6;

[0090] A micro-spectral analysis device 6 is used to analyze and process the second light beam received by the optical fiber coupling system 5 to obtain spectrum information of the object 1;

[0091] The algorithm storage control system 7 is used to receive and store pixel information and spectral information, calibrate the pixel information based on the spectral information, obtain the target colorimetric information of the object under test that meets the preset conditions, and complete the calibration.

[0092] In this embodiment, the objective lens receives the measuring light and images the object to be measured, the dual optical path system receives the measuring light from the objective lens and forms a dual optical path, which is divided into a first light beam and a second light beam, the industrial camera system receives the first light beam, obtains an image and extracts RGB information, the optical fiber coupler receives the second light beam and transmits it to the spectrometer, the second light beam is the light emitted from the center point of the object to be measured, the spectrometer extracts and detects the spectral information of the second light beam, the algorithm control system receives and stores the spectral information obtained by the spectrometer and the pixel information obtained by the industrial camera system, specifically the values ​​of different colors in the color gamut triangle measured by the spectrometer and the RGB information of the same position measured by the industrial camera system, and the algorithm control system converts the industrial camera system into an image. The RGB information measured by the industrial camera system is extracted and processed, and compared with the data measured by the spectrometer. Then, based on the two, the algorithm corresponding to the RGB colorimetric parameters XYZ is used for comparison and fitting. During the comparison and fitting, a three-step approximation fitting algorithm is used, such as the least squares method (the first step approximation fitting algorithm), the generalized linear regression algorithm (the second step approximation fitting algorithm) and the gradient ascent tree algorithm (the third step approximation fitting algorithm) to calibrate the RGB data measured by the industrial camera system. When the analysis result after calibration meets the preset conditions, the calibrated color coordinates Cx and Cy data are obtained, and the calibrated color coordinates Cx and Cy data are stored to complete the mapping from RGB to XYZ.

[0093] See also Figure 2 , which shows a light measurement method provided by an embodiment of the present application, comprising:

[0094] Step S101: the light receiving unit receives the measuring light emitted by the object to be measured;

[0095] Step S102: the light receiving unit transmits the measuring light to the dual optical path system;

[0096] Step S103: the dual-optical path system divides the measurement light into a first light beam and a second light beam, and transmits the light beams to the industrial camera system and the optical fiber coupling system respectively;

[0097] Step S104: the industrial camera system receives the first light beam and analyzes to obtain pixel information of the object under test. Meanwhile, the optical fiber coupling system receives the second light beam and guides the second light beam into the spectrum analysis device, so that the spectrum analysis device analyzes the second light beam to obtain spectrum information of the object under test.

[0098] Step S105: the algorithm control system receives and stores the pixel information and the spectral information, and calibrates the pixel information based on the spectral information to obtain the target colorimetric information of the object under test that meets the preset conditions, thereby completing the calibration.

[0099] In an optional embodiment, step S105 may include:

[0100] Step 201: The storage module receives and stores pixel information and spectrum information;

[0101] Step S202: the first conversion module obtains pixel information from the storage module and converts the pixel information into first chromaticity information;

[0102] Step S203: the second conversion module obtains the spectrum information from the storage module, and converts the spectrum information into second colorimetric information;

[0103] Step S204: The calibration module obtains the first colorimetric information and the second colorimetric information from the first conversion module and the second conversion module respectively, and compares and analyzes the first colorimetric information and the second colorimetric information to obtain an analysis result;

[0104] Step S205: the determination module obtains the analysis result from the calibration module and determines whether the analysis result meets the preset conditions;

[0105] Step S2061: the determination module obtains the analysis result and the determination result from the calibration module and the determination module respectively, and when the determination result indicates that the analysis result meets the preset condition, the analysis result is used as the target colorimetric information to complete the calibration;

[0106] Step S207: the transmission module obtains the target colorimetric information from the determination module, and outputs and visualizes the target colorimetric information.

[0107] In an optional embodiment, step S105 may further include:

[0108] Step S2062: the determination module obtains the analysis result and the determination result from the calibration module and the determination module respectively, and starts the circulation module when the determination result indicates that the analysis result does not meet the preset condition;

[0109] Step S2063: The circulation module restarts the calibration module to perform another comparative analysis.

[0110] In an optional embodiment, the analysis and comparison of the calibration module adopts a multi-step approximation fitting algorithm.

[0111] Preferably, the multi-step approximation fitting algorithm includes at least a three-step approximation fitting algorithm; the three-step approximation fitting algorithm may include:

[0112] The first step approximation fitting algorithm is, but not limited to, the double Newton algorithm;

[0113] The second step approximation fitting algorithm is, but not limited to, the generalized linear regression algorithm;

[0114] The third step approximation fitting algorithm is, but not limited to, the gradient boosting tree algorithm.

[0115] In an optional embodiment, step S101 may include:

[0116] Adjusting the optical module and the aperture module on the light receiving unit;

[0117] The light receiving unit receives the measurement light emitted by the object to be measured.

[0118] The embodiment of the present invention also provides a training method for an algorithm control system, comprising:

[0119] The light receiving part receives the reference light source emitted by the object to be measured;

[0120] The light receiving unit transmits the reference light source to the dual light path system;

[0121] The dual-optical path system divides the reference light source into a first reference beam and a second reference beam, and transmits them to the industrial camera system and the optical fiber coupling system respectively;

[0122] The industrial camera system receives the first reference beam and analyzes to obtain reference pixel information of the object under test. At the same time, the fiber coupling system receives the second reference beam and introduces the second reference beam into the spectral analysis device, so that the spectral analysis device analyzes and obtains reference spectral information of the object under test based on the second reference beam. Optionally, the spectral analysis device uses standard colorimetry measurement equipment, including but not limited to: point colorimeter, imaging colorimeter, spectroradiometer, imaging spectroradiometer.

[0123] The sample algorithm control system receives and stores reference pixel information and reference spectrum information, and obtains first reference colorimetry information based on the reference pixel information and obtains second reference colorimetry information based on the reference spectrum information;

[0124] The sample algorithm control system will calibrate and fit the second reference colorimetric information based on the first reference colorimetric information in accordance with a multi-step approximation fitting algorithm, and adjust the calibration parameters of the sample algorithm control system according to the fitting situation until the target reference colorimetric information of the object under test that meets the preset conditions is obtained, thereby completing the calibration.

[0125] See also Figure 3 , which is a training flow chart of an algorithm control system provided by an embodiment of the present invention.

[0126] In a specific embodiment, the light emitted by the light source of the object to be measured is first converged by an imaging lens located at the front end of the colorimeter to form a clear and accurate image on the focal plane of the lens;

[0127] The clear image is divided into two images, called image A and image B, after passing through a semi-transparent and semi-reflective dual-path beam splitter placed behind the imaging objective lens. The two images propagate in two optical paths respectively. Image A is focused on the CMOS array of the industrial camera through a converging lens group in optical path 1. The pixel points on the array correspond to the image points on the image of the object being measured.

[0128] In front of the CMOS area array camera, there is a Bayer filter that can selectively transmit red, green, and blue specific light wavelengths. After the Bayer filter filters the light, each pixel on the CMOS photosensitive chip produces a different induction intensity. After photoelectric conversion and digital-to-analog conversion, the original image with specific RGB values ​​is output. By selecting a part of the image, you can use the point distribution tool to calculate the pixel values ​​of the three channels of R, G, and B;

[0129] The image B formed by the beam splitter passes through the convergent lens in the optical path 2 and is received by the light receiver, and then propagates through the optical fiber to the inside of the spectrometer, which has a slit for receiving the transmitted light and a grating for splitting the light. The image B is processed by the internal splitting of the spectrometer and the data integration of the built-in algorithm, and the light intensity value of each wavelength in the spectrum data is divided by the maximum intensity value to normalize the intensity values ​​at all wavelengths. Then, the normalized spectrum data is mapped to the CIE XYZ chromaticity coordinate space through the CIE chromaticity diagram algorithm. The CIE XYZ coordinates describe the three basic attributes of color: brightness, chromaticity and hue. Therefore, the light signal input to the spectrometer 6 is converted into the three stimulus values ​​XYZ describing the color and the standard Cx and Cy values ​​through the above steps.

[0130] Then, in the storage and calculation processing module, the original TOC algorithm of the present invention is used to map and cross-validate the three-channel pixel values ​​(R, G, B) and the three stimulus values ​​(X, Y, Z) corresponding to the two optical paths, and directly display the chromaticity parameters corresponding to the pixels in the selected area on the image formed by the industrial camera, including the color coordinates (Cx, Cy) and the three stimulus values ​​(X, Y, Z).

[0131] The process and logic of the above TOC algorithm are as follows:

[0132] The idea of ​​TOC algorithm is to use a large amount of appropriate and comprehensive measured data to establish an accurate mathematical model, so as to accurately predict the specified colorimetric parameters.

[0133] First, test data is selected for the area to be measured in the color gamut triangle. In the embodiment, a triangular area including the entire sRGB color gamut triangle is specified for point arrangement. By selecting 800 feature points of different colors inside the selected color gamut triangle area, the color coordinates Cx and Cy specified by the feature points are obtained.

[0134] After the color coordinates are determined, a standard uniform light field is used to emit accurate colors. Each color has a fixed 5 brightness values. The uniform light field is used as the above light source, and the uniform light field is imaged using the above light splitting method. Each brightness value of each color is measured in turn to obtain 4000 sets of one-to-one corresponding color three-channel (R, G, B) values ​​and three stimulus values ​​(X, Y, Z);

[0135] The (4000×3) matrix composed of 2500 RGB values ​​is recorded as matrix R;

[0136] The (4000×3) matrix composed of 2500 tristimulus values ​​is recorded as matrix X;

[0137] Then, in the data preprocessing stage, in order to ensure the reliability and generalization ability of the model, we divide the matrix R and matrix X into training set (R 1 , X 1 ) and the test set (R 2 ,X 2 ), and a data volume ratio of 4:1 is used. This ratio is chosen to make full use of most of the data for model training, while retaining a small amount of data for subsequent testing and verification.

[0138] Training set (R 1 , X 1 ) is used for model learning and parameter adjustment to make it fit the training data better. 2 ,X 2) is a data subset independent of the training process and is used to evaluate the generalization performance of the model to new data. Such a strict division helps to avoid overfitting the training set during training, thereby improving its effectiveness in practical applications.

[0139] There is a standard matrix formula T for converting RGB to the three stimulus values ​​XYZ of the human eye. 0 :

[0140]

[0141] in

[0142] In the data set, this formula is used to calculate the matrix R composed of 3200 (R, G, B) three-channel values. 1 Perform the conversion to standard XYZ to obtain a set of initial target matrices A(X 0 ,Y 0 ,Z 0 ), to ensure the consistent representation of color information in the color space. Then, we introduce the standard least squares method. The least squares method is an optimization technique whose core principle is to find a fitting curve or plane that minimizes the residual sum of squares between the fitting value and the actual observation value. The residual is the difference between the observation value and the fitting value. By adjusting the parameters of the fitting curve or plane, the least squares method iteratively finds the best match to minimize the residual sum of squares. This ensures that the fitting result has good robustness to noise and changes, and improves the accuracy and reliability of the matrix B. By fitting the initial target value, the matrix B (X 1 ,Y 1 ,Z 1 ), where the transformation matrix is ​​T 1 :

[0143]

[0144] After the least squares fitting is completed, we perform linear regression fitting on the processed matrix B and the standard three-stimulus value matrix X to obtain the fitted result matrix C (X 2 ,Y 2 ,Z 2 ). The goal of this linear regression process is to build a model M 1 , the model can accurately describe the relationship between the matrix B and the standard tristimulus value matrix X.

[0145] Linear regression model M 1 By adjusting the coefficients, the data in the matrix B is best fitted so that the fitted values ​​are as close as possible to the actual values ​​of the standard tristimulus value matrix X. This method can be regarded as further adjusting the processed color information to fit the standard reference value.

[0146] The subtlety of the above process lies in that through multi-stage fitting and adjustment, we not only complete the accurate mapping of the color space, but also establish a linear regression model M 1 , the model can capture the more complex relationship between the matrix B and the standard tristimulus value matrix X. This provides further guarantee for the accuracy and reliability of the data and provides a more sophisticated foundation for subsequent applications.

[0147] After linear regression fitting, we used the gradient boosting tree algorithm to further optimize the matrix C, and introduced the genetic algorithm for parameter search. Finally, we used the test set R 2 ,X 2 to verify.

[0148] Gradient Ascent Tree is an ensemble learning algorithm that builds decision trees based on the gradient ascent optimization criterion. Unlike the traditional gradient descent method, the goal of the gradient ascent tree is to maximize the loss function rather than minimize it. The core idea of ​​the algorithm is to iteratively add simple decision trees, each time adding a tree to increase the value of the objective function (usually the loss function). In each round of iteration, the gradient ascent tree builds a new decision tree based on the residual of the current model (the difference between the actual value and the predicted value), and then adds the output of the tree to the model by gradient ascent. This process is repeated until the predetermined number of iterations is reached or the loss function converges.

[0149] Therefore, the advantages of gradient boosting trees are mainly reflected in the following aspects:

[0150] 1. Nonlinear relationship processing: Gradient boosting trees can adapt to nonlinear relationships more flexibly. Since the relationship between color information and human eye perception may be a complex nonlinear relationship, gradient boosting trees help better fit the data by iteratively adding trees, each tree focusing on capturing different patterns.

[0151] 2. Interaction effect capture: By combining multiple decision trees, the gradient boosting tree can effectively capture the interaction effects between features. In the processing of color information, there may be a mutual influence relationship between different channels, and the gradient boosting tree helps to capture these complex interaction effects.

[0152] 3. Robustness and generalization ability: Gradient boosting tree can improve the robustness of the model by continuously iterating and gradually improving the model fitting effect. This is critical for dealing with complex color information mapping problems in the real world, because color perception may be affected by many factors.

[0153] 4. Handling residuals: The construction process of gradient boosting trees is based on the residuals of the current model, which means that the goal of each tree is to capture the part that the previous model failed to explain. This strategy helps to gradually improve the performance of the model.

[0154] A genetic algorithm is an optimization algorithm that simulates natural selection and genetic mechanisms. It searches for the optimal solution in the parameter space by simulating operations such as gene mutation and crossover.

[0155] In this process, we encode the hyperparameters of the gradient boosting tree (n_est imators, learning_rate, max_depth) into the genes of individuals. Initially, a group of individuals is randomly generated to represent the parameter combination. Through genetic operations such as selection, crossover and mutation, new individuals are generated, and according to the fitness evaluation of the objective function, a population with higher fitness is gradually formed. This evolutionary process continues until the convergence condition is met, and finally the optimal individual and its corresponding parameter combination are obtained.

[0156] Through genetic algorithm search, we obtained the optimal hyperparameter combination as follows:

[0157] n_est imators:50,learning_rate:0.000000008,max_depth:1, and the corresponding gradient boosting model M 2 ;

[0158] Finally, in the testing phase, we use the test set (R 2 ,X 2 ) to conduct test verification to ensure that the error between the predicted XYZ value and the standard XYZ value is within 0.5% to verify the accuracy and generalization ability of the model.

[0159] So far, we have successfully established a complete model, which includes the color space conversion matrix T 1 , linear regression model M 1 And the gradient boosted tree model M 2 This series of complex and sophisticated algorithms ensures the high accuracy and reliability of the model in the process of color information mapping and fitting, providing strong support for subsequent applications.

[0160] This model and the measured data are stored in the calculation and processing unit. Before detecting the object to be measured, a calibration operation is required. Red, green, blue and white are used as the objects to be measured for calibration in turn. During calibration, the RGB data in optical path one is converted into XYZ data through this model and stored in the calculation and processing unit. At the same time, the spectrometer in optical path two displays the XYZ data of the object to be measured. The two data are compared and cross-validated to determine the accuracy of the model.

[0161] After the calibration is completed, the object to be measured is photographed directly, and the resulting image will directly display the chromaticity value parameters of the object to be measured. At this time, since the photo is taken at one time, the imaging speed is very fast, generally 0.1s-2s; and because of the support of a large amount of data and the accuracy of the model, the measured error is very small, not exceeding 0.5%, which meets the requirements of most manufacturers for colorimetric parameter testing.

[0162] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. An imaging colorimeter, characterized in that: include: The object to be measured is used to emit the measuring light; A light receiving unit, used for receiving the measuring light emitted by the object to be measured; A dual-light path system, used for dividing the measuring light received by the light receiving unit into a first light beam and a second light beam; An industrial camera system, used for receiving the first light beam and obtaining pixel information of the object to be measured; an optical fiber coupling system, used for receiving the second light beam and guiding the second light beam into a spectrum analysis device; A spectrum analysis device, used for analyzing and processing the second light beam received by the optical fiber coupling system to obtain spectrum information of the object under test; An algorithm control system, used for receiving and storing the pixel information and the spectral information, calibrating the pixel information based on the spectral information, obtaining target colorimetric information of the object under test that meets preset conditions, and completing calibration; The algorithm control system comprises: A storage module, used for receiving and storing the pixel information and the spectrum information; A first conversion module, used for converting the pixel information into first chromaticity information; A second conversion module, used for converting the spectral information into second colorimetric information; A calibration module, configured to perform comparative analysis based on the first colorimetric information and the second colorimetric information to obtain an analysis result; A determination module, used to determine whether the analysis result meets the preset condition; A determination module, configured to use the analysis result as the target colorimetric information if the analysis result meets the preset condition; A transmission module, used for outputting and visualizing the target colorimetric information; The comparative analysis of the calibration module adopts a multi-step approximation fitting algorithm; specifically, the multi-step approximation fitting algorithm includes a three-step approximation fitting algorithm, and the three-step approximation fitting algorithm includes: The first step of the approximate fitting algorithm is the least squares method; The second step approximation fitting algorithm is the generalized linear regression algorithm; The third step approximation fitting algorithm is the gradient boosting tree algorithm.

2. An imaging colorimeter according to claim 1, characterized in that: The algorithm control system also includes: The circulation module is used to restart the calibration module to perform another comparative analysis when the analysis result does not meet the preset condition.

3. An imaging colorimeter according to claim 1 or 2, characterized in that: The light receiving part is also provided with an optical module and an aperture module, and the optical module is fixedly connected to the aperture module, wherein: The optical module is used to change the imaging surface position of the object to be measured; The aperture module is used to change the field angle and aperture size of the object to be measured.

4. A light measurement method, applied to an imaging colorimeter as claimed in any one of claims 1 to 3, characterized in that: include: The light receiving part receives the measuring light emitted by the object to be measured; The light receiving unit transmits the measuring light to the dual light path system; The dual-light path system divides the measurement light into a first light beam and a second light beam, and transmits the light beams to the industrial camera system and the optical fiber coupling system respectively; The industrial camera system receives the first light beam and analyzes to obtain pixel information of the object under test. Meanwhile, the optical fiber coupling system receives the second light beam and guides the second light beam into the spectrum analysis device, so that the spectrum analysis device analyzes the second light beam to obtain spectrum information of the object under test. The algorithm control system receives and stores the pixel information and the spectral information, and calibrates the pixel information based on the spectral information to obtain the target colorimetric information of the object under test that meets preset conditions, thereby completing the calibration.

5. A light measurement method according to claim 4, characterized in that: The algorithm control system receives and stores the pixel information and the spectral information, and calibrates the pixel information based on the spectral information to obtain target colorimetric information of the object under test that meets preset conditions, and completes calibration, including: The storage module receives and stores the pixel information and the spectrum information; A first conversion module obtains the pixel information from the storage module and converts the pixel information into first chromaticity information; A second conversion module obtains the spectral information from the storage module and converts the spectral information into second chromaticity information; The calibration module obtains the first colorimetric information and the second colorimetric information from the first conversion module and the second conversion module respectively, and compares and analyzes the first colorimetric information and the second colorimetric information to obtain an analysis result; The determination module obtains the analysis result from the calibration module and determines whether the analysis result meets the preset condition; The determination module obtains the analysis result and the determination result from the calibration module and the determination module respectively, and when the determination result indicates that the analysis result meets the preset condition, uses the analysis result as the target colorimetric information to complete the calibration; The transmission module obtains the target colorimetric information from the determination module, and outputs and visualizes the target colorimetric information.

6. A light measurement method according to claim 5, characterized in that: The algorithm control system receives and stores the pixel information and the spectral information, and calibrates the pixel information based on the spectral information to obtain target colorimetric information of the object under test that meets preset conditions, and completes calibration, further comprising: The determination module obtains the analysis result and the determination result from the calibration module and the determination module respectively, and starts the circulation module when the determination result indicates that the analysis result does not meet the preset condition; The circulation module restarts the calibration module to perform another comparative analysis.

7. A light measurement method according to any one of claims 4 to 6, characterized in that: The light receiving unit receives the measuring light emitted by the object to be measured, including: Adjusting the optical module and the aperture module on the light receiving part; The light receiving unit receives the measuring light emitted by the object to be measured.

Images