Spectral model adaptive method based on standard calibration plate

Abstract

The present invention relates to the field of near-infrared spectrum processing. In order to solve the problem of poor consistency of miniaturized near-infrared spectrum equipment, a spectrum model adaptation method based on a standard calibration plate is provided, including: step 1, using N original near-infrared spectrometers of the same model to collect M labeled sample spectrum data, and constructing spectrum models according to the labeled sample spectrum data to form an original spectrum model library; step 2, using N original near-infrared spectrometers and near-infrared spectrometers to be matched to collect standard calibration plate spectrum data respectively, and matching the spectrum model for the near-infrared spectrometer to be matched in the original spectrum model library based on the collected spectrum data. The above steps do not require multiple modeling, and the equipment consistency is higher.

Description

Technical Field

[0001] The invention relates to the field of near-infrared spectrum processing, and in particular to a spectrum model self-adaptation method based on a standard calibration plate. Background Art

[0002] In recent years, near-infrared spectroscopy analysis technology has developed rapidly and has been applied in many fields such as chemical industry, pharmaceutical industry, military industry, food, etc. Near-infrared spectroscopy technology belongs to molecular spectroscopy technology, which can show the composition and properties of substances at the molecular level. It has achieved very high benefits in terms of both economic and social impact and has great development potential.

[0003] However, currently most material composition and property information detection is mainly carried out using large laboratory near-infrared spectroscopic instruments. Although these methods are quantitatively accurate and highly sensitive, the equipment required is bulky and expensive, the sample preparation time is long and the sample preparation method is strict. The detection equipment and sample preparation require professional operation, the detection environment is fixed, and the analysis time is long. They are not suitable for on-site detection and are not easy to promote and use.

[0004] With the development of micro-electromechanical technology, in recent years, miniaturized near-infrared spectrometers have been widely used in various fields such as food, beverages, medicine, coal, etc. Compared with the traditional large-scale Fourier transform spectral analysis system, it has the advantages of simple structure, low cost, easy to carry, and real-time visible results. It has become a popular product in the field of spectroscopy. However, compared with large-scale near-infrared spectroscopy equipment in laboratories, miniaturized near-infrared spectrometers are affected by light sources, detectors, manufacturing processes, usage methods, environmental conditions, etc., and the performance of equipment consistency, stability, reproducibility and other indicators has been significantly reduced, especially equipment consistency. Different miniaturized near-infrared spectroscopy equipment of the same model has poor consistency, and the spectral models established by different equipment are not universal. Therefore, in the actual application of miniaturized near-infrared spectroscopy analysis, for the same type of samples to be tested, each miniaturized near-infrared spectroscopy equipment needs to be modeled separately, which wastes a lot of manpower and material resources and greatly hinders the development of miniaturized near-infrared spectroscopy analysis technology. Summary of the invention

[0005] In order to solve the problems of poor consistency of miniaturized near-infrared spectroscopy equipment and non-universal spectral models, the present application provides a spectral model adaptation method based on a standard calibration plate.

[0006] The technical solution adopted by the present invention to solve the above problems is:

[0007] The spectral model adaptation method based on the standard calibration plate includes:

[0008] Step 1: Use N original near-infrared spectrometers of the same model to collect M labeled sample spectral data, and construct spectral models according to the labeled sample spectral data to form an original spectral model library;

[0009] Step 2: Use N original near-infrared spectrometers and the near-infrared spectrometer to be matched to respectively collect spectral data of the standard calibration plate, and match the spectral model for the near-infrared spectrometer to be matched in the original spectral model library based on the collected spectral data.

[0010] Furthermore, in step 2, N original near-infrared spectrometers and the near-infrared spectrometer to be matched are used to collect spectral data of Y kinds of standard calibration plates respectively, and Y is greater than 1.

[0011] Furthermore, the step 2 is specifically as follows:

[0012] Step 21, using N original near-infrared spectrometers and the near-infrared spectrometer to be matched to respectively collect spectral data of Y kinds of standard calibration plates;

[0013] Step 22: Map the spectral data of Y kinds of standard calibration plates corresponding to each near-infrared spectrometer to a single standard calibration plate in combination with the reflectivity of the standard calibration plate, and calculate the mean value of each near-infrared spectrometer according to the mapped value;

[0014] Step 23: Calculate the trend offset according to the mean value corresponding to each original near-infrared spectrometer and the mean value corresponding to the near-infrared spectrometer to be matched, and match the spectral model for the near-infrared spectrometer to be matched in the original spectral model library according to the trend offset.

[0015] Furthermore, the step 23 is specifically as follows:

[0016] Step 231, respectively calculating the offset coefficients between the two types of devices according to the mean values ​​corresponding to the original near-infrared spectrometers and the mean values ​​corresponding to the near-infrared spectrometers to be matched;

[0017] Step 232, correcting the mean value according to the offset coefficient;

[0018] Step 233: Use the corrected mean to calculate the trend offset, and match the spectral model for the near-infrared spectrometer to be matched in the original spectral model library according to the trend offset.

[0019] Furthermore, the calculation method of the offset coefficient β is: w is the number of spectral wavelength points, Sn is the mean value corresponding to the original near-infrared spectrometer, S' n is the mean value corresponding to the near-infrared spectrometer to be matched.

[0020] Furthermore, the trend offset is calculated as:

[0021] Furthermore, the method also includes step 3: correcting the unlabeled sample spectral data in combination with the offset coefficient, and using the matched spectral model to perform predictive analysis on the corrected spectral data.

[0022] Furthermore, the single standard calibration plate in step 22 is a standard calibration plate with a medium reflectivity among the Y types of standard calibration plates.

[0023] Furthermore, M is 150-250, and N is 15-25.

[0024] Furthermore, Y is 3, and the three standard calibration plates are: a black calibration plate, a white calibration plate, and a gray calibration plate.

[0025] Compared with the prior art, the present invention has the following beneficial effects: the method described in the present application first uses a plurality of miniaturized near-infrared spectrometers of the same model to respectively collect spectral data of labeled samples, establishes an original spectral model library, and then uses the same batch of miniaturized near-infrared spectrometers to collect spectral data of standard calibration plates, and matches the spectral model for the near-infrared spectrometer to be matched in the original spectral model library in combination with the spectral data of the standard calibration plates, without the need for re-modeling, and effectively solves the problems of poor consistency of miniaturized near-infrared spectral equipment and non-universality of spectral models. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a flow chart of the spectral model adaptation method based on the standard calibration plate;

[0027] Figure 2 Flow chart corresponding to the embodiment. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. 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.

[0029] like Figure 1 As shown, the spectral model adaptation method based on the standard calibration plate includes:

[0030] Step 1: Use N original near-infrared spectrometers of the same model to collect M labeled sample spectral data, and construct spectral models according to the labeled sample spectral data to form an original spectral model library;

[0031] Step 2: Use N original near-infrared spectrometers and the near-infrared spectrometer to be matched to respectively collect spectral data of the standard calibration plate, and match the spectral model for the near-infrared spectrometer to be matched in the original spectral model library based on the collected spectral data.

[0032] The original near-infrared spectrometer can be compared with the near-infrared spectrometer to be matched through the standard calibration plate. The spectral model can be matched for the near-infrared spectrometer to be matched based on the difference in data collected by the two, without the need for re-modeling.

[0033] In order to reduce the error caused by the standard calibration plate and improve the model matching accuracy, in step 2, N original near-infrared spectrometers and the near-infrared spectrometer to be matched are used to collect Y types of standard calibration plate spectral data respectively, and Y is greater than 1.

[0034] Specifically, the step 2 is as follows:

[0035] Step 21, using N original near-infrared spectrometers and the near-infrared spectrometer to be matched to respectively collect spectral data of Y kinds of standard calibration plates;

[0036] Step 22: Map the spectral data of Y kinds of standard calibration plates corresponding to each near-infrared spectrometer to a single standard calibration plate in combination with the reflectivity of the standard calibration plate, and calculate the mean value of each near-infrared spectrometer according to the mapped value;

[0037] Step 23: Calculate the trend offset according to the mean value corresponding to each original near-infrared spectrometer and the mean value corresponding to the near-infrared spectrometer to be matched, and match the spectral model for the near-infrared spectrometer to be matched in the original spectral model library according to the trend offset.

[0038] In order to eliminate the difference between the data collected by the original near-infrared spectrometer and the near-infrared spectrometer to be matched and further improve the model matching accuracy, the step 23 is specifically as follows:

[0039] Step 231, respectively calculating the offset coefficients between the two types of devices according to the mean values ​​corresponding to the original near-infrared spectrometers and the mean values ​​corresponding to the near-infrared spectrometers to be matched;

[0040] Step 232, correcting the mean value according to the offset coefficient;

[0041] Step 233: Use the corrected mean to calculate the trend offset, and match the spectral model for the near-infrared spectrometer to be matched in the original spectral model library according to the trend offset.

[0042] Specifically, the calculation method of the offset coefficient β is: w is the number of spectral wavelength points, Sn is the mean value corresponding to the original near-infrared spectrometer, S' n is the mean value corresponding to the near-infrared spectrometer to be matched. The trend offset is calculated as follows:

[0043] In order to make the predicted data more accurate, step 3 is also included: correcting the unlabeled sample spectral data in combination with the offset coefficient, and using the matched spectral model to perform predictive analysis on the corrected spectral data.

[0044] Preferably, the single standard calibration plate in step 22 is a standard calibration plate with a medium reflectivity among the Y standard calibration plates; M is 150-250, N is 15-25; Y is 3, and the three standard calibration plates are: black calibration plate, white calibration plate, and gray calibration plate.

[0045] Example

[0046] like Figure 2 As shown, step 101 is to use multiple miniaturized near-infrared spectrometers of the same model to respectively collect spectral data of labeled samples and unlabeled samples, and establish an original spectral model library according to the spectral model corresponding to the labeled samples. Multiple miniaturized near-infrared spectrometers of the same model are divided into two types of equipment, one is the original spectral model library establishment equipment, and its collected samples are samples with known calibration values; the other is the adaptive spectral model equipment, and its collected samples are samples with unknown calibration values, and are directly predicted by the model in the adaptive original spectral model library without the need for spectral modeling.

[0047] In this embodiment, multiple miniaturized near-infrared spectrometers of the same model are specifically classified as (N+1), wherein N devices are original spectral database establishment devices, and 1 device is an adaptive spectral model device for implementing the spectral model adaptation method. In specific practical applications, the number of original spectral database establishment devices is preferably 15-25, and 20 are selected in this embodiment. Too many devices will consume a lot of manpower and material resources for spectral data collection and modeling. Too few devices cover fewer spectral models, have lower adaptive accuracy, and larger prediction errors. The number of adaptive spectral model devices can be (1~+∞), and the spectral model adaptation of all devices can be achieved by processing them using the same spectral model adaptation method. The specific number of labeled samples and unlabeled samples is (M+1), where M are labeled samples and 1 is unlabeled sample. In specific practical applications, the number of labeled samples is 150-250, and this embodiment uses 200. The spectral model established by this number of samples has high stability and strong prediction ability, meeting the requirements of miniaturized near-infrared spectral analysis technology; too many samples will cause waste of manpower and material resources, and the ability to improve the effect of the spectral model is very limited. The spectral model established by too few samples has poor stability, is very easy to overfit, and has weak prediction ability; the number of unlabeled samples can be (1~+∞), and the spectral data can be collected and processed by using the processing method of a single unlabeled sample to complete the component prediction of all unlabeled samples. The specific collection method is: use N devices to collect M labeled samples respectively, each device collects M spectral data, and a total of (M×N) spectral data is collected; use 1 adaptive spectral model device to collect spectral data of 1 unlabeled sample to obtain a single spectral data.

[0048] Furthermore, (M×N) spectral data are used to establish the original spectral model library. For a single device among the N devices, M spectral data are collected and corresponded one by one with the known M component calibration values. The partial least squares method is used for spectral modeling to obtain a single spectral model. Similarly, the same modeling process is performed on the remaining (N-1) original spectral database establishment devices to obtain (N-1) spectral models. Therefore, N original spectral database establishment devices have established N spectral models, that is, the original spectral model library contains N spectral models.

[0049] Step 102 is to use the same batch of miniaturized near-infrared spectrometers to collect spectral data of three standard calibration plates. The three standard calibration plates are black calibration plates, white calibration plates, and gray calibration plates. These standard calibration plates are minimally affected by the environment, the material is not easy to change, the oil resistance is strong, the test surface is evenly distributed, and it is extremely stable within the near-infrared spectrum wavelength range. It is a calibration material that is extremely suitable for miniaturized near-infrared spectroscopy equipment.

[0050] In this embodiment, the wavelength point arrangement method of the miniaturized near-infrared spectroscopy device adopted in this patent is an evenly distributed method, the spectral band range is 1100nm~1500nm, and the resolution is 10nm. It can be calculated that the number of wavelength points contained in the miniaturized near-infrared device is M=1+(1500-1100) / 10=41. It can be further known that the spectral data collected by the miniaturized near-infrared spectroscopy device is actually represented as a matrix set of light intensity values ​​at 41 wavelength points.

[0051] Furthermore, the reflectivities of the black calibration plate, the white calibration plate, and the gray calibration plate are α1=3%, α2=98%, and α3=47%, respectively. It can be approximated that the black calibration plate absorbs all the near-infrared light emitted by the device, the white calibration plate reflects all the near-infrared light emitted by the device, and the gray calibration plate absorbs exactly half of the near-infrared light emitted by the device. Selecting three representative calibration plates for the adaptation of spectral equipment can minimize the adaptation error and improve the adaptation accuracy of spectral equipment.

[0052] Step 103 is to map the spectral data of the three standard calibration plates to a single standard calibration plate in combination with the reflectivity of the standard calibration plate. If a single standard calibration plate is directly used for spectral model matching, the matching error is large due to the singleness of the calibration plate and the poor stability of the miniaturized near-infrared device. Mapping the spectral data of the three standard calibration plates through the correspondence between reflectivity and spectral data can effectively retain the spectral data characteristics of the three calibration plates, reduce errors, and improve the matching accuracy of spectral devices.

[0053] In this embodiment, it is assumed that a single device among the N original spectral database establishment devices collects and obtains the spectral data of the black calibration plate, the white calibration plate, and the gray calibration plate, respectively B1, W1, and G1, and a single spectral model device to be matched collects and obtains the spectral data of the black calibration plate, the white calibration plate, and the gray calibration plate, respectively B2, W2, and G2. Since the working principle of the miniaturized near-infrared spectrometer is that the illumination system emits near-infrared light to the standard calibration plate, the standard calibration plate partially absorbs the near-infrared light, and the unabsorbed near-infrared light is reflected to the spectral device sensor, and the sensor converts the optical signal into an electrical signal, and then further converts it into a numerical signal, namely, spectral data. It can be seen that the spectral data of the standard calibration plate is linearly positively correlated with its reflectivity, namely:

[0054]

[0055] The reflectivity of the gray calibration plate is in the middle. Compared with the total absorption of the black calibration plate and the total reflection of the white calibration plate, which are too extreme, the gray calibration plate is more suitable as the standard reference plate. Taking the gray calibration plate as the standard reference plate, the black calibration plate and the white calibration plate are mapped to the gray calibration plate. Then, the spectral data of the black calibration plate mapped to the gray calibration plate, the white calibration plate mapped to the gray calibration plate, and the gray calibration plate mapped to the gray calibration plate of a single device in the N original spectral database establishment devices are respectively G1×1, we can further calculate the mean value of the spectrum data of three standard calibration plates mapped to the spectrum data of a single standard calibration plate:

[0056]

[0057] Similarly, the average value of the spectral data of three standard calibration plates of a single spectral model device to be matched mapped to the spectral data of a single standard calibration plate can be calculated as:

[0058]

[0059] Step 104 is to calculate the spectral data offset coefficient of a single standard calibration plate, calculate the trend offset in combination with the offset coefficient, and perform spectral device model matching in combination with the trend offset. For the same single standard calibration plate, the smaller the trend offset of the spectral data collected by the two types of miniaturized near-infrared spectroscopy devices, the more consistent the trends between the spectral data of the two types, and thus it can be known that their spectral models are also consistent, so the two devices with the smallest trend offset in the two types of devices are selected for spectral model matching.

[0060] In this embodiment, there is a certain offset in the light intensity values ​​of the same standard calibration plate spectrum data collected by the original spectrum database establishment device and the device to be matched with the spectrum model. Therefore, it is necessary to first calculate the offset coefficient of the single standard calibration plate of the two types of equipment, and set the spectrum data of the three standard calibration plates of the single original spectrum database establishment device to be mapped to the spectrum data mean value S1=(S 01 , S 02 , ... S 41 ), the spectral data of three standard calibration plates of a single adaptive spectral model device are mapped to the mean spectral data of a single standard calibration plate S2 = (S' 01 , S' 02 , ... S' 41 ), then the offset coefficient β between the two types of devices can be calculated as:

[0061]

[0062] The spectral data collected by the two types of miniaturized near-infrared devices are kept at the same benchmark through the offset coefficient. The offset coefficient is substituted into the spectral data of the three standard calibration plates of a single adaptive spectral model device and mapped to the spectral data mean S2 of a single standard calibration plate. The spectral data mean S3 of a single original spectral database establishment device under the same benchmark conditions is calculated as follows:

[0063] S3=S2×β=(S' 01 ×β,S' 02 ×β, ...S' 41 ×β)

[0064] The trend offset of the spectral mean values ​​S1 and S3 is calculated, and the trend offset C1 between the spectral data of the standard calibration plate between the two miniaturized near-infrared spectroscopy devices is calculated as:

[0065]

[0066] Similarly, the trend offsets of the spectral data of the standard calibration plate collected by a single adaptive spectral model device compared to the spectral data of the standard calibration plate collected by N original spectral database establishment devices are calculated respectively (C1, C2, ... C N ).

[0067] For the same standard calibration plate, the smaller the trend offset of the spectral data collected by the spectral equipment, the more consistent the trends between the two spectral data are, and thus it can be known that their spectral models are also consistent. From the above, from the trend offset (C1, C2, ... C N ) selects the two devices with the smallest offset value as the adaptive devices.

[0068] Step 105 is to perform isoquantization calculation on the unlabeled sample spectral data in combination with the offset coefficient, and call the spectral model of the adaptive spectral device to perform prediction analysis on the isoquantized spectral data. Since there may be a certain magnitude difference between the spectral data of the same type of samples collected by the two types of miniaturized near-infrared spectral devices, when calling the spectral model of the adaptive spectral device to predict the unlabeled sample spectral data, it is necessary to perform isoquantization calculation on the unlabeled sample spectral data to unify the magnitude of the unlabeled sample spectral data to the labeled sample spectral data.

[0069] In this embodiment, the offset coefficient calculated by the standard calibration plate is also applicable to the magnitude conversion between the unlabeled sample spectral data and the labeled sample spectral data. Assuming that a single adaptive spectral model device collects spectral data of one unlabeled sample, P1=(P 01 , P 02 , ... P 41), firstly perform isometry calculation on it, unify the magnitude of the unlabeled sample spectral data to the labeled sample spectral data, and calculate and obtain the isometry unlabeled sample spectral data P2 as:

[0070] P2=(P 01 ×β,P 02 ×β, ...P 41 ×β)

[0071] The spectral model created by the original spectral database building device that is adaptive to the single adaptive spectral model device is further called to perform predictive analysis on the quantified unlabeled sample spectral data P2 to obtain a predictive analysis value, which is the predicted value of the component content of the unlabeled sample.

Claims

1. A spectral model adaptation method based on a standard calibration plate, characterized in that: include: Step 1: Use N original near-infrared spectrometers of the same model to collect M labeled sample spectral data, and construct spectral models according to the labeled sample spectral data to form an original spectral model library; Step 2: Use N original near-infrared spectrometers and the near-infrared spectrometer to be matched to respectively collect spectral data of the standard calibration plate, and match the spectral model for the near-infrared spectrometer to be matched in the original spectral model library based on the collected spectral data; The step 2 is specifically as follows: step 21, using N original near-infrared spectrometers and the near-infrared spectrometer to be matched to respectively collect spectral data of Y kinds of standard calibration plates; Step 22: Map the spectral data of Y kinds of standard calibration plates corresponding to each near-infrared spectrometer to a single standard calibration plate in combination with the reflectivity of the standard calibration plate, and calculate the mean value of each near-infrared spectrometer according to the mapped value; Step 23, calculating the trend offset according to the mean value corresponding to each original near-infrared spectrometer and the mean value corresponding to the near-infrared spectrometer to be matched, and matching the spectral model for the near-infrared spectrometer to be matched in the original spectral model library according to the trend offset; The step 23 is specifically as follows: Step 231, respectively calculating the offset coefficients between the two types of devices according to the mean values ​​corresponding to the original near-infrared spectrometers and the mean values ​​corresponding to the near-infrared spectrometers to be matched; Step 232, correcting the mean value according to the offset coefficient; Step 233: Use the corrected mean to calculate the trend offset, and match the spectral model for the near-infrared spectrometer to be matched in the original spectral model library according to the trend offset.

2. The spectral model adaptation method based on the standard calibration plate according to claim 1, characterized in that: In step 2, N original near-infrared spectrometers and the near-infrared spectrometer to be matched are used to collect spectral data of Y kinds of standard calibration plates respectively, and Y is greater than 1.

3. The spectral model adaptation method based on the standard calibration plate according to claim 1, characterized in that: The calculation method of the offset coefficient β is: w is the number of spectral wavelength points, Sn is the mean value corresponding to the original near-infrared spectrometer, S' n is the mean value corresponding to the near-infrared spectrometer to be matched.

4. The spectral model adaptation method based on the standard calibration plate according to claim 3 is characterized in that: The trend shift is calculated as:

5. The spectral model adaptation method based on the standard calibration plate according to claim 1, characterized in that: The method also includes step 3: correcting the unlabeled sample spectral data in combination with the offset coefficient, and using the matched spectral model to perform predictive analysis on the corrected spectral data.

6. The spectral model adaptation method based on the standard calibration plate according to claim 1, characterized in that: The single standard calibration plate in step 22 is a standard calibration plate with a medium reflectivity among the Y types of standard calibration plates.

7. The spectral model adaptation method based on a standard calibration plate according to any one of claims 1 to 6, characterized in that: M is between 150 and 250, and N is between 15 and 25.

8. The spectral model adaptation method based on a standard calibration plate according to any one of claims 1 to 6, characterized in that: Y is 3, and the three standard calibration plates are: black calibration plate, white calibration plate, and gray calibration plate.

Images