Method and device for improving repeatability of spectrum prediction result and storage medium

Abstract

The invention relates to a method and device for improving repeatability of a spectrum prediction result and a storage medium, and belongs to the technical field of spectrum measurement, the method comprises the following steps: obtaining sampling data of a to-be-measured sample, each sampling data comprising N sub-sample data; for each piece of sampling data, based on the median of the N sub-sample data in the sampling data and a preset range threshold value, carrying out data rejection on the N sub-sample data to obtain processed sub-sample data; based on the median, M normal distributions are constructed for the processed subsample data; determining a sampling result of the sampling data based on an average value of the median of each normal distribution, wherein the sampling result is used for performing spectrum prediction; as the subsample data in any distribution form is converted to be formed by connecting the M normal distributions, the repeatability of the spectrum prediction result can be improved, meanwhile, spectrum acquisition equipment with higher repeatability precision does not need to be used, and the research and development cost can be reduced.

Description

Technical Field

[0001] The present application relates to a method, a device and a storage medium for improving the repeatability of spectral prediction results, and belongs to the technical field of spectral measurement. Background Art

[0002] The repeatability of spectral prediction results refers to the consistency between the spectral prediction results obtained when the spectral measurements of the same sample are performed multiple times under the same experimental conditions and the spectral model is used for prediction. In other words, the repeatability of spectral prediction results is used to measure the stability and reliability of spectral prediction results. If the spectral prediction results of the spectral model have good repeatability, the performance of spectral analysis will be better. Therefore, it is necessary to improve the repeatability of spectral prediction results.

[0003] Traditional methods to improve the repeatability of spectral prediction results include the following:

[0004] 1. Increase the number of spectrum acquisition averages to improve the signal-to-noise ratio; however, this method will cause the detection time to be too long, and the temperature effect will be very obvious when the sample is exposed to light for a long time, affecting the measurement accuracy of the sample;

[0005] 2. Increase the number of sample collection sub-samples and average multiple results; however, this will also cause the test time to be too long, the impact of temperature effect will also be very large, and the difference in test results between multiple times will be greater;

[0006] 3. Improve the spectral repeatability of spectral acquisition equipment (i.e. improve equipment performance); however, this method has high R&D costs, low efficiency, and a long time period, and cannot produce significant results in the short term;

[0007] 4. The spectral prediction results are eliminated by the median upper and lower limits; however, if the median deviates, the repeatability of the spectral prediction results will deteriorate even if the elimination mechanism is added;

[0008] 5. Use Mahalanobis distance to eliminate spectra; although this method can effectively eliminate spectra that deviate from the spectral prediction results, the prerequisite is that the quality of the spectrum is good (i.e., high signal-to-noise ratio and good repeatability), and it is also necessary to consider improving equipment performance to solve the problem of poor repeatability.

[0009] 6. Segmented modeling strategy, and in quantitative analysis, different contents of the same substance are divided into segments for modeling. Although it can improve the repeatability within the content range of the analyte, segmented modeling is complicated to use, and when the sample is located at the inflection point of the modeling with different contents, if the wavelength drift occurs in the equipment, the accuracy of the sample quantitative analysis will deteriorate. Summary of the invention

[0010] This application provides a method, device and storage medium for improving the repeatability of spectral prediction results, which can solve the problem of poor repeatability of multiple measurement results and low user credibility; at the same time, there is no need to use spectral acquisition equipment with higher repeatability accuracy, which can reduce research and development costs. This application provides the following technical solutions:

[0011] In a first aspect, a method for improving the repeatability of spectral prediction results is provided, the method comprising:

[0012] Acquire sampling data of the sample to be tested, each sampling data includes N sub-sample data; N is an integer greater than 1;

[0013] For each sampled data, based on the median of N sub-sampled data in the sampled data and a preset range threshold, data removal is performed on the N sub-sampled data to obtain processed sub-sampled data;

[0014] Based on the median, constructing M normal distributions for the processed sub-sample data; wherein M is an odd number greater than or equal to 3;

[0015] The sampling result of the sampling data is determined based on the average value of the median of each normal distribution, and the sampling result is used for spectrum prediction.

[0016] Optionally, constructing M normal distributions for the processed sub-sample data based on the median includes:

[0017] Based on the median and a preset first distribution range threshold, determining a first data range consisting of data located at both ends of the median; wherein the first distribution range threshold is less than the range threshold;

[0018] Within the first data range, taking the two ends of the median with the smaller amount of sample data as a reference, constructing an intermediate normal distribution;

[0019] For each sub-data range located within the processed sub-sample data and outside the first data range, determining a sub-median of the sub-sample data within the sub-data range;

[0020] Determine a second data range based on the sub-median and a second distribution range threshold corresponding to the sub-data range; the second distribution threshold is less than the range threshold, and the sum of the second distribution range threshold corresponding to each sub-data range and the first distribution range threshold is the range threshold;

[0021] Within the second data range, a two-end normal distribution is constructed with the two ends of the sub-median having the smaller amount of sub-sample data as a reference.

[0022] Optionally, the method further comprises:

[0023] The first distribution range threshold and the second distribution range threshold are set based on the distribution of the processed sub-sample data.

[0024] Optionally, the setting of the first distribution range threshold and the second distribution range threshold based on the distribution of the processed sub-sample data includes:

[0025] If the distribution of the processed sub-sample data is a normal distribution, setting the difference between the first distribution range threshold and the range threshold to be smaller than a preset threshold, and setting the second distribution range threshold to be smaller than the difference;

[0026] If the distribution of the processed sub-sample data is an M-type distribution, the second distribution range threshold is set to be greater than the first distribution range threshold.

[0027] Optionally, for each sampled data, based on the median of N sub-sampled data in the sampled data and a preset range threshold, data removal is performed on the N sub-sampled data to obtain processed sub-sampled data, including:

[0028] Determine the median of the N sub-sample data;

[0029] The sum of the median and the range threshold is used as the upper limit value, and the difference between the median and the range threshold is used as the lower limit value to obtain a data screening range;

[0030] The data outside the data screening range is eliminated to obtain the processed sub-sample data.

[0031] Optionally, the range threshold is set based on the repeatability of sample sub-sample results in the modeling set of the spectral model.

[0032] Optionally, the value of M is 3.

[0033] In a second aspect, a device for improving the repeatability of spectral prediction results is provided, the device comprising:

[0034] A data acquisition module, used to obtain sampling data of the sample to be tested, each sampling data includes N sub-sample data; N is an integer greater than 1;

[0035] A data screening module is used to, for each sampled data, remove the N sub-sampled data based on the median of the N sub-sampled data in the sampled data and a preset range threshold to obtain processed sub-sampled data;

[0036] A data processing module, configured to construct M normal distributions for the processed sub-sample data based on the median; wherein M is an odd number greater than or equal to 3;

[0037] The result determination module is used to determine the sampling result of the sampling data based on the average value of the median of each normal distribution, and the sampling result is used for spectrum prediction.

[0038] In a third aspect, a device for improving the repeatability of spectral prediction results is provided, the device comprising a processor and a memory; a program is stored in the memory, and the program is loaded and executed by the processor to implement the method for improving the repeatability of spectral prediction results described in the first aspect.

[0039] In a fourth aspect, a computer-readable storage medium is provided, wherein a program is stored in the storage medium, and the program is loaded and executed by the processor to implement the method for improving the repeatability of spectral prediction results described in the first aspect.

[0040] The beneficial effects of the present application include: by acquiring sampling data of a sample to be tested, each sampling data includes N sub-sample data; for each sampling data, based on the median of the N sub-sample data in the sampling data and a pre-set extreme value, the N sub-sample data are eliminated to obtain processed sub-sample data; based on the median, M normal distributions are constructed for the processed sub-sample data; based on the average value of the median of each normal distribution, the sampling result of the sampling data is determined, and the sampling result is used for spectral prediction; the problem of poor repeatability of multiple measurement results and low user credibility can be solved; since the sub-sample data of any distribution form is converted into a connection composed of M normal distributions, the repeatability of the spectral prediction result can be improved, and at the same time, there is no need to use spectral acquisition equipment with higher repeatability accuracy, which can reduce R&D costs.

[0041] In addition, in this application, different samples to be tested can adjust the range of repeatability results by setting different range thresholds. This is suitable for solid samples analyzed by the near-infrared industry on the market, and there is no need to consider the degree of mixing between samples. The repeatability of the results can still be controlled under the condition of multiple mixed samples of one substance. The system integration cost is low and the efficiency is higher.

[0042] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application and implement it according to the contents of the specification, the following is a detailed description of the preferred embodiments of the present application in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of a method for improving the repeatability of spectral prediction results provided by an embodiment of the present application;

[0044] Figure 2 is a flow chart of a method for improving the repeatability of spectral prediction results provided by another embodiment of the present application;

[0045] Figure 3 is a block diagram of a device for improving the repeatability of spectral prediction results provided by an embodiment of the present application;

[0046] Figure 4 It is a block diagram of a device for improving the repeatability of spectral prediction results provided by an embodiment of the present application. DETAILED DESCRIPTION

[0047] The specific implementation methods of the present application are further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present application but are not intended to limit the scope of the present application.

[0048] Usually, when quantitative analysis is used to measure solid particles, the data collected once is obtained through multiple sampling times. At this time, more acquisition times will be used to average the final measurement result. However, when the collected spectrum is abnormal, the model prediction value will also be abnormal, causing data deviation. At this time, when the results of multiple acquisitions have poor repeatability, the credibility of the data will be low, which will affect user use.

[0049] Based on this, the present application provides a data processing method to reduce costs and improve the repeatability of multiple acquisition results, thereby enhancing users' credibility of measurement data.

[0050] Figure 1 This is a flow chart of a method for improving the repeatability of spectral prediction results provided by an embodiment of the present application. The method includes at least the following steps:

[0051] Step 101: Acquire sampling data of a sample to be tested, each sampling data includes N sub-sample data.

[0052] Wherein, N is an integer greater than 1.

[0053] Usually, the collected data of the sample to be tested is obtained by collecting multiple sub-samples, that is, each sample data includes N sub-sample data.

[0054] For example, if the value of N is 100, the sampling data obtained each time includes 100 spectra corresponding to 100 sub-sample results.

[0055] Step 102 : for each sampled data, based on the median of the N sub-sampled data in the sampled data and a preset range threshold, data are eliminated from the N sub-sampled data to obtain processed sub-sampled data.

[0056] In one example, for each sampled data, N sub-sample data are eliminated based on their medians and a preset range threshold in the sampled data to obtain processed sub-sample data, including: determining the median of the N sub-sample data; using the sum of the median and the range threshold as the upper limit value and the difference between the median and the range threshold as the lower limit value to obtain a data screening range; and eliminating data outside the data screening range to obtain processed sub-sample data.

[0057] In other words, the data screening range is the median ± m, m is the range threshold, the sub-sample data within the data screening range is retained, and the sub-sample data outside the data screening range is removed to obtain the processed sub-sample data. It can be seen that the sub-sample range of the processed sub-sample data is 2|m|, that is, twice the absolute value of m.

[0058] In one example, the range threshold is set based on the repeatability of the sample sub-sample results in the modeling set of the spectral model. That is, the measurement results of all sub-samples are collected from the modeling set of the spectral model. These measurement results are obtained based on multiple measurements under the same or similar conditions; for each sub-sample, the range of its measurement results, that is, the difference between the maximum and minimum values, is calculated. By analyzing these range values, the repeatability of the sub-sample results is evaluated. If the range of most sub-samples is very small, it means that the results have high repeatability; conversely, if the range is large, it may indicate that the repeatability of the results is poor. Based on the analysis results of the repeatability, a range threshold can be set. This range threshold can include most of the sub-sample results with high repeatability and exclude those results with extremely large ranges. The set range threshold is used to eliminate those sub-sample results whose range exceeds the threshold. Doing so helps to improve the accuracy and reliability of the model.

[0059] Step 103: construct M normal distributions for the processed sub-sample data based on the median.

[0060] Here, M is an odd number greater than or equal to 3.

[0061] Due to the multiple acquisitions, there are various distributions of sub-sample data within the range, such as normal distribution, random distribution, M-type distribution, and trend distribution. Different distributions of sub-sample data will cause most of the sub-sample data to deviate from the median, resulting in the overall data deviating from the median after averaging, thus resulting in the problem of poor repeatability of the final averaged result. In this embodiment, the N sub-sample data are not directly averaged, but M normal distributions are constructed for the processed sub-sample data, that is, the sub-sample data of any distribution form is converted into a combination of M normal distributions, which can solve the problem of different distributions of sub-sample data acquired multiple times, thereby improving the repeatability of the final averaged result.

[0062] In one example, based on the median, M normal distributions are constructed for the processed sub-sample data, including steps 1031-1035:

[0063] Step 1031, based on the median and a preset first distribution range threshold, determine a first data range consisting of data located at both ends of the median, wherein the first distribution range threshold is smaller than the range threshold.

[0064] Assuming that a first distribution range threshold n is given, the first data range can be determined based on the median, that is, the data range consisting of the median ±n.

[0065] Step 1032, within the first data range, construct an intermediate normal distribution based on the ends of the median with the smaller amount of sample data.

[0066] In this step, a standard normal distribution is constructed based on the side with fewer sub-sample results within the range of both ends of the median, and the redundant data within the first data range is eliminated, thus completing the construction of the first normal distribution. At this point, the influence of redundant deviating data on the average result can be eliminated.

[0067] Specifically, the middle normal distribution is constructed based on the end with less sample data at both ends of the median, including: taking the side with lower probability density as the standard and eliminating the data with higher probability density on the other side, so that the constructed normal distribution is a standard normal distribution that is symmetrical on both sides.

[0068] Step 1033: for each sub-data range that is within the processed sub-sample data and outside the first data range, determine the sub-median of the sub-sample data within the sub-data range.

[0069] Taking the value of M as 3 as an example, each sub-data range located within the processed sub-sample data and outside the first data range includes: sub-data ranges located on both sides outside the first data range.

[0070] When the value of M is an odd number greater than 3, each sub-data range within the processed sub-sample data and outside the first data range can be obtained by evenly dividing the data ranges on both sides of the first data range according to (M-1) / 2, or by dividing according to a preset data interval. This embodiment does not limit the way of dividing the sub-data ranges.

[0071] Step 1034, determine the second data range based on the second distribution range threshold corresponding to the sub-median and the sub-data range; wherein the second distribution threshold is less than the range threshold, and the sum of the second distribution range threshold corresponding to each sub-data range and the first distribution range threshold is the range threshold.

[0072] Take the value of M as 3 as an example. At this time, in addition to the first data range, two sub-data ranges are also included. Assuming that the second distribution range thresholds k1 and k2 corresponding to the two sub-data ranges are given, one second data range includes a data range obtained by the sub-median ±k1 of one sub-data range; the other second data range includes a data range obtained by the sub-median ±k2 of another sub-data range. At this time, |m|>|n|+|k1|+|k2|.

[0073] When the value of M is an odd number greater than 3, different sub-data ranges correspond to a second distribution range threshold, and the second distribution range thresholds corresponding to different sub-data ranges are the same or different. This embodiment does not limit the value of the second distribution range threshold.

[0074] Step 1035 , within the second data range, taking the two ends of the sub-sample data with the smaller amount of data as a reference, construct a two-end normal distribution.

[0075] For each second data range, the side with lower probability density at both ends of the sub-median is taken as the standard, and the data with higher probability density on the other side is eliminated, so that the constructed normal distribution is a standard normal distribution that is symmetrical on both sides, and the normal distribution at both ends corresponding to the second data range is obtained.

[0076] In this embodiment, before step 1031, the method further includes: setting a first distribution range threshold and a second distribution range threshold based on the distribution of the processed sub-sample data.

[0077] In one example, if the distribution of the processed sub-sample data is a normal distribution, the difference between the first distribution range threshold and the range threshold is set to be smaller than a preset threshold, and the second distribution range threshold is set to be smaller than the difference.

[0078] In another example, if the distribution of the processed sub-sample data is an M-type distribution, the second distribution range threshold is set to be greater than the first distribution range threshold. The M in the M-type distribution does not represent a numerical value. The M-type distribution means that the data distribution graph presents a shape similar to the letter "M". This distribution feature indicates that there are two obvious peaks in the data and a valley in the middle, forming a special form of bimodal distribution.

[0079] Step 104: determining a sampling result of the sampled data based on the average value of the median of each normal distribution, and the sampling result is used for spectrum prediction.

[0080] In this way, by selecting different distribution range thresholds, the range threshold of the results of multiple acquisition repeatability can be set. That is, within the range of |m|, the values ​​of |k1| and |k2| are greater than |n|, and the larger the value, the greater the range of the repeatability of multiple acquisitions. If |k1| and |k2| are extremely small or infinitesimal, the range of the repeatability results of multiple acquisitions can be 0. The range of the repeatability results can be controlled.

[0081] In summary, the method for improving the repeatability of spectral prediction results provided in this embodiment is to obtain sampling data of the sample to be tested, each sampling data includes N sub-sample data; for each sampling data, the N sub-sample data are eliminated based on the median of the N sub-sample data in the sampling data and a pre-set extreme value to obtain processed sub-sample data; based on the median, M normal distributions are constructed for the processed sub-sample data; the sampling results of the sampling data are determined based on the average value of the median of each normal distribution, and the sampling results are used for spectral prediction; the problems of poor repeatability of multiple measurement results and low user credibility can be solved; since the sub-sample data of any distribution form are converted into a combination of M normal distributions, the repeatability of the spectral prediction results can be improved. At the same time, there is no need to use spectral acquisition equipment with higher repeatability accuracy, which can reduce research and development costs.

[0082] In addition, in this application, different samples to be tested can adjust the range of repeatability results by setting different range thresholds. This is suitable for solid samples analyzed by the near-infrared industry on the market, and there is no need to consider the degree of mixing between samples. The repeatability of the results can still be controlled under the condition of multiple mixed samples of one substance. The system integration cost is low and the efficiency is higher.

[0083] In order to more clearly understand the method for improving the repeatability of spectral prediction results provided by the present application, the following takes M=3 as an example to illustrate the method. Figure 2 , the method includes the following steps:

[0084] Step 21, obtaining sampling data of the sample to be tested, each sampling data includes N sub-sample data;

[0085] Step 22, calculate the median value of N sub-samples, set the range threshold to m, remove the sub-sample data outside the range of ±m around the median value, and obtain the processed sub-sample data;

[0086] Step 23, constructing an intermediate normal distribution according to the range of ±n with the median value of the processed sub-sample data, and eliminating the sub-sample data with a large deviation within the current range of ±n;

[0087] Step 24, using the sub-sample data outside the ±n range and within the ±m range, respectively setting k1 and k2 on the left and right to construct two normal distributions, to obtain a two-end normal distribution, so as to eliminate the sub-sample data with large deviations on the left and right sides between the ±n range and the ±m range;

[0088] Step 25, averaging the medians of the three normal distributions to obtain a sampling result.

[0089] Figure 3 The block diagram of a device for improving the repeatability of spectrum prediction results provided by an embodiment of the present application includes at least the following modules: a data acquisition module 310 , a data screening module 320 , a data processing module 330 and a result determination module 340 .

[0090] The data acquisition module 310 is used to obtain sampling data of the sample to be tested, each sampling data includes N sub-sample data; N is an integer greater than 1;

[0091] The data screening module 320 is used to, for each sampled data, remove the N sub-sampled data based on the median of the N sub-sampled data in the sampled data and a preset range threshold to obtain processed sub-sampled data;

[0092] A data processing module 330 is used to construct M normal distributions for the processed sub-sample data based on the median; M is an odd number greater than or equal to 3;

[0093] The result determination module 340 is used to determine the sampling result of the sampling data based on the average value of the median of each normal distribution, and the sampling result is used for spectrum prediction.

[0094] For relevant details, refer to the above method embodiment.

[0095] It should be noted that: the device for improving the repeatability of spectral prediction results provided in the above embodiments only uses the division of the above functional modules as an example to illustrate when improving the repeatability of spectral prediction results. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device for improving the repeatability of spectral prediction results is divided into different functional modules to complete all or part of the functions described above. In addition, the device for improving the repeatability of spectral prediction results provided in the above embodiments and the method embodiment for improving the repeatability of spectral prediction results belong to the same concept. The specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0096] Figure 4 4 is a block diagram of a device for improving the repeatability of spectrum prediction results provided by an embodiment of the present application. The device at least includes a processor 401 and a memory 402.

[0097] The processor 401 may include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 401 may be implemented in at least one hardware form of DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 401 may also include a main processor and a coprocessor. The main processor is a processor for processing data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in the standby state. In some embodiments, the processor 401 may be integrated with a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the display screen. In some embodiments, the processor 401 may also include an AI (Artificial Intelligence) processor, which is used to process computing operations related to machine learning.

[0098] The memory 402 may include one or more computer-readable storage media, which may be non-transitory. The memory 402 may also include a high-speed random access memory, and a non-volatile memory, such as one or more disk storage devices, flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 402 is used to store at least one instruction, which is used to be executed by the processor 401 to implement the method for improving the repeatability of spectral prediction results provided by the method embodiment of the present application.

[0099] In some embodiments, the device for improving the repeatability of the spectrum prediction result may also optionally include: a peripheral device interface and at least one peripheral device. The processor 401, the memory 402 and the peripheral device interface may be connected via a bus or a signal line. Each peripheral device may be connected to the peripheral device interface via a bus, a signal line or a circuit board. Schematically, the peripheral devices include but are not limited to: a radio frequency circuit, a touch display screen, an audio circuit, and a power supply.

[0100] Of course, the device for improving the repeatability of spectral prediction results may also include fewer or more components, which is not limited in this embodiment.

[0101] Optionally, the present application also provides a computer-readable storage medium, in which a program is stored, and the program is loaded and executed by a processor to implement the method for improving the repeatability of spectral prediction results of the above method embodiment.

[0102] Optionally, the present application also provides a computer product, which includes a computer-readable storage medium, in which a program is stored, and the program is loaded and executed by a processor to implement the method for improving the repeatability of spectral prediction results of the above method embodiment.

[0103] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0104] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for improving the repeatability of spectral prediction results, characterized in that: The method comprises: Acquire sampling data of the sample to be tested, each sampling data includes N sub-sample data; N is an integer greater than 1; For each sampled data, based on the median of N sub-sampled data in the sampled data and a preset range threshold, data removal is performed on the N sub-sampled data to obtain processed sub-sampled data; Based on the median, constructing M normal distributions for the processed sub-sample data; wherein M is an odd number greater than or equal to 3; The sampling result of the sampling data is determined based on the average value of the median of each normal distribution, and the sampling result is used for spectrum prediction.

2. The method according to claim 1, characterized in that The constructing M normal distributions for the processed sub-sample data based on the median includes: Based on the median and a preset first distribution range threshold, determining a first data range consisting of data located at both ends of the median; wherein the first distribution range threshold is less than the range threshold; Within the first data range, taking the two ends of the median with the smaller amount of sample data as a reference, constructing an intermediate normal distribution; For each sub-data range located within the processed sub-sample data and outside the first data range, determining a sub-median of the sub-sample data within the sub-data range; Determine a second data range based on the sub-median and a second distribution range threshold corresponding to the sub-data range; the second distribution threshold is less than the range threshold, and the sum of the second distribution range threshold corresponding to each sub-data range and the first distribution range threshold is the range threshold; Within the second data range, a two-end normal distribution is constructed with the two ends of the sub-median having the smaller amount of sub-sample data as a reference.

3. The method according to claim 2, characterized in that The method further comprises: The first distribution range threshold and the second distribution range threshold are set based on the distribution of the processed sub-sample data.

4. The method according to claim 3, characterized in that The setting of the first distribution range threshold and the second distribution range threshold based on the distribution of the processed sub-sample data includes: If the distribution of the processed sub-sample data is a normal distribution, setting the difference between the first distribution range threshold and the range threshold to be smaller than a preset threshold, and setting the second distribution range threshold to be smaller than the difference; If the distribution of the processed sub-sample data is an M-type distribution, the second distribution range threshold is set to be greater than the first distribution range threshold.

5. The method according to any one of claims 1 to 4, characterized in that: For each sampled data, based on the median of N sub-sampled data in the sampled data and a preset range threshold, data removal is performed on the N sub-sampled data to obtain processed sub-sampled data, including: Determine the median of the N sub-sample data; The sum of the median and the range threshold is used as the upper limit value, and the difference between the median and the range threshold is used as the lower limit value to obtain a data screening range; The data outside the data screening range is eliminated to obtain the processed sub-sample data.

6. The method according to claim 5, characterized in that The range threshold is set based on the repeatability of the sample sub-sample results in the modeling set of the spectral model.

7. The method according to any one of claims 1 to 4, characterized in that: The value of M is 3.

8. A device for improving the repeatability of spectral prediction results, characterized in that: The device comprises: A data acquisition module, used to obtain sampling data of the sample to be tested, each sampling data includes N sub-sample data; N is an integer greater than 1; A data screening module is used to, for each sampled data, remove the N sub-sampled data based on the median of the N sub-sampled data in the sampled data and a preset range threshold to obtain processed sub-sampled data; A data processing module, configured to construct M normal distributions for the processed sub-sample data based on the median; wherein M is an odd number greater than or equal to 3; The result determination module is used to determine the sampling result of the sampling data based on the average value of the median of each normal distribution, and the sampling result is used for spectrum prediction.

9. A device for improving the repeatability of spectral prediction results, characterized in that: The device includes a processor and a memory; a program is stored in the memory, and the program is loaded and executed by the processor to implement the method for improving the repeatability of spectral prediction results as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The storage medium stores a program, and the program is loaded and executed by the processor to implement the method for improving the repeatability of spectral prediction results as described in any one of claims 1 to 7.

Images