A method and system for automatic evaluation and grading of gamma spectroscopy data quality
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HEBEI XIONGAN BAIZE INFORMATION TECHNOLOGY CO LTD
- Filing Date
- 2026-03-18
- Publication Date
- 2026-07-21
AI Technical Summary
In existing technologies, the quality assessment of gamma-ray spectral data relies on human experience, which has problems such as strong subjectivity, low efficiency, difficulty in batch processing, and difficulty in unifying standards. In particular, it is difficult to meet the requirements of speed, accuracy, and standardization in large-scale measurement and high-throughput data acquisition scenarios.
An automatic assessment and classification method for gamma-ray spectral data quality was adopted, including data acquisition and preprocessing, index calculation and feature extraction, automatic quality scoring and data classification. The method combines machine learning models for objective and standardized assessment. Through noise filtering, baseline correction, outlier removal and normalization, the method calculates indicators such as peak signal-to-noise ratio, peak integrity, background noise level and peak position shift, generates a comprehensive quality score and classifies the data.
It enables automated and standardized evaluation of gamma spectrum data, improves the consistency and reliability of evaluation results, enhances data processing efficiency, supports batch processing and real-time management of large-scale data, and reduces the subjectivity of manual evaluation.
Smart Images
Figure CN122432476A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of radioactivity measurement and nuclear detection technology, specifically to a method and system for automatic quality assessment and classification of gamma-ray spectrum data. Background Technology
[0002] Gamma spectroscopy is an important tool for radionuclide analysis and environmental nuclear monitoring. It is widely used in geological exploration, environmental monitoring, nuclear facility safety and other fields. By analyzing the energy of gamma rays emitted by radionuclides in samples or the environment, the corresponding energy spectrum information can be obtained, thereby enabling qualitative and quantitative analysis of radioactive elements.
[0003] However, in actual measurement processes, due to the influence of various factors such as instrument status, measurement environment, operation method, and sample characteristics, the acquired gamma spectral data often suffers from problems such as noise interference, peak position drift, background anomalies, and low signal-to-noise ratio. These data quality issues not only affect subsequent spectral analysis and elemental quantitative calculations, but may also lead to inaccurate analytical results, increase the workload of data processing personnel, and affect the reliability and comparability of the data.
[0004] Currently, the assessment of gamma spectral data quality mainly relies on manual inspection and judgment, including manual judgment of peak shape, noise level, baseline drift, and instrument malfunction. This method has disadvantages such as strong subjectivity, low efficiency, difficulty in batch processing, and difficulty in unifying standards. Especially in large-scale measurement and high-throughput data acquisition scenarios, traditional manual assessment methods can hardly meet the requirements of fast, accurate, and standardized data quality.
[0005] Therefore, there is an urgent need for a method and system that can automate, objectively and standardize the quality assessment and classification of gamma-ray spectral data in order to improve data processing efficiency and ensure the reliability and availability of measurement data. Summary of the Invention
[0006] To achieve the above objectives, this invention provides an automatic assessment and grading method for gamma data quality, comprising the following steps:
[0007] S1, Data Acquisition and Preprocessing: First, the sample and environment are subjected to energy spectrum measurements using a gamma spectroscopy instrument to obtain raw gamma spectral data. After data acquisition, the raw data is preprocessed to ensure the accuracy and stability of subsequent analysis. The preprocessing process includes noise filtering of the acquired signal, correction of the energy spectrum baseline, automatic removal of outliers, and normalization of the energy channel data. These preprocessing steps can effectively reduce the impact of environmental interference, instrument drift, and sampling errors on data quality, providing a reliable data foundation for subsequent quality assessment.
[0008] S2, Indicator Calculation and Feature Extraction: After preprocessing, quality indicators and features are calculated for each gamma spectrum data. By quantitatively calculating key features such as peak signal-to-noise ratio, peak integrity, background noise level, and peak shift amplitude, a set of indicators for quality assessment is formed. At the same time, combined with historical high-quality data, data quality features are extracted through statistical analysis and machine learning methods to ensure that the indicators can comprehensively reflect the quality status of the energy spectrum data. This step provides a scientific and quantitative basis for data scoring and classification.
[0009] S3, Automatic Quality Scoring: Based on the results of index calculation and feature extraction, the gamma spectrum data is automatically scored. The scoring process uses a weighted scoring algorithm and machine learning model to comprehensively calculate various quality indicators and generate an overall quality score for each data point. The scoring model is trained and calibrated using historical data to ensure high accuracy and repeatability, thereby avoiding subjectivity and inconsistency in manual evaluation and achieving objectivity and standardization of data quality assessment.
[0010] S4, Data Classification and Labeling: Based on the calculated quality scores, the gamma spectrum data is divided into different levels: excellent, good, medium, and poor. Each data point is labeled accordingly. The classification criteria are determined based on the numerical range output by the scoring model, enabling data of different qualities to be quickly classified, which facilitates subsequent data screening, analysis, and storage management. Through data classification and labeling, we can not only intuitively understand the overall quality distribution of the data, but also achieve automated management in high-throughput measurement and large-scale data processing.
[0011] S5, Output and Visualization: Output the quality scoring and grading results to an information system, which is one or more of a database or a visualization terminal. It provides data query, statistical and visualization analysis functions, and displays data quality distribution, abnormal data prompts and trend changes through charts and interfaces. It provides decision-making basis and data management support for operators, and can also realize batch processing and real-time updates, providing reliable data quality assurance for high-throughput measurement and long-term monitoring.
[0012] Furthermore, in the S1 data acquisition and preprocessing, the gamma spectrum data is first processed by a measuring instrument to obtain the raw energy spectrum signal sequence. ,in This indicates the number of channels collected, each... To determine the corresponding channel count values, during preprocessing, a moving average filter is first used to filter noise in the original signal, using a window length of [value missing]. The formula for moving average filtering is:
[0013] ;
[0014] in, The filtered signal value is padded symmetrically at the boundaries; secondly, the sliding minimum value is used to adjust the energy spectrum baseline. Correction is performed to eliminate the influence of background noise; the corrected signal It can be represented as:
[0015] ;
[0016] in, The count value is the baseline corrected value;
[0017] To eliminate the influence of differences in measurement intensity, the channel count is normalized to its maximum value:
[0018] ;
[0019] Through this normalization step, the energy spectrum data is converted into an interval. The values within this range ensure that subsequent quality indicator calculations are not affected by the magnitude of the absolute count value; furthermore, for isolated peaks where the detected channel count values are significantly higher or lower than the surrounding average, a threshold is set. Remove:
[0020] ;
[0021] in, The mean of the neighborhood energy channels. The preprocessing steps described above result in noise filtering, baseline correction, normalization, and outlier removal of the original gamma spectrum data, providing a reliable input data foundation for subsequent quality index calculations and feature extraction.
[0022] Furthermore, in the S2 index calculation and feature extraction, the preprocessed gamma-ray spectrum data... Calculate key metrics for quality assessment to quantify the integrity, stability, and reliability of the data;
[0023] First, the signal-to-noise ratio (SNR) is calculated to evaluate the significance of the spectral peak signal relative to the background noise. The SNR is calculated as follows:
[0024] ;
[0025] in, This represents the energy channel set corresponding to the spectral peak. For the peak energy level, SNR is the standard deviation of the baseline region count values. The higher the SNR, the more obvious the spectral peak signal and the higher the data quality.
[0026] Secondly, peak integrity is calculated to evaluate the integrity and shape stability of spectral peaks. Peak integrity can be defined as the ratio of the total peak count to the ideal Gaussian fit count of the peak region:
[0027] ;
[0028] in, The values are the Gaussian fitting results for the spectral peak region. The closer the value is to 1, the more complete the spectral peak shape is and the less obvious the defects are.
[0029] Next, the background noise level is calculated to quantify the stability of the spectral baseline:
[0030] ;
[0031] in, The mean of the counts for the background region. The standard deviation of the count in the background region; the lower the value, the more stable the background noise.
[0032] In addition, peak shift is calculated to detect whether the spectral peak position deviates from the expected energy channel center:
[0033] ;
[0034] in, To observe the center value of the energy channel of the spectral peak, For reference, the center value of the energy channel of the spectral peak, A smaller value indicates better peak stability. A set of quantitative characteristics is formed through calculations using the above indicators. This feature set is used for subsequent quality scoring and grading. It can not only comprehensively reflect the quality status of energy spectrum data, but also be standardized by combining historical high-quality data to enhance the accuracy and repeatability of the scoring model.
[0035] Furthermore, in the S3 automatic quality scoring, the feature set obtained based on the index calculation and feature extraction steps is as follows:
[0036] ;
[0037] Each gamma-ray spectrum data point is comprehensively scored to quantify data quality. A weighted scoring method is used in the scoring process, assigning different weights to each indicator according to its importance to data quality. Calculate the overall quality score :
[0038] ;
[0039] in, For the first The normalized value of each quality indicator. For the corresponding weights, satisfying , The total number of indicators is used to adjust the weights so that the scoring system focuses on key indicators with high signal-to-noise ratio, good peak integrity, and stable background, thereby ensuring that the scoring results are highly correlated with the actual data quality.
[0040] To improve the objectivity and repeatability of the ratings, historical high-quality data was used as the training set, and a quality rating model was built using machine learning algorithms. The input to the rating model was a feature set. The output is the overall quality score. and the probability of level :
[0041] ;
[0042] in, This means that the trained scoring model automatically calculates the data quality score based on the input indicators and makes predictions for new data. This method avoids the subjectivity problem of traditional manual scoring and realizes automated, standardized and repeatable evaluation of gamma spectrum data quality. Through this step, each gamma spectrum data can obtain a comprehensive quality score, providing a quantitative basis for subsequent data classification and labeling, and supporting batch processing and real-time scoring, providing reliable data quality assurance for high-throughput measurement and long-term monitoring.
[0043] Furthermore, in the S4 data grading and labeling process, the comprehensive quality score obtained from the automatic quality scoring step... The gamma-ray spectral data was divided into different levels to visually demonstrate data quality and guide subsequent analysis. The data grading used a threshold system, mapping the overall quality score to four levels: Excellent, Good, Fair, and Poor. The specific divisions are as follows:
[0044] ;
[0045] in, These are preset mass fraction thresholds, which are obtained through statistical analysis of historical high-quality data to ensure that the classification results can reasonably reflect the actual quality of the energy spectrum data.
[0046] At the same time, a corresponding level label is generated for each data point. With confidence level The confidence level indicates the reliability of the rating model's determination of the level.
[0047] ;
[0048] in, For the first The overall quality score of the data points The higher the value, the more reliable the classification judgment. Through the classification label and confidence output, operators can quickly identify high-quality data and filter low-quality data for processing. The processing method is either retesting or elimination, thereby improving data management efficiency.
[0049] In addition, by combining batch processing mechanisms, a large amount of gamma-ray spectral data can be scored and classified simultaneously to achieve automated management of data quality. At the same time, statistical analysis is performed on data of each level to generate data quality distribution charts, providing a reference for long-term monitoring, instrument calibration and measurement scheme optimization.
[0050] Furthermore, in the S5 output and visualization process, the automatic quality scoring and data classification results are output to the database terminal system for operators to query, statistically analyze, and perform subsequent analysis. The output includes the comprehensive quality score for each gamma-ray spectrum data point. Corresponding level and confidence level At the same time, a set of quality indicators is generated. Detailed records are provided to facilitate retrospective analysis of the data. A visualization interface is also provided to graphically display the scoring and grading results. The visualization methods include bar charts, line charts, heat maps, and radar charts to intuitively reflect the distribution of data quality, the proportion of data at different levels, and the trend over time.
[0051] In addition, operators perform batch screening of data based on grade labels and quality score thresholds, eliminating low-quality data and individually labeled abnormal data, providing reliable high-quality data for subsequent spectral analysis, quantitative calculation, and scientific research. The output results are linked with the measurement instrument system, laboratory management system, and cloud database to achieve real-time updates and automated management. Through the above operations, the gamma energy spectrum data quality assessment results are made intuitive, operable, and automated, which not only improves data processing efficiency but also provides reliable data quality assurance for long-term monitoring and large-scale measurements.
[0052] On the other hand, the technical solution of the present invention also provides an automatic assessment and grading system for gamma-ray spectral data quality, comprising:
[0053] Data acquisition module: configured for gamma spectrum acquisition, related metadata recording, and parameter measurement;
[0054] Data processing module: configured for noise filtering, baseline correction, outlier removal, and normalization.
[0055] The indicator calculation and feature extraction module is configured to perform signal-to-noise ratio calculation, peak integrity calculation, background noise calculation, and peak position shift calculation.
[0056] The automatic scoring and grading module is configured to generate a comprehensive quality score, classify data into grades, and output confidence levels.
[0057] Output and Visualization Module: Configured to output quality scores, output grade information, and visualizations of data distribution and trends.
[0058] Compared with the prior art, the present invention has the following advantages:
[0059] 1. This invention enables automatic evaluation and classification of gamma-ray spectrum data quality, avoiding the subjective differences caused by traditional manual interpretation and improving the consistency and reliability of evaluation results.
[0060] 2. This invention uses comprehensive calculation of multi-dimensional quality indicators to quantify and grade energy spectrum data, which can comprehensively reflect the data quality status and improve the accuracy of identifying low-quality data.
[0061] 3. This invention supports batch processing of large-scale gamma-ray spectral data, significantly improving the efficiency of data quality assessment, and is suitable for high-throughput measurement and long-term continuous monitoring scenarios.
[0062] 4. The evaluation and grading results of this invention can be directly used for data screening, anomaly labeling and subsequent analysis, effectively reducing the impact of invalid data on nuclide identification and quantitative analysis results.
[0063] 5. This invention adopts a modular system structure, which is easy to expand and integrate, and can be seamlessly connected with existing measurement systems and data management systems, thus having good engineering application value. Attached Figure Description
[0064] The present invention will be further described below with reference to the accompanying drawings and embodiments.
[0065] Figure 1 This is a schematic diagram of an automatic assessment and classification method and system for gamma-ray spectrum data quality according to the present invention. Detailed Implementation
[0066] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.
[0067]
Example 1
[0068] like Figure 1This invention provides an automatic assessment and classification method for the quality of gamma-ray spectral data, comprising the following steps:
[0069] S1, Data Acquisition and Preprocessing: In this embodiment, a NaI(Tl) gamma spectrum analyzer is used for measurement. A single spectrum contains 1024 channels, and a single measurement takes 60 seconds. A total of 1000 gamma spectrum data points are collected. Let the original spectrum data be:
[0070] ;
[0071] The raw energy spectrum data is preprocessed, firstly by using a window with a length of... Smoothing is performed using a moving average filter:
[0072] ;
[0073] Subsequently, the background baseline was estimated using the sliding minimum method. And perform background subtraction:
[0074] ;
[0075] The energy spectrum data after background subtraction is normalized to its maximum value.
[0076] ;
[0077] After preprocessing, standardized energy spectrum data for quality assessment are obtained;
[0078] S2, Index Calculation and Feature Extraction: In this embodiment, multiple quality indices are calculated as a feature set for the preprocessed energy spectrum data, and the peak signal-to-noise ratio (SNR) is calculated using the following formula:
[0079] ;
[0080] in For the peak energy channel range, The standard deviation of the background region;
[0081] Peak integrity is calculated as the ratio of the actual peak count to the Gaussian fitted count:
[0082] ;
[0083] Background noise level is defined as:
[0084] ;
[0085] The peak shift is calculated as follows:
[0086] ;
[0087] Therefore, a set of energy spectrum quality characteristics is constructed:
[0088] ;
[0089] S3, Automatic Quality Scoring: In this embodiment, a weighted scoring method is used to comprehensively score the feature set, with the weights of each indicator as follows: , , , Overall quality score The calculation is as follows:
[0090] ;
[0091] in The score is a normalized index value, and the score is used to characterize the overall quality level of a single energy spectrum data.
[0092] S4, Data Classification and Labelling: In this embodiment, the energy spectrum data is classified according to the comprehensive quality score, and the classification rules are as follows:
[0093] ;
[0094] The system generates a corresponding quality level label for each energy spectrum data point, which is used for subsequent data filtering and management;
[0095] S5, Results Output and Visualization: In this embodiment, the mass fraction and grading results are stored in the database, and the distribution of energy spectrum data of different quality levels and the trend of mass change are displayed in the form of bar charts and line charts, providing an intuitive basis for the screening and analysis of measurement data.
[0096]
Example 2
[0097] In a typical embodiment of the present invention, this embodiment discloses an automatic assessment and grading system for gamma-ray spectral data quality, comprising:
[0098] Data acquisition module: configured for gamma spectrum acquisition, related metadata recording, and parameter measurement;
[0099] Data processing module: configured for noise filtering, baseline correction, outlier removal, and normalization.
[0100] The indicator calculation and feature extraction module is configured to perform signal-to-noise ratio calculation, peak integrity calculation, background noise calculation, and peak position shift calculation.
[0101] The automatic scoring and grading module is configured to generate a comprehensive quality score, classify data into grades, and output confidence levels.
[0102] Output and Visualization Module: Configured to output quality scores, output grade information, and visualizations of data distribution and trends.
[0103] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention can have various modifications and variations; any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. An automatic assessment and classification method for the quality of gamma-ray spectral data, characterized in that: Includes the following steps: S1, Data Acquisition and Preprocessing: The sample and environment are measured by a gamma energy spectrum measuring instrument to obtain raw gamma energy spectrum data. After the data acquisition is completed, the raw data is preprocessed. The preprocessing process includes noise filtering of the acquired signal, correction of the energy spectrum baseline, automatic removal of outliers, and normalization of the energy channel data. S2, Index Calculation and Feature Extraction: After preprocessing, quality indicators and features are calculated for each gamma spectrum data. By quantitatively calculating key features such as peak signal-to-noise ratio, peak integrity, background noise level, and peak position drift amplitude, a set of indicators for quality assessment is formed. Combined with historical high-quality data, data quality features are extracted through statistical analysis and machine learning methods to ensure that the indicators can comprehensively reflect the quality status of the energy spectrum data. S3, Automatic Quality Scoring: Based on the results of index calculation and feature extraction, the gamma spectrum data is automatically scored. The scoring process uses a weighted scoring algorithm and machine learning model to comprehensively calculate various quality indicators and generate an overall quality score for a single data point. The scoring model is trained and calibrated using historical data. S4, Data Classification and Labeling: Based on the calculated quality score, the gamma spectrum data is divided into different levels, and each data point is labeled accordingly. The classification standard is determined based on the numerical range output by the scoring model. The different levels include excellent, good, medium and poor. S5, Output and Visualization: Output the quality score and grading results to an information system, which includes a database and a visualization terminal, and provides data query, statistical and visualization analysis functions.
2. The method for automatic assessment and classification of gamma-ray spectral data quality according to claim 1, characterized in that: In the S1 data acquisition and preprocessing process, the gamma spectrum data is first obtained by measuring instruments to obtain the raw energy spectrum signal sequence. ,in This indicates the number of channels collected, each... To determine the corresponding channel count values, during preprocessing, a moving average filter is first used to filter noise in the original signal, using a window length of [value missing]. The formula for moving average filtering is: ； in, The filtered signal value is filled with symmetrical padding at the boundaries. Secondly, the sliding minimum is used to target the energy spectrum baseline. The signal is corrected. Represented as: ； in, The count value is the baseline corrected value; To eliminate the influence of differences in measurement intensity, the channel count is normalized to its maximum value: ； Through this normalization step, the energy spectrum data is converted into an interval. The values within the range ensure that subsequent quality indicator calculations are not affected by the magnitude of the absolute count value; For isolated peaks with abnormal channel counts detected, a threshold is set. Remove: ； in, The mean of the neighborhood energy channels. This is the final processed energy spectrum data.
3. The method for automatic assessment and classification of gamma-ray spectral data quality according to claim 1, characterized in that: In the S2 index calculation and feature extraction, the preprocessed gamma spectrum data To calculate the key metrics used for quality assessment, first, calculate the peak signal-to-noise ratio (SNR): ； in, This represents the energy channel set corresponding to the spectral peak. For the peak energy level, SNR is the standard deviation of the baseline region count values. The higher the SNR, the more obvious the spectral peak signal and the higher the data quality. Secondly, calculate peak integrity: ； in, The values are the Gaussian fitting results for the spectral peak region. The closer the value is to 1, the more complete the spectral peak shape is and the less obvious the defects are. Next, calculate the background noise level (NoiseLevel): ； in, The mean of the counts for the background region. The standard deviation of the count in the background region; the lower the value, the more stable the background noise. In addition, peak shift is calculated: ； in, To observe the center value of the energy channel of the spectral peak, For reference, the center value of the energy channel of the spectral peak, A smaller value indicates better peak stability. A set of quantitative characteristics is formed through calculations using the above indicators. This is used for subsequent quality scoring and grading.
4. The method for automatic assessment and classification of gamma-ray spectral data quality according to claim 1, characterized in that: In the S3 automatic quality scoring, the feature set is based on the indicator calculation and feature extraction steps: ； Each gamma-ray spectrum data point is comprehensively scored to quantify data quality. A weighted scoring method is used in the scoring process, assigning different weights to each indicator according to its importance to data quality. Calculate the overall quality score : ； in, For the first The normalized value of each quality indicator. For the corresponding weights, satisfying , The total number of indicators; Historical high-quality data is used as the training set to build a quality scoring model using machine learning algorithms. The input to the scoring model is the feature set. The output is the overall quality score. and the probability of level : ； in, This indicates a trained scoring model that automatically calculates data quality scores based on input metrics and makes predictions for new data.
5. The method for automatic assessment and classification of gamma-ray spectral data quality according to claim 1, characterized in that: In the S4 data grading and labeling process, the comprehensive quality score is obtained based on the automatic quality scoring step. The gamma-ray spectral data was divided into different levels to visually demonstrate data quality and guide subsequent analysis. The data grading used a threshold system, mapping the overall quality score to four levels: Excellent, Good, Fair, and Poor. The specific divisions are as follows: ； in, These are preset mass fraction thresholds, which are obtained through statistical analysis of historical high-quality data to ensure that the classification results can reasonably reflect the actual quality of the energy spectrum data. Generate a corresponding level label for each data point. With confidence level Confidence level indicates the reliability of the rating model's determination of the level: ； in, For the first The overall quality score of the data points A higher value indicates a more reliable classification. Furthermore, by combining a batch processing mechanism to simultaneously score and classify a large amount of gamma spectrum data, automated management of data quality is achieved. At the same time, statistical analysis is performed on data of each level to generate data quality distribution charts.
6. The method for automatic assessment and classification of gamma-ray spectral data quality according to claim 1, characterized in that: In the S5 output and visualization process, the automatic quality scoring and data classification results are output to the database terminal system for operators to query, statistically analyze, and perform subsequent analysis. The output includes the comprehensive quality score for each gamma-ray spectrum data point. Corresponding level and confidence level At the same time, a set of quality indicators is generated. The system provides detailed records and a visual interface to graphically display the scoring and grading results. The visualization methods include bar charts, line charts, heat maps, and radar charts. In addition, operators can perform batch filtering of data based on grade labels and quality score thresholds, removing low-quality data and individually labeled abnormal data. The output results are linked with the information management and analysis platform to achieve real-time updates and automated management. The information management and analysis platform includes a measuring instrument system, a laboratory management system, and a cloud database.
7. An automatic assessment and grading system for gamma-ray spectral data quality, characterized in that: include: Data acquisition module: includes gamma spectrum acquisition, related metadata recording, and parameter measurement; Data processing module: configured for noise filtering, baseline correction, outlier removal, and normalization. The indicator calculation and feature extraction module is configured to perform signal-to-noise ratio calculation, peak integrity calculation, background noise calculation, and peak position shift calculation. The automatic scoring and grading module is configured to generate a comprehensive quality score, classify data into grades, and output confidence levels. Output and Visualization Module: Configured to output quality scores, output grade information, and visualizations of data distribution and trends.