Radiographic map feature extraction method, system, equipment and medium
By employing multi-scale analysis and feature optimization, the problem of misjudgment in the identification of seafood radionuclide features under harsh environments using traditional methods has been solved, achieving high-precision extraction of radiometric spectral features and improving the accuracy of identification and analysis.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-18
- Publication Date
- 2026-04-21
AI Technical Summary
Traditional machine learning methods struggle to effectively distinguish the characteristics of radionuclides in seafood under harsh detection environments, leading to misjudgments or missed detections. In particular, the quality of spectral images deteriorates under high-dimensional, high-noise data conditions, failing to meet the needs of practical applications.
Through steps such as multi-scale analysis, feature extraction, decomposition, noise reduction, dimensionality reduction, and weight optimization, a hyperspectral radiometric spectrum dataset is obtained, and a radiometric spectrum feature set is generated. This includes resolution generation, multi-scale analysis, wavelet transform, median filtering, principal component analysis, and backpropagation algorithm optimization to improve feature extraction accuracy.
It enhances the characterization ability of radionuclides, improves the accuracy of radioactive contamination identification, can more accurately reflect the characteristics of radioactive substances, reduces false positives and false negatives, and improves the accuracy of detection.
Smart Images
Figure CN121901689A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of image processing, and in particular to a method, system, device and medium for extracting radiometric spectral features. Background Technology
[0002] In the fields of import and export inspection and food safety monitoring, especially for seafood products, accurate detection of radioactive contamination and rapid analysis of radioactive spectra to identify radioactive features are crucial for developing effective countermeasures. However, in harsh testing environments, the limitations of traditional machine learning methods in handling high-dimensional, high-noise data become apparent when faced with complex seafood radioactive spectra. These methods struggle to effectively distinguish the characteristics of different radionuclides, easily leading to misjudgments or missed detections. For example, in noisy environments, noise directly degrades the quality of spectra generated by existing feature extraction methods, blurring key features and ultimately resulting in recognition and analysis accuracy failing to meet practical application requirements. Summary of the Invention
[0003] This application aims to improve at least one technical problem in the background art.
[0004] This application provides a method for extracting radiometric spectral features, which includes: Obtain hyperspectral radiometric dataset; Multi-scale analysis and feature extraction were performed on the hyperspectral radiometric dataset to obtain a radiometric feature set; The hyperspectral radiometric dataset is decomposed to obtain the first sub-band coefficient and the second sub-band coefficient; The hyperspectral radiometric dataset is denoised based on the first sub-band coefficient and the second sub-band coefficient to obtain a denoised dataset. The denoised dataset is subjected to dimensionality reduction processing to obtain principal component feature vectors; A weighted dataset is generated based on the pre-defined characteristics of radioactive materials. The weight dataset is optimized to obtain an optimized weight dataset; Based on the optimized weight dataset, feature extraction is performed on the principal component feature vector to obtain the radiometric feature set.
[0005] According to some technical solutions of this application, the acquisition of hyperspectral radiometric dataset includes: The acquisition resolution is generated based on the preset wavelength range; Based on the acquisition resolution and preset time interval, corresponding data are extracted from the preset hyperspectral radiometric spectrum raw data to obtain a hyperspectral radiometric spectrum dataset.
[0006] According to some technical solutions of this application, the step of decomposing the hyperspectral radiometric dataset to obtain the first sub-band coefficient and the second sub-band coefficient includes: The hyperspectral radiometric dataset is analyzed according to the preset number of decomposition layers to obtain the analysis results; Based on the analysis results, the first sub-band coefficient and the second sub-band coefficient are generated.
[0007] According to some technical solutions of this application, the step of denoising the hyperspectral radiometric dataset based on the subband coefficient to obtain a denoised dataset includes: Generate filtered median data based on preset time series data and preset filter window size; The median data from the filter is mapped to obtain the mapping interval. The hyperspectral radiometric dataset is denoised based on the first sub-band coefficient and the second sub-band coefficient to obtain the initial denoised data. The initial denoised data is reconstructed using wavelet transform and the mapping interval to obtain the denoised dataset.
[0008] According to some technical solutions of this application, the step of optimizing the weight dataset to obtain an optimized weight dataset includes: The weight dataset is optimized based on the preset backpropagation algorithm and the noise reduction dataset to obtain the optimized weight dataset.
[0009] According to some technical solutions of this application, the step of performing dimensionality reduction processing on the denoised dataset to obtain principal component feature vectors includes: The noise reduction dataset is dimensionality reduced based on the preset characteristics of radioactive materials to obtain the first principal component feature vector and the second principal component feature vector. The principal component eigenvectors are obtained by integrating the eigenvectors of the first and second principal components.
[0010] According to some technical solutions of this application, the step of extracting features from the principal component feature vector based on the optimized weight dataset to obtain a radiometric feature set includes: The principal component feature vectors are weighted and fused according to the optimized weight dataset to obtain a weighted feature vector; The weighted feature vector is mapped and calculated according to a preset mapping function to obtain the radioactive spectrum feature set.
[0011] Secondly, a radiometric spectrum feature extraction system is provided, comprising: an acquisition module, an analysis module, a decomposition module, a noise reduction module, a dimensionality reduction module, a generation module, an optimization module, and... Extraction module; The system comprises the following modules: an acquisition module for acquiring a hyperspectral radiometric spectrum dataset; an analysis module for performing multi-scale analysis and feature extraction on the hyperspectral radiometric spectrum dataset to obtain a radiometric spectrum feature set; a decomposition module for decomposing the hyperspectral radiometric spectrum dataset to obtain subband coefficients; a denoising module for denoising the hyperspectral radiometric spectrum dataset based on the subband coefficients to obtain a denoised dataset; a dimensionality reduction module for dimensionality reduction of the denoised dataset to obtain principal component feature vectors; a generation module for generating a weighted dataset based on preset radiometric material characteristics; an optimization module for optimizing the weighted dataset to obtain an optimized weighted dataset; and an extraction module for extracting features from the principal component feature vectors based on the optimized weighted dataset to obtain the radiometric spectrum feature set.
[0012] Thirdly, a radiometric feature extraction device is provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a radiometric feature extraction method as described in the above embodiments.
[0013] Fourthly, a computer-readable storage medium is provided having a computer program stored thereon, which is executed by a processor to implement the radiometric feature extraction method as described in the above embodiments.
[0014] The radiometric spectrum feature extraction method provided in this application has at least the following beneficial effects: by performing multi-scale feature extraction on the radiometric spectrum, the feature information of the radiometric spectrum can be comprehensively obtained from different frequencies and spatial resolutions, thereby enhancing the feature characterization ability of key radionuclides, making the extracted features more accurately reflect the characteristics of radioactive substances, and thus providing conditions for improving the identification accuracy of radioactive pollution. Attached Figure Description
[0015] Figure 1 This is a schematic diagram of the first process of the radioactive spectrum feature extraction method provided in the embodiments of this application; Figure 2 A schematic diagram of a second process for a radiometric spectrum feature extraction method provided in an embodiment of this application; Figure 3 A schematic diagram of the third process of the radioactive spectrum feature extraction method provided in the embodiments of this application; Figure 4 A schematic diagram of the fourth process of the radioactive spectrum feature extraction method provided in the embodiments of this application; Figure 5 A schematic diagram of the fifth process for the radioactive spectrum feature extraction method provided in the embodiments of this application; Figure 6 A schematic diagram of the sixth process of the radioactive spectrum feature extraction method provided in the embodiments of this application; Figure 7 A schematic diagram of the seventh process for the radioactive spectral feature extraction method provided in the embodiments of this application; Figure 8 A structural diagram of a radioactive map feature extraction system provided in an embodiment of this application; Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0016] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.
[0017] In the description of this application, it should be understood that the orientation descriptions, such as up, down, front, back, left, right, etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation or be constructed or operated in a specific orientation. Therefore, they should not be construed as limitations on this application.
[0018] In the description of this application, unless otherwise expressly defined, terms such as "setup," "installation," and "connection" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this application in conjunction with the specific content of the technical solution.
[0019] The following is combined with Figures 1 to 9 The embodiments of this application are described below.
[0020] This application provides a method for extracting radioactive spectral features, specifically including: S100: Acquire a hyperspectral radiometric spectrum dataset. Based on this dataset, and according to preset parameters such as wavelength range and time interval, read data from the raw data storage device and perform preliminary processing to provide a data foundation for subsequent processing. In practical applications, the acquisition module first reads the raw hyperspectral radiometric spectrum data from connected sensors, databases, or other data storage devices according to preset parameters such as wavelength range and time interval. For example, when detecting radioactivity in seafood products, the acquisition module connects to a hyperspectral radiometric sensor placed in the seafood detection area. It acquires data in real time according to preset parameters of once every 5 seconds, wavelength range [300nm, 1000nm], and resolution of 1nm. The data is then preliminarily processed and formatted into a hyperspectral radiometric spectrum dataset format that the system can process, and then analyzed.
[0021] In some embodiments, step S100, obtaining the hyperspectral radiometric dataset, includes: S110 generates the acquisition resolution based on the preset wavelength range; S120: Based on the acquisition resolution and preset time interval, corresponding data is extracted from the preset hyperspectral radiometric spectrum raw data to obtain a hyperspectral radiometric spectrum dataset. Specifically, in practical applications, the preset wavelength range can be determined according to the target radioactive substance and the capability of the detection equipment. For example, when mainly detecting radionuclides such as cesium-137 and strontium-90 in seafood products, the characteristic wavelengths produced by their radiation have a certain range. The preset wavelength range is set to [300nm, 1000nm]. Based on the spectral resolution limitations of the equipment and the detection accuracy requirements, the acquisition resolution is calculated to be 1nm. The preset time interval is determined according to the real-time requirements of the detection and the performance of the data acquisition system, such as being set to 5 seconds. From the database or file system storing the hyperspectral radiometric spectrum raw data, the corresponding data is extracted according to the time order and wavelength resolution to obtain the hyperspectral radiometric spectrum dataset. For example, the raw data is stored in a two-dimensional array indexed by timestamps and wavelength values. By traversing the time range and wavelength range, data that meets the conditions is extracted, thus forming the hyperspectral radiometric spectrum dataset.
[0022] S200, perform multi-scale analysis and feature extraction on the hyperspectral radiometric spectrum dataset to obtain a radiometric spectrum feature set; exemplarily, use multi-scale analysis methods, such as wavelet transform or pyramid decomposition, to analyze the obtained hyperspectral radiometric spectrum dataset. In wavelet transform, select appropriate wavelet basis functions to decompose the dataset at different scales, obtaining sub-band coefficients at different frequencies. Through analysis and processing of these sub-band coefficients, extract multi-scale features of the radiometric spectrum to form a preliminary radiometric spectrum feature set.
[0023] Specifically, multi-scale analysis methods are used to process the acquired hyperspectral radiometric spectrum dataset. For example, wavelet transform is used, employing the Daubechies wavelet basis function, which exhibits good time-frequency localization characteristics in signal processing and is suitable for analyzing non-stationary signals, such as radiometric spectrum data. The decomposition scale is set to four levels. By performing wavelet decomposition on the dataset at different scales, subband coefficients at different frequencies are obtained. In the first level of decomposition, the original dataset is decomposed into a low-frequency subband A1 and a high-frequency subband D1. The low-frequency subband A1 contains the main trend and low-frequency components of the signal, while the high-frequency subband D1 contains the details and high-frequency components of the signal. As the decomposition level increases, the low-frequency subband is further decomposed; for example, the second level decomposes A1 into A2 and D2. Thus, through analysis and processing of these subband coefficients, such as calculating the energy, mean, variance, and other statistical characteristics of the subband coefficients, multi-scale features of the radiometric spectrum are extracted, forming a preliminary radiometric spectrum feature set. Therefore, multi-scale analysis and feature extraction were performed on the collected hyperspectral radiometric spectrum dataset. Various multi-scale analysis algorithms and feature extraction methods were used to extract the key features of the radiometric spectrum from the dataset, resulting in a preliminary radiometric spectrum feature set.
[0024] S300, decompose the hyperspectral radiometric dataset to obtain the first sub-band coefficient and the second sub-band coefficient; In some embodiments, in S300, the step of decomposing the hyperspectral radiometric dataset to obtain subband coefficients includes: S310, Analyze the hyperspectral radiometric dataset according to the preset number of decomposition layers to obtain the analysis results; S320, Based on the analysis results, first sub-band coefficients and second sub-band coefficients are generated. The hyperspectral radiometric dataset is analyzed according to a preset decomposition level N. For example, the decomposition level N is set to 5, and the dataset is decomposed using Discrete Wavelet Transform (DWT). DWT is an effective method for decomposing a signal into different frequency sub-bands; it filters and downsamples the signal using low-pass and high-pass filters. After 5 levels of decomposition, a series of low-frequency and high-frequency sub-bands are obtained. Based on these decomposition results, first and second sub-band coefficients are generated. For example, using the coefficients of the low-frequency sub-bands as the first sub-band coefficients can reflect the characteristics of the low-frequency smooth portion of the signal; using the coefficients of the high-frequency sub-bands as the second sub-band coefficients can reflect the characteristics of the high-frequency detailed portion of the signal.
[0025] Therefore, the hyperspectral radiometric dataset is decomposed according to the preset number of decomposition layers, and an appropriate decomposition algorithm, such as discrete wavelet transform, is used to generate the first sub-band coefficients and the second sub-band coefficients, thereby providing data support for noise reduction processing.
[0026] S400, the hyperspectral radiometric dataset is denoised based on the first sub-band coefficient and the second sub-band coefficient to obtain a denoised dataset; In some embodiments, in step S400, the denoising process performed on the hyperspectral radiometric dataset based on the first sub-band coefficient and the second sub-band coefficient to obtain a denoised dataset includes: S410, generates filtered median data based on preset time series data and preset filter window size; S420: Perform mapping calculations on the filtered median data to obtain the mapping interval; S430, Denoise the hyperspectral radiometric dataset based on the first sub-band coefficient and the second sub-band coefficient to obtain initial denoised data; S440, Perform wavelet reconstruction on the initial denoised data according to the wavelet transform method and the mapping interval to obtain the denoised dataset.
[0027] Specifically, filtered median data is generated based on preset time-series data and a preset filter window size W. Assume the preset time-series data consists of radioactive data collected every 5 seconds over the past hour, and the filter window size W is set to 11. For a given time-series data, the window slides with a size of 11, calculating the median of the data within each window to obtain the filtered median data. For example, for time-series data [x1,x2,x3,…,xn], when the window starts sliding from x1, the median of [x1,x2,…,x11] is calculated as the first filtered median data. Then, the window moves one data point to the right, and the median of [x2,x3,…,x12] is calculated as the next filtered median data. The filtered median data is then mapped, for example using a linear mapping function y=ax+b, where a and b are determined based on the data range and the desired mapping interval. For example, a=2, b=-1, mapping the filtered median data to the interval [-1,1] to obtain the mapping interval. The hyperspectral radiometric dataset is denoised based on the first and second sub-band coefficients. The characteristics of the sub-band coefficients are utilized to remove noise interference, yielding initial denoised data. Then, wavelet transform and mapping intervals are used to reconstruct the initial denoised data. By selecting appropriate wavelet basis functions and reconstruction algorithms, the integrity of the signal is restored, resulting in the denoised dataset. Therefore, denoising the hyperspectral radiometric dataset based on sub-band coefficients, by generating filtered median data, calculating mapping intervals, and combining wavelet transform methods, denoises and reconstructs the dataset, resulting in a denoised dataset that improves data quality and reliability.
[0028] S500, Dimensionality reduction is performed on the denoised dataset to obtain principal component feature vectors; In some embodiments, S500, the dimensionality reduction processing of the denoised dataset to obtain the principal component feature vector includes: S510, the dimensionality reduction of the noise reduction dataset is performed according to the preset characteristics of radioactive materials to obtain the first principal component feature vector and the second principal component feature vector. S520, integrate the first principal component eigenvector and the second principal component eigenvector to obtain the principal component eigenvector.
[0029] Specifically, the dimensionality reduction of the denoised dataset is performed based on the pre-defined characteristics of the radioactive materials. For example, using principal component analysis, the main radioactive materials of interest in seafood products are cesium-137 and strontium-90, whose characteristic wavelengths and intensity distributions in the radiometric spectrum exhibit certain patterns. First, the covariance matrix of the denoised dataset is calculated, reflecting the correlation between the variables in the dataset. Then, the eigenvalues and eigenvectors of the covariance matrix are solved. The first K eigenvectors are selected based on the magnitude of the eigenvalues. For example, when K=3, the denoised dataset is projected onto these three eigenvectors to obtain the first principal component eigenvector and the second principal component eigenvector. The first and second principal component eigenvectors are then integrated to obtain the principal component eigenvector. For example, by calculating the eigencorrelation matrix of the two, a weighted concatenation method is used to integrate them to obtain the principal component eigenvector containing multi-dimensional key information, thereby achieving data dimensionality reduction.
[0030] Therefore, the denoised dataset is dimensionality reduced by using principal component analysis and other dimensionality reduction algorithms based on the pre-defined characteristics of radioactive materials, thereby obtaining the first principal component eigenvector and the second principal component eigenvector, reducing data dimensionality and computational complexity.
[0031] S600 generates a weighted dataset based on preset radioactive material characteristics. Specifically, it generates the dataset based on preset radioactive material characteristics, such as the types, half-lives, and radiation intensities of different radionuclides. For key radionuclides such as cesium-137 and strontium-90, due to their significant impact on human health and the environment, and their prominent characteristics in radiometric spectra, higher weights are assigned to their corresponding spectral data points through quantitative analysis based on their radiation intensity and importance in actual detection. For example, a weight of 0.8 is assigned to the spectral data point corresponding to the characteristic wavelength of cesium-137, based on the relative magnitude of its radiation intensity; a weight of 0.7 is assigned to the spectral data point corresponding to the characteristic wavelength of strontium-90. Lower weights, such as 0.1, are assigned to spectral data points corresponding to background noise or interference signals. Thus, a weighted dataset is generated based on preset radioactive material characteristics, and a weight is assigned to each element in the dataset through the analysis and quantification of these characteristics.
[0032] S700, The weight dataset is optimized to obtain an optimized weight dataset; In some embodiments, step S700, optimizing the weight dataset to obtain an optimized weight dataset, includes: S710, the weight dataset is optimized according to a preset backpropagation algorithm and the denoised dataset to obtain an optimized weight dataset. Specifically, the weight dataset is optimized according to the preset backpropagation algorithm and the denoised dataset. Using the backpropagation algorithm, the gradient of the weight dataset is calculated, and the weights are adjusted based on the gradient information. During backpropagation, a learning rate of 0.01 is set; this is a parameter that controls the step size of weight updates, affecting the optimization speed and effect. Through multiple iterations, based on the interaction between the denoised dataset and the weight dataset, the weights are continuously adjusted so that the weight dataset can better reflect the characteristics of the radiometric spectrum, thus obtaining the optimized weight dataset.
[0033] Therefore, the weight dataset is optimized by using a pre-defined backpropagation algorithm and a denoised dataset to adjust the parameters of the weight dataset so that it can more accurately reflect the characteristics of the radiometric spectrum, thus obtaining an optimized weight dataset.
[0034] S800, based on the optimized weight dataset, feature extraction is performed on the principal component feature vector to obtain the radiometric spectrum feature set.
[0035] In some embodiments, in S800, the step of extracting features from the principal component feature vectors based on the optimized weight dataset to obtain a radiometric feature set includes: S810, The principal component feature vectors are weighted and fused according to the optimized weight dataset to obtain a weighted feature vector; S820: The weighted feature vectors are mapped and calculated according to a preset mapping function to obtain the radiometric spectrum feature set. The principal component feature vectors are then weighted and fused according to the optimized weight dataset. The weights in the optimized weight dataset are multiplied by the corresponding elements of the principal component feature vectors and summed to obtain the weighted feature vector. For example, assuming there are three principal component feature vectors PC1, PC2, and PC3, with corresponding weights w1, w2, and w3, then the weighted feature vector WF = w1. PC1+w2 PC2+w3 PC3. The weighted feature vectors are mapped using a preset mapping function, assuming the preset mapping function is y=sigmoid(x)=1 / (1+e^(-x)). Each element of the weighted feature vector is substituted into this function for calculation, yielding the final radiometric feature set. Therefore, based on the optimized weight dataset, features are extracted from the principal component feature vectors. Through weighted fusion and mapping calculation, the final radiometric feature set is obtained, providing crucial data for subsequent analysis and applications.
[0036] Therefore, by extracting multi-scale features from seafood radiometric spectra such as gamma-ray spectroscopy and X-ray fluorescence spectroscopy, and enhancing the characterization capabilities of key radionuclides such as cesium-137 and strontium-90, the feature extraction process can be automatically optimized, reducing manual intervention and significantly improving feature detection accuracy and generalization ability. This solves the problem of radiometric spectrum data extraction and identification, enabling more efficient and accurate identification and analysis of radioactive contamination characteristics in seafood products.
[0037] Secondly, see Figure 8 This application provides a system for extracting radioactive spectral features. It includes: acquisition module 100, analysis module 200, decomposition module 300, noise reduction module 400, dimensionality reduction module 500, generation module 600, optimization module 700 and extraction module 800; The system comprises the following modules: acquisition module 100 acquires a hyperspectral radiometric spectrum dataset; analysis module 200 performs multi-scale analysis and feature extraction on the hyperspectral radiometric spectrum dataset to obtain a radiometric spectrum feature set; decomposition module 300 decomposes the hyperspectral radiometric spectrum dataset to obtain subband coefficients; denoising module 400 performs denoising processing on the hyperspectral radiometric spectrum dataset based on the subband coefficients to obtain a denoised dataset; dimensionality reduction module 500 performs dimensionality reduction processing on the denoised dataset to obtain principal component feature vectors; generation module 600 generates a weighted dataset based on preset radiometric material characteristics; optimization module 700 optimizes the weighted dataset to obtain an optimized weighted dataset; and extraction module 800 extracts features from the principal component feature vectors based on the optimized weighted dataset to obtain the radiometric spectrum feature set.
[0038] Figure 9 This is a schematic diagram of a radiometric feature extraction device 900 provided in an embodiment of this application. The radiometric feature extraction device 900 can vary significantly due to different configurations or performance characteristics. It may include one or more central processing units (CPUs) 910 (e.g., one or more processors) and a memory 920, and one or more storage media 930 (e.g., one or more mass storage devices) for storing application programs 933 or data 932. The memory 920 and storage media 930 can be temporary or persistent storage. The program stored in the storage media 930 may include one or more modules (not shown in the diagram), and each module may include a series of instruction operations on the radiometric feature extraction device 900.
[0039] Furthermore, the processor 910 can be configured to communicate with the storage medium 930 and execute a series of instruction operations in the storage medium 930 on the radiometric spectrum feature extraction device 900 to implement the steps of the radiometric spectrum feature extraction method provided in the above-described method embodiments.
[0040] The radiometric feature extraction device 900 may also include one or more power supplies 940, one or more wired or wireless network interfaces 950, one or more input / output interfaces 960, and / or one or more operating systems 931, such as Windows Server, Mac OSX, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 9 The illustrated electronic device structure does not constitute a limitation on the electronic device and may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.
[0041] This application also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the step of extracting radiometric spectral features.
[0042] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system, device, or unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0043] The preferred embodiments of this application have been described in detail above, but this disclosure is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this disclosure.
Claims
1. A method for extracting features from a radiometric spectrum, characterized in that: include: Obtain a hyperspectral radiometric dataset; Multi-scale analysis and feature extraction were performed on the hyperspectral radiometric dataset to obtain a radiometric feature set; The hyperspectral radiometric dataset is decomposed to obtain the first sub-band coefficient and the second sub-band coefficient; The hyperspectral radiometric dataset is denoised based on the first sub-band coefficient and the second sub-band coefficient to obtain a denoised dataset. The denoised dataset is subjected to dimensionality reduction processing to obtain principal component feature vectors; A weighted dataset is generated based on the pre-defined characteristics of radioactive materials. The weight dataset is optimized to obtain an optimized weight dataset; Based on the optimized weight dataset, feature extraction is performed on the principal component feature vector to obtain the radiometric feature set.
2. The method for extracting radioactive spectral features according to claim 1, characterized in that: The acquisition of the hyperspectral radiometric dataset includes: The acquisition resolution is generated based on the preset wavelength range; Based on the acquisition resolution and preset time interval, corresponding data are extracted from the preset hyperspectral radiometric spectrum raw data to obtain a hyperspectral radiometric spectrum dataset.
3. The method for extracting radioactive spectral features according to claim 1, characterized in that: The step of decomposing the hyperspectral radiometric dataset to obtain the first sub-band coefficient and the second sub-band coefficient includes: The hyperspectral radiometric dataset is analyzed according to the preset number of decomposition layers to obtain the analysis results; Based on the analysis results, the first sub-band coefficient and the second sub-band coefficient are generated.
4. The method for extracting radioactive spectral features according to claim 1, characterized in that: The hyperspectral radiometric dataset is denoised based on the first sub-band coefficient and the second sub-band coefficient to obtain a denoised dataset, including: Generate filtered median data based on preset time series data and preset filter window size; The median data from the filter is mapped to obtain the mapping interval. The hyperspectral radiometric dataset is denoised based on the first sub-band coefficient and the second sub-band coefficient to obtain the initial denoised data. The initial denoised data is reconstructed using wavelet transform and the mapping interval to obtain the denoised dataset.
5. The method for extracting radioactive spectral features according to claim 1, characterized in that: The optimization process of the weight dataset to obtain an optimized weight dataset includes: The weight dataset is optimized based on the preset backpropagation algorithm and the noise reduction dataset to obtain the optimized weight dataset.
6. The method for extracting radioactive spectral features according to claim 1, characterized in that: The dimensionality reduction process of the denoised dataset to obtain the principal component feature vector includes: The noise reduction dataset is dimensionality reduced based on the preset characteristics of radioactive materials to obtain the first principal component feature vector and the second principal component feature vector. The principal component eigenvectors are obtained by integrating the eigenvectors of the first and second principal components.
7. The method for extracting radiometric spectral features according to claim 1, characterized in that: The step of extracting features from the principal component feature vectors based on the optimized weight dataset to obtain a radiometric feature set includes: The principal component feature vectors are weighted and fused according to the optimized weight dataset to obtain a weighted feature vector; The weighted feature vector is mapped and calculated according to a preset mapping function to obtain the radioactive spectrum feature set.
8. A system for extracting radiometric spectral features, characterized in that: include: The acquisition module is used to acquire hyperspectral radiometric datasets. The analysis module is used to perform multi-scale analysis and feature extraction on the hyperspectral radiometric dataset to obtain a radiometric feature set; The decomposition module is used to decompose the hyperspectral radiometric dataset to obtain sub-band coefficients; The noise reduction module is used to perform noise reduction processing on the hyperspectral radiometric dataset based on the subband coefficients to obtain a noise-reduced dataset. The dimensionality reduction module is used to perform dimensionality reduction processing on the denoised dataset to obtain the principal component feature vectors. The generation module is used to generate a weighted dataset based on preset characteristics of radioactive materials. The optimization module is used to optimize the weight dataset to obtain an optimized weight dataset. The extraction module is used to extract features from the principal component feature vectors based on the optimized weight dataset to obtain the radiometric feature set.
9. A device for extracting radiometric spectral features, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the radiometric feature extraction method as described in any one of claims 1-7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the radiometric feature extraction method as described in any one of claims 1-7.