A visual analytics system and method for astronomical spectral type inspection

Abstract

The present application relates to a kind of visual analysis system and method for astronomical spectral type inspection, belong to the field of visualization and man-machine interaction, the system includes: data preprocessing module is used to the reference spectrum and the pre-processing of the spectrum to be checked, completes the normalization of spectrum, spectrum feature detection and spectrum characterization;Selection module is used to help user select a spectrum to be checked, and determine possible red shift and the important spectral line under the red shift;Checking module is used to help user determine the red shift and type of the spectrum to be checked, checking module includes adjustment submodule, recommendation submodule and verification submodule, respectively for adjusting the red shift of the spectrum to be checked, recommend suitable spectral line to user and evaluate the inspection result;Generalization module is used to help user obtain other spectrum similar to the checked spectrum by searching.The system and method provided by the present application can significantly improve the classification efficiency under the premise of ensuring the accuracy of expert spectrum classification.

Description

Technical Field

[0001] This invention belongs to the field of visualization and human-computer interaction technology, specifically a visual analysis system and method for astronomical spectrum type inspection. Background Technology

[0002] In recent years, various sky surveys, such as the LAMOST (Large Sky Area Multi-Object Fiber Spectroscopic Telescope) survey, have generated tens of millions of spectra. Spectral classification is a crucial step before using spectra for scientific research. Due to the sheer volume of spectra, classification is primarily accomplished through automated methods. While automated methods are becoming increasingly sophisticated, they require high spectral quality and cannot guarantee complete accuracy even when processing high-quality spectra. Therefore, visual inspection remains essential for ensuring classification accuracy. However, current tools for visual inspection are relatively simple. They fail to provide experts with effective data organization and analysis strategies to handle large volumes of spectra, and they also lack sufficient support for experts examining individual spectra. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to provide a visual analysis system and method for astronomical spectral type inspection. This system and method follow a "selection-inspection-promotion" process and incorporates various forms of visual guidance to support experts in performing spectral inspections more efficiently while ensuring classification accuracy.

[0004] To achieve the above objectives, the present invention adopts the following technical solution:

[0005] A visual analysis system for astronomical spectral type inspection, characterized in that the system comprises:

[0006] Data preprocessing module: used to preprocess the reference spectrum and the spectrum to be inspected. The data preprocessing module includes a normalization processing submodule, a spectral feature detection submodule, and a spectral characterization submodule, which are used to perform spectral normalization, spectral feature detection, and spectral characterization, respectively.

[0007] Selection module: Used to help users select a spectrum to be examined and determine possible redshifts and important spectral lines at those redshifts;

[0008] The inspection module helps users determine the redshift and type of the spectrum to be inspected. The inspection module includes an adjustment submodule, a recommendation submodule, and a verification submodule, which are used to adjust the redshift of the spectrum to be inspected, recommend suitable spectral lines to the user, and evaluate the inspection results, respectively.

[0009] Promotion module: Used to help users find other spectra similar to the already checked spectra through search.

[0010] Furthermore, in the visual analysis system for astronomical spectral type inspection as described above, the normalization processing submodule normalizes the spectral data to be inspected by: using a minimum-maximum normalization method to adjust the flux range of each spectrum to [0,1].

[0011] Furthermore, in the visual analysis system for astronomical spectral type inspection as described above, the spectral feature detection submodule records the spectral type, wavelength range, flow range, and peak value when performing spectral feature detection. It uses topographic peak value to detect the local maximum flow in the spectrum and uses vertical spectral flipping and topographic peak value to detect the local minimum flow in the spectrum.

[0012] Furthermore, in the visual analysis system for astronomical spectral type inspection as described above, the spectral characterization submodule uses a given redshift to characterize the reference spectrum; for the spectrum to be inspected, the spectral characterization submodule improves the reliability of the representation by selecting n most likely redshift candidates, calculating a representation for each redshift candidate, and combining them together as the final spectral representation.

[0013] Furthermore, in the visual analysis system for astronomical spectral type inspection described above, the selection module completes the selection based on the following two types of information:

[0014] 1) Relationship between reference spectrum and spectrum to be inspected: If the representation of a spectrum to be inspected is located in a region where the representation of a certain type of reference spectrum is concentrated, then the spectrum to be inspected may also belong to this type.

[0015] 2) The type of the spectrum to be inspected is determined by the 1D pipeline and the independent pipeline: If the two types are the same, the spectrum is more likely to be correctly classified; otherwise, the classification may be incorrect.

[0016] Furthermore, in the visual analysis system for astronomical spectral type inspection as described above, the interface of the selection module includes:

[0017] Projection view: Used to show the relationship between the reference spectrum and the spectrum to be examined, with the reference spectrum as the background and the spectrum to be examined as the focus;

[0018] Parallel coordinate view: Used to display the parameters of the spectrum under examination, i.e. the category and redshift given by the 1D pipeline and the independent pipeline;

[0019] Table view: Used to display information for each spectrum to be examined, including the spectrum, the category and redshift given by the 1D pipeline and independent pipelines, and the redshift and type determined by experts;

[0020] Candidate Spectrum View: Displays n candidate redshifts for the selected spectrum to be examined, the important spectral lines of each candidate redshift, and the two reference spectra that are most similar to it.

[0021] Furthermore, in the visual analysis system for astronomical spectral type inspection as described above, the interface of the inspection module mainly includes: a spectral view, a spectral line view, a horizon plot, a local spectral view, a local flux view, and an overall importance view, wherein:

[0022] The adjustment submodule includes the horizon plot and the local flow view. The horizon plot converts the wavelength-flow relationship of each spectral line into a redshift-flow relationship, specifically:

[0023] According to the formula λ=(1+z)λ0, the wavelength range corresponding to the redshift range of each spectral line is calculated, and then the wavelength-flux points in this range are converted into redshift-flux points, where: λ, z, λ0 represent the current wavelength of the spectral line, the spectral redshift, and the stationary wavelength of the spectral line, respectively.

[0024] The local flow view is used to display the flow of each spectral line across multiple wavelengths around the current redshift, for fine-tuning of the redshift.

[0025] Furthermore, in the visual analysis system for astronomical spectral type inspection as described above, the recommendation submodule includes the spectral view and the overall importance view, with recommendation strategies as follows:

[0026] The spectral view displays the wavelength-flux relationship of the spectrum. When the user selects significant features in the spectrum in the spectral view, the system will automatically recommend redshifts of the spectral lines corresponding to the selected features.

[0027] The overall importance view shows the overall importance of all spectral lines at different redshifts, and redshifts with higher overall importance may correspond to the true redshift.

[0028] Furthermore, in the visual analysis system for astronomical spectral type inspection as described above, the verification submodule includes the spectral view, spectral line view, overall importance view, and local spectral view, and the corresponding verification methods are as follows:

[0029] Display the template spectrum in the spectrum view, and check the degree of matching between the current spectrum and the template spectrum in the spectrum view;

[0030] The opacity of each spectral line in the spectral view varies depending on how well it matches the feature. Check the degree of matching between the spectral feature and the spectral line in the spectral view.

[0031] Check the overall importance of the selected spectral line in the overall importance view;

[0032] Examine the local spectral profile of each spectral line in the local spectral view to determine whether the redshift corresponds to a selected, false, or real but misplaced spectral line.

[0033] Furthermore, in the visual analysis system for astronomical spectral type inspection described above, the screening strategy of the promotion module is as follows:

[0034] After a user has inspected a spectrum, the system extracts the user's judgment criteria, namely the selected spectral lines and their importance; using the vectorized judgment criteria as the target pattern, it searches for all spectra to be inspected.

[0035] The user determines whether each spectrum is similar to the corresponding vector at the previously calculated n candidate redshifts. That is, the user extracts an importance vector composed of selected spectral lines from the representation of each candidate redshift; then compares each importance value between the vector and the target mode, thereby removing vectors that have significant differences on a single spectral line.

[0036] Next, the similarity between each retained vector and the target vector is calculated. The user sets a similarity threshold, and vectors with similarity lower than the threshold are filtered out again. If multiple final vectors correspond to the same spectrum, only the vector with the highest similarity is retained.

[0037] Furthermore, in the visual analysis system for astronomical spectral type inspection as described above, the interface of the promotion module is updated from the interface of the selection module, and the update mainly includes:

[0038] Projection view: With the projection of the reference spectrum as the background and the projections generated by the spectrum under inspection and its similar spectra as the foreground, the user can perceive the similarity between the spectrum under inspection and the similar spectra by the proximity of the two, and determine the possible category of the similar spectra by the position of the two on the background contour map.

[0039] Similar Spectral View: In place of the original candidate spectral view, the similar spectral view includes an upper half and a lower half. The upper half displays the spectrum being examined, the identified type, and the redshift. The lower half displays information including similar spectra, their quantity, type statistics, and examination progress.

[0040] A visual analysis method for astronomical spectral type inspection using the visual analysis system for astronomical spectral type inspection as described above includes the following steps:

[0041] S1. Preprocess the spectral data to complete the normalization of the spectrum, detection of spectral features, and spectral characterization;

[0042] S2. Select a spectrum to be examined and determine the possible redshift and the important spectral lines at that redshift;

[0043] S3. Examine the spectrum to determine the redshift and its type;

[0044] S4. Find other spectra similar to the examined spectra by searching.

[0045] The visual analysis system and method for astronomical spectral type inspection described in this invention have the following significant technical advantages:

[0046] This invention first preprocesses the reference spectrum and the spectrum to be inspected using a data preprocessing module, completing spectral normalization, spectral feature detection, and spectral characterization. Then, a selection module selects a spectrum to be inspected and determines possible redshifts and important spectral lines under those redshifts. Next, an inspection module determines the redshift and type of the spectrum to be inspected. Finally, a generalization module obtains other spectra similar to the inspected spectrum. Based on the "selection-inspection-generalization" workflow and various introduced visual guidelines, this system significantly improves the efficiency of spectral inspection while maintaining accuracy comparable to current visual inspection tools. Attached Figure Description

[0047] Figure 1 This is a structural framework diagram of a visual analysis system for astronomical spectral type inspection provided in an embodiment of the present invention;

[0048] Figure 2 for Figure 1 Workflow diagram of the aforementioned visual analytics system;

[0049] Figure 3 for Figure 1 A schematic diagram of the interface for selecting modules;

[0050] Figure 4 for Figure 1 A schematic diagram of the interface of the inspection module;

[0051] Figure 5 for Figure 1 A schematic diagram of the interface of the promotion module;

[0052] Figure 6 This is a flowchart of a visual analysis method for astronomical spectral type inspection provided in an embodiment of the present invention. Detailed Implementation

[0053] The present invention will now be further described with reference to specific embodiments and the accompanying drawings.

[0054] Figure 1 This diagram illustrates the structural framework of a visual analysis system for astronomical spectral type inspection provided in an embodiment of the present invention. Figure 2The system's workflow diagram is shown. The system mainly includes a data preprocessing module 100, a selection module 200, an inspection module 300, and a promotion module 400. First, the user uses the data preprocessing module 100 to preprocess the reference spectrum and the spectrum to be inspected, completing spectral normalization, spectral feature detection, and spectral characterization. Then, the selection module 200 selects a spectrum to be inspected and determines the possible redshift and important spectral lines under that redshift. Next, the inspection module 300 determines the redshift and type of the spectrum to be inspected. Finally, the promotion module 400 obtains other spectra similar to the inspected spectrum. The functions and working methods of each module are described in detail below.

[0055] Data preprocessing module 100: Used to preprocess spectral data, completing spectral normalization, spectral feature detection, and spectral characterization. The reference spectrum and the spectrum to be examined serve as inputs to the system. The user extracts spectral features through the data preprocessing module 100 and calculates the spectral representation based on these features. The data preprocessing module 100 includes a normalization processing submodule 101, a spectral feature detection submodule 102, and a spectral characterization submodule 103, wherein:

[0056] Normalization processing submodule 101: This module normalizes the spectral data to be inspected, adjusting the flow range of each spectrum to [0,1]. LAMOST lacks components for obtaining corresponding photometric measurement data; therefore, the obtained spectral data cannot be calibrated for absolute flow rate and can only be calibrated for relative flow rate. After calibration, the flow range of each spectrum differs significantly. To ensure comparability between spectra and facilitate subsequent processing, the user first adjusts the flow range of each spectrum to [0,1] using the minimum-maximum normalization method provided by the normalization processing submodule.

[0057] Spectral Feature Detection Submodule 102: This submodule detects normalized spectral features and records their type, wavelength range, flow range, and peak value. Spectral features refer to the wavelengths corresponding to local maximum or minimum flow rates in the spectrum; these may correspond to emission or absorption lines, respectively. The spectral feature detection submodule uses topographic peak value to detect local maximum flow rates. By vertically flipping the spectrum, topographic peak value can also be used to detect local minimum flow rates. For each detected feature, the spectral feature detection submodule records its type (local maximum or minimum), wavelength range, flow range, and peak value.

[0058] Spectral Characterization Submodule 103: Used to calculate the representation of the spectrum based on spectral characteristics. This system obtains the spectral representation based on spectral characteristics. The system uses two types of spectra: a reference spectrum and a spectrum to be inspected. The reference spectrum has already been inspected by experts and can be used as a benchmark for subsequent inspections; the spectrum to be inspected is the spectrum that needs to be inspected by this system.

[0059] The main difference between characterizing the reference spectrum and the spectrum to be examined lies in the redshift. This invention directly uses the given redshift to characterize the reference spectrum. For the spectrum to be examined, since neither automated 1D pipelines nor independent pipelines may provide accurate redshifts, the spectral characterization submodule 103 of this invention improves the reliability of the representation by selecting n most likely redshift candidates. The spectral characterization submodule 103 computes a representation for each candidate and combines them together as the final spectral representation.

[0060] Selection module 200: Primarily used to assist expert users in selecting a spectrum to be examined and identifying possible redshifts and important spectral lines at those redshifts for further in-depth examination. This typically occurs at the beginning or after the expert user has examined similar spectra of an already examined spectrum. Selection module 200 primarily completes the selection based on the following two types of information:

[0061] 1) Relationship between reference spectrum and spectrum under test. If the representation of a spectrum under test is located in a region where the representation of a certain type of reference spectrum is concentrated, then the spectrum under test may also belong to this type.

[0062] 2) The type of the spectrum to be examined is determined by the 1D pipeline and the independent pipeline. If both types are the same, the spectrum is more likely to be correctly classified. Otherwise, its classification may be incorrect and requires careful examination.

[0063] Figure 3 The diagram illustrates the interface of the selection module provided in a specific embodiment of the present invention, which mainly includes the following four views:

[0064] Projection View (a): The projection view shows the relationship between the reference spectrum and the spectrum to be examined, with the reference spectrum as the background and the spectrum to be examined as the focus. This invention uses t-SNE to first project the reference spectrum, and then projects the spectrum to be examined. This invention uses OOS extended technology bi-kernel t-SNE to implement a layout based on the reference spectrum to project the spectrum to be examined. The projected distribution of the reference spectrum is visualized as a contour plot, and the projection of the spectrum to be examined is displayed as a scatter plot.

[0065] Parallel coordinate view (b): The parallel coordinate view displays the parameters of the spectrum under examination, namely the category and redshift given by the 1D pipeline and the independent pipeline.

[0066] Table View (c): The table view displays information for each spectrum to be examined, including the spectrum, the category and redshift given by the 1D pipeline and independent pipelines, and the redshift and type determined by experts. The table view provides expert users with a basic understanding of these spectra and helps users review and schedule the examination process. The table view uses a color-based representation, or color fields, to visualize the spectrum. Color fields are a one-dimensional heatmap that encodes aggregate values ​​through the saturation or brightness of vertical stripes. The table view further uses Event Striping to highlight outliers (i.e., features) by plotting them as additional vertical stripes and overlaying them onto the original stripes.

[0067] Candidate Spectral View (d): The candidate spectral view displays n candidate redshifts for the selected spectrum to be examined and the significant spectral lines for each candidate redshift. Additionally, for each candidate redshift, the candidate spectral view provides two reference spectra that are most similar to its representation.

[0068] Inspection module 300: mainly used to determine the redshift and type of the spectrum to be inspected. Figure 4 The diagram illustrates the interface of the inspection module provided in a specific embodiment of the present invention. The interface view is divided into three main categories based on function: adjustment submodule 301, recommendation submodule 302, and verification submodule 303.

[0069] Adjust submodule 301 corresponding Figure 4 The horizon plot (d) and local flow view (f) are primarily used for redshift adjustment. The main idea is to transform the wavelength-flow relationship for each spectral line into a redshift-flow relationship. Specifically, the wavelength range corresponding to the redshift range of each spectral line is calculated using the formula λ = (1 + z)λ0. Then, the wavelength-flow points within this range are converted into redshift-flow points (λ, z, and λ0 represent the current wavelength, spectral redshift, and rest wavelength of the spectral line, respectively). The redshift range is set according to the 1D pipeline. Flow rates at the boundaries do not have corresponding wavelengths and are obtained through interpolation. This conversion eliminates the differences in spectral wavelength changes during redshift and allows for consistent horizontal shift, thus enabling synchronous adjustment of redshift across multiple spectral lines.

[0070] The horizon plot (d) displays the redshift-flux relationship, with each plot corresponding to one spectral line. It uses spatial segmentation techniques to allow for a aligned view of all selected spectral lines. By segmenting and layering the original redshift-flux line plots, the horizon plot effectively compresses the vertical screen space used. The local flux view (f) displays the flux of each spectral line across 10 wavelengths around the current redshift, which can be used for fine-tuning the redshift.

[0071] Recommended submodule 302 correspondence Figure 4The spectral view (a) and overall importance view (g) are used to recommend suitable spectral lines to the user. Spectral lines are not always available; if a selected spectral line does not work, the user needs to select another. Recommendation submodule 302 provides the following two recommendation strategies:

[0072] 1) Users select significant features in the spectrum, and the system automatically recommends redshifts for the selected spectral lines. The spectral view (a) displays the wavelength-flux relationship of the spectrum. Expert users can select several significant features in the spectrum, and the system will automatically identify the spectral lines. The set of spectral lines that meet the requirements is marked in the spectral view. After selection, the redshift will be automatically adjusted so that experts can easily check the results.

[0073] 2) Displays the overall importance of all spectral lines at different redshifts. Redshifts with higher overall importance may correspond to the true redshift. The overall importance view (g) shows the overall importance of all spectral lines at different redshifts. Each band represents a redshift, and the height represents the overall importance. It is obtained by traversing the redshift range and then calculating the sum of the importance. When an expert selects a band, the redshift is adjusted, and the five most important spectral lines are selected.

[0074] Verify submodule 303 corresponding Figure 4 The spectral view (a), spectral line view (b), overall importance view (g), and local spectral view (e) are used to evaluate the inspection results. Correspondingly, the verification submodule 303 provides the following four verification methods:

[0075] 1) Check the matching degree between the current spectrum and the template spectrum: Expert users can display the template spectrum in the spectrum view (a) to compare the matching degree between the current spectrum and the template spectrum;

[0076] 2) Check the degree of matching between spectral features and spectral lines: The opacity of each spectral line in the spectral line view (b) varies depending on its degree of matching with the features. If the opacity of all selected spectral lines is high at the current redshift, then the expert's decision is reasonable;

[0077] 3) Check the overall importance of the selected spectral lines: The overall importance view (g) can be switched to display only the overall importance of the selected spectral lines. If the redshift is within a very high band, it indicates that the correct redshift may have been selected;

[0078] 4) Examine the local spectral profile of each spectral line to determine whether the redshift corresponds to the selected, false, or real but misplaced spectral line: The local spectral view (e) displays the local spectral profile of each spectral line, which expert users can use to determine whether the redshift corresponds to the selected, false, or real but misplaced spectral line.

[0079] Promotion Module 400: Used to find other spectra similar to the examined spectra through searching. After a user examines a spectrum, the system extracts the user's judgment criteria, namely the selected spectral lines and their importance, to search for other similar spectra. The search uses vectorized judgment criteria as the target pattern. It covers all spectra to be examined. For each spectrum, the user judges whether it is similar to the corresponding vector at the previously calculated n candidate redshifts. The user extracts an importance vector composed of the selected spectral lines from the representation of each redshift; then compares each importance value between this vector and the target pattern. This is the first filtering strategy, designed to remove vectors that are significantly different on a single spectral line. Afterwards, the user calculates the similarity between each retained vector and the target vector. The user sets a similarity threshold, filtering out vectors that do not meet the criteria again. If multiple final vectors correspond to the same spectrum, only the one with the highest similarity is retained.

[0080] Figure 5 A schematic diagram of the promotion module provided in a specific embodiment of the present invention is shown. The promotion interface is updated from the selection interface. In the projection view (a), the projection of the reference spectrum as the background remains unchanged, and the foreground displays the projection generated by the spectrum being examined and its similar spectra. Experts can perceive the similarity between the spectrum being examined and the similar spectra by their proximity, and determine the possible categories of the similar spectra by their positions on the background contour map. The spectral candidate view is updated to the similar spectrum view (b), which contains two parts: the upper part displays the spectrum being examined and the determined type and redshift; the lower part displays the similar spectra and their quantity, type statistics, and examination progress. Since the number of similar spectra may be large, the system uses pagination instead of scrolling to display them to avoid excessive rendering overhead and ensure smooth interface operation.

[0081] In this invention, the user uses a traditional keyboard and mouse for input, the computing module uses an Intel Xeon dual-core processor with 8GB of memory, and the display module uses a 23-inch 1920×1080 pixel display driven by an NVIDIA Quadro2000 graphics card.

[0082] refer to Figure 6 As shown, the visual analysis method for checking astronomical spectral types using the above system includes the following steps:

[0083] S1. Preprocess the spectral data to complete the normalization of the spectrum, detection of spectral features, and spectral characterization;

[0084] S2. Select a spectrum to be examined and determine the possible redshift and the important spectral lines at that redshift;

[0085] S3. Examine the spectrum to determine the redshift and its type;

[0086] S4. Find other spectra similar to the examined spectra by searching.

[0087] This invention provides a visual analysis system and method for astronomical spectral type inspection. The user first uses a data preprocessing module to preprocess the reference spectrum and the spectrum to be inspected, completing spectral normalization, spectral feature detection, and spectral characterization. Then, a selection module selects a spectrum to be inspected and determines possible redshifts and important spectral lines at those redshifts. Next, an inspection module determines the redshift and type of the spectrum to be inspected. Finally, a generalization module obtains other spectra similar to the inspected spectrum. Based on a "selection-inspection-generalization" workflow and various introduced visual guidelines, this system significantly improves the efficiency of spectral inspection while maintaining accuracy comparable to current visual inspection tools.

[0088] The above embodiments are merely illustrative examples of the present invention. The present invention may also be implemented in other specific ways or forms without departing from its spirit or essential characteristics. Therefore, the described embodiments should be considered illustrative rather than limiting in any respect. The scope of the present invention should be defined by the appended claims, and any variations equivalent to the intent and scope of the claims should also be included within the scope of the present invention.

Claims

1. A visual analytics system for astronomical spectral type inspection, characterized by, The system comprises: A data preprocessing module for preprocessing reference spectra and spectra to be checked, the data preprocessing module comprising a normalization processing submodule, a spectral feature detection submodule and a spectral characterization submodule for completing normalization of spectra, spectral feature detection and spectral characterization respectively; the normalization processing submodule normalizes the spectral data to be checked, specifically adjusting the flow range of each spectrum to [0, 1] by using a minimum-maximum normalization method; the spectral feature detection submodule records spectral types, wavelength ranges, flow ranges and protrusion values when performing spectral feature detection, detects local maximum flow in the spectrum by using terrain protrusion, and detects local minimum flow in the spectrum by vertically flipping the spectrum; the spectral characterization submodule uses a given redshift to characterize the reference spectrum, and for the spectrum to be checked, selects n most likely redshift candidates, calculates a representation for each redshift candidate, and combines them together as the final spectral representation, thereby improving the expression credibility; A selection module for helping a user select a spectrum to be checked and determine possible redshifts and important spectral lines under the redshift; the selection module completes selection based on the following two types of information: 1) Relationship between reference spectra and spectra to be checked: if the representation of a spectrum to be checked is in the region where the representations of reference spectra of a certain type are distributed, then the spectrum to be checked is likely to belong to this type; 2) Type of the spectrum to be checked determined by the 1D pipeline and the independent pipeline: if the two types are the same, the spectrum is more likely to be correctly classified; otherwise, the classification may be incorrect; The interface of the selection module comprises: A projection view for showing the relationship between reference spectra and spectra to be checked, taking reference spectra as the background and the spectrum to be checked as the focus; A parallel coordinate view for displaying parameters of the spectrum to be checked, i.e. the categories and redshifts given by the 1D pipeline and the independent pipeline; A table view for displaying information of each spectrum to be checked, including the spectrum, the categories and redshifts given by the 1D pipeline and the independent pipeline, and the redshift and type determined by experts; A spectrum candidate view for displaying n candidate redshifts of the selected spectrum to be checked, important spectral lines of each candidate redshift, and the two most similar reference spectra; An inspection module for helping a user determine the redshift and type of the spectrum to be checked, the inspection module comprising an adjustment submodule, a recommendation submodule and a verification submodule for adjusting the redshift of the spectrum to be checked, recommending appropriate spectral lines to the user and evaluating the inspection result respectively; A promotion module for helping a user search for other spectra similar to the checked spectrum; the screening strategy of the promotion module is: After the user checks a spectrum, the system extracts the user's judgment criteria, i.e. the selected spectral lines and their importance; the vectorized judgment criteria are used as the target pattern to search all spectra to be checked. The user judges whether each spectrum is similar to the corresponding vector at the previously calculated n candidate redshifts, i.e. the user extracts an importance vector consisting of selected spectral lines from the representation of each candidate redshift; then compares each importance value between the vector and the target pattern, thereby removing vectors with significant differences in a single spectral line; Then, the similarity between each retained vector and the target vector is calculated, and the user sets a similarity threshold to filter out vectors with a similarity lower than the similarity threshold again; if multiple final vectors correspond to the same spectrum, only the vector with the highest similarity is retained.

2. The visual analytics system for astronomical spectral type inspection according to claim 1, characterized in that, The interface of the checking module mainly includes: a spectrum view, a spectral line view, a horizon plot, a local spectrum view, a local flux view and an overall importance view, wherein: The adjusting submodule includes the horizon plot and the local flux view, the horizon plot converts the wavelength-flux relationship of each spectral line into a redshift-flux relationship, specifically: According to the formula λ=(1+z)λ0, the wavelength range corresponding to the redshift range of each spectral line is calculated, and then the wavelength-flux points in the range are converted into redshift-flux points, wherein: λ, z, λ0 respectively represent the current wavelength of the spectral line, the spectral redshift and the static wavelength of the spectral line; The local flux view is used to display the flux of each spectral line at multiple wavelengths around the current redshift, which is used for fine tuning of the redshift.

3. The visual analytics system for astronomical spectral type inspection according to claim 2, characterized in that, The recommendation submodule includes the spectrum view and the overall importance view, and the recommendation strategies are respectively: The spectrum view displays the wavelength-flux relationship of the spectrum, and the user selects the significant features in the spectrum in the spectrum view, and the system automatically recommends the redshift of the spectral line corresponding to the selected features; The overall importance view displays the overall importance of all spectral lines under different redshifts, and the redshift with high overall importance may correspond to the true redshift.

4. The visual analytics system for astronomical spectral type inspection according to claim 3, characterized in that, The verification submodule includes the spectrum view, the spectral line view, the overall importance view and the local spectrum view, and the corresponding provided verification methods are respectively: The template spectrum is displayed in the spectrum view, and the matching degree between the current spectrum and the template spectrum is checked in the spectrum view; The opacity of each spectral line in the spectral line view changes with the matching degree of the feature, and the matching degree of the spectral feature and the spectral line is checked in the spectral line view; The overall importance of the selected spectral line is checked in the overall importance view; The local spectrum profile of each spectral line is checked in the local spectrum view, and whether the redshift corresponds to the selected, false or true but position error spectral line is determined accordingly.

5. The visual analytics system for astronomical spectral type inspection of claim 1, wherein, The interface of the promotion module is updated from the interface of the selection module, and the update mainly includes: Projection view: taking the projection of the reference spectrum as the background and the projection of the checked spectrum and its similar spectra as the foreground, the user perceives the similarity between the checked spectrum and the similar spectrum through the closeness of the two, and determines the possible category of the similar spectrum through the position of the two on the background contour map; Similar spectrum view: instead of the original spectrum candidate view, the similar spectrum view includes an upper half and a lower half, the upper half displays the checked spectrum and the determined type and redshift, and the lower half displays information including the similar spectrum, the type statistics and the checking progress.

6. A visual analysis method for astronomical spectral type inspection based on the visual analysis system for astronomical spectral type inspection according to any one of claims 1-5, comprising the following steps: S1, preprocessing the spectral data, completing normalization of the spectrum, spectral feature detection and spectral characterization; S2, selecting a spectrum to be inspected, and determining possible redshift and important spectral lines under the redshift; S3, inspecting the spectrum to determine the redshift and type of the spectrum; S4, searching for other spectra similar to the inspected spectrum.

Images