Visual analysis system and method for astronomical spectrum type inspection

Abstract

The present invention relates to the technical field of visualization and human-computer interaction, and in particular to a visual analysis system and method for astronomical spectrum type inspection. The system comprises: a data preprocessing module used for preprocessing reference spectra and spectra to be inspected, to complete spectrum normalization, spectrum feature detection and spectrum characterization; a selection module used for helping a user to select a spectrum to be inspected and determine a possible redshift and important spectral lines under the redshift; an inspection module used for helping the user to determine the redshift and type of the spectrum to be inspected, the inspection module comprising an adjustment sub-module, a recommendation sub-module and a verification sub-module, which are respectively used for adjusting the redshift of the spectrum to be inspected, recommending a proper spectral line to the user and evaluating an inspection result; and a promotion module used for helping the user to obtain other spectra similar to the inspected spectrum by means of searching. The system and method provided by the present invention can significantly improve the classification efficiency while ensuring the spectrum classification accuracy of experts.

Description

A visual analysis system and method for astronomical spectrum type inspection Technical Field

[0001] The present invention belongs to the technical field of visualization and human-computer interaction, and specifically relates to a visual analysis system and method for astronomical spectrum type inspection. Background Art

[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 critical step before using spectra for scientific research. Due to the enormous number of spectra, classification is primarily performed using automated methods. While automated methods are improving in performance, they place high demands on spectral quality and cannot guarantee completely accurate classification even when processing high-quality spectra. Therefore, visual inspection remains crucial to ensuring classification accuracy. However, current tools for visual inspection are relatively simple. They fail to provide experts with effective data organization and analysis strategies for processing large numbers of spectra, nor do they provide sufficient support for experts in examining individual spectra.

[0003] Summary of the Invention

[0004] To address the deficiencies in the prior art, the present invention aims to provide a visual analysis system and method for astronomical spectral type inspection. This system and method, based on a "select-inspect-promote" process and combined with various forms of visual guidance, support experts in conducting spectral inspections more efficiently while ensuring classification accuracy.

[0005] In order to achieve the above purpose, a technical solution adopted by the present invention is:

[0006] A visual analysis system for astronomical spectrum type inspection, characterized in that the system comprises:

[0007] Data preprocessing module: used to preprocess the reference spectrum and the spectrum to be checked. The data preprocessing module includes a normalization processing submodule, a spectrum feature detection submodule and a spectrum characterization submodule, which are respectively used to complete spectrum normalization, spectrum feature detection and spectrum characterization;

[0008] Selection module: used to help users select a spectrum to be examined and determine the possible redshift and important spectral lines under the redshift;

[0009] Inspection module: used to help 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 respectively used to adjust the redshift of the spectrum to be inspected, recommend appropriate spectral lines to the user, and evaluate the inspection results;

[0010] Promotion module: It is used to help users obtain other spectra similar to the checked spectrum through search.

[0011] Furthermore, in the visual analysis system for astronomical spectrum type inspection as described above, the normalization processing submodule normalizes the spectral data to be inspected by adjusting the flow range of each spectrum to [0, 1] using a minimum-maximum normalization method.

[0012] Furthermore, in the visual analysis system for astronomical spectrum type inspection as described above, the spectral feature detection submodule records the spectrum type, wavelength range, flow range and protrusion value when performing spectral feature detection, uses the terrain protrusion to detect the local maximum flow in the spectrum, and detects the local minimum flow in the spectrum by vertically flipping the spectrum and terrain protrusion.

[0013] Furthermore, in the visual analysis system for astronomical spectrum type inspection as described above, the spectrum characterization submodule uses a given redshift to characterize a reference spectrum; for the spectrum to be inspected, the spectrum characterization submodule selects n most likely redshift candidates, calculates a representation for each redshift candidate, and combines them together as the final spectrum representation, thereby improving the credibility of the representation.

[0014] Furthermore, in the visual analysis system for checking astronomical spectrum types as described above, the selection module performs selection based on the following two types of information:

[0015] 1) Relationship between reference spectra and the spectrum to be checked: If the representation of a spectrum to be checked is in an area where a certain type of reference spectrum is concentrated, then the spectrum to be checked may also belong to this type;

[0016] 2) The type of the spectrum to be checked is determined by the 1D pipeline and the independent pipeline: If the two types are the same, then the spectrum is likely to be correctly classified; otherwise, there may be an error in the classification.

[0017] Furthermore, in the visual analysis system for astronomical spectrum type inspection as described above, the interface of the selection module includes:

[0018] Projection view: used to show the relationship between the reference spectrum and the spectrum to be checked, with the reference spectrum as the background and the spectrum to be checked as the focus;

[0019] Parallel coordinates view: used to display the parameters of the spectrum to be examined, namely the category and redshift given by the 1D pipeline and the independent pipeline;

[0020] Table view: used to display information about each spectrum to be examined, including the category and redshift given by the spectrum, 1D pipeline and independent pipeline, as well as the redshift and type determined by the expert;

[0021] Spectral Candidate View: It is used to display the n candidate redshifts of the spectrum selected for inspection, the important spectral lines of each candidate redshift, and the two reference spectra that are most similar to it.

[0022] Furthermore, in the visual analysis system for astronomical spectrum type inspection as described above, the interface of the inspection module mainly includes: spectrum view, spectrum line view, horizon view, local spectrum view, local flow view and overall importance view, wherein:

[0023] The adjustment submodule includes the horizon diagram and the local flow view. The horizon diagram converts the wavelength-flow relationship of each spectral line into a redshift-flow relationship, specifically:

[0024] 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 point in this range is converted into a redshift-flux point, where: λ, z, λ0 represent the current wavelength of the spectral line, the spectral redshift, and the rest wavelength of the spectral line respectively;

[0025] 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 the redshift.

[0026] Furthermore, in the visual analysis system for checking astronomical spectrum types as described above, the recommendation submodule includes the spectral view and the overall importance view, and the recommendation strategies are:

[0027] The spectrum view displays the wavelength-flow relationship of the spectrum. The user can select a significant feature in the spectrum in the spectrum view, and the system will automatically recommend a spectral line with a redshift corresponding to the selected feature.

[0028] The overall importance view shows the overall importance of all spectral lines at different redshifts, and redshifts with higher overall importance may correspond to true redshifts.

[0029] Furthermore, in the visual analysis system for checking astronomical spectrum types as described above, the verification submodule includes the spectrum view, spectral line view, overall importance view, and local spectrum view, and the corresponding verification methods provided are:

[0030] Display the template spectrum in the spectrum view, and check the matching degree between the current spectrum and the template spectrum in the spectrum view;

[0031] The opacity of each spectral line in the spectral line view varies with its match to the feature. Check the match between the spectral feature and the spectral line in the spectral line view;

[0032] Check the overall importance of the selected line in the Overall Importance view;

[0033] Inspect the local spectral profile of each spectral line in the local spectral view to determine whether the redshift corresponds to a selected, spurious, or real but misplaced spectral line.

[0034] Furthermore, in the visual analysis system for checking astronomical spectrum types as described above, the screening strategy of the promotion module is:

[0035] After the user has checked a spectrum, the system extracts the user's judgment criteria, i.e., the selected spectral lines and their importance; uses the vectorized judgment criteria as a target pattern to search all spectra to be checked;

[0036] The user determines whether each spectrum is similar to the corresponding vectors at the n candidate redshifts previously calculated, i.e., the user extracts an importance vector consisting of the selected spectral lines from the representation of each candidate redshift; then compares each importance value between this vector and the target pattern, thereby removing vectors that have significant differences on a single spectral line;

[0037] Afterwards, the similarity between each retained vector and the target vector is calculated. The user sets a similarity threshold and again filters out vectors with similarity below the threshold. If multiple final vectors correspond to the same spectrum, only the vector with the highest similarity is retained.

[0038] Furthermore, in the visual analysis system for checking astronomical spectrum types as described above, the interface of the promotion module is updated by updating the interface of the selection module. The update mainly includes:

[0039] Projection view: With the projection of the reference spectrum as the background and the projections of the inspected spectrum and its similar spectra as the foreground, the user perceives the similarity between the inspected spectrum and the similar spectra by their proximity, and determines the possible categories of the similar spectra by their positions on the background contour map;

[0040] Similar Spectrum View: Replaces the original spectrum candidate view. The Similar Spectrum View includes an upper half and a lower half. The upper half displays the inspected spectrum and the determined type and redshift. The lower half displays information including similar spectra and their number, type statistics, and inspection progress.

[0041] A visual analysis method for astronomical spectrum type inspection using the visual analysis system for astronomical spectrum type inspection as described above comprises the following steps:

[0042] S1. Preprocess spectral data to complete spectral normalization, spectral feature detection and spectral characterization;

[0043] S2. Select a spectrum to be examined and determine the possible redshift and the important spectral lines at the redshift;

[0044] S3. Check the spectrum to determine the red shift and type of the spectrum;

[0045] S4. Other spectra similar to the examined spectrum are obtained by searching.

[0046] The visual analysis system and method for astronomical spectrum type inspection according to the present invention have the following significant technical effects:

[0047] The present 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. A selection module then selects a spectrum to be inspected and determines possible redshifts and important spectral lines associated with those redshifts. The inspection module then determines the redshift and type of the inspected spectrum. Finally, a generalization module generates additional spectra similar to the inspected spectrum. This system, based on a "select-inspect-generalize" workflow and the introduction of various visual guidance tools, significantly improves the efficiency of spectral inspection while maintaining comparable accuracy to current visual inspection tools. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] FIG1 is a structural framework diagram of a visual analysis system for astronomical spectrum type inspection provided in an embodiment of the present invention;

[0049] FIG2 is a flowchart of the workflow of the visual analysis system described in FIG1 ;

[0050] FIG3 is a schematic diagram of the interface for selecting a module in FIG1 ;

[0051] FIG4 is a schematic diagram of the interface of the inspection module in FIG1 ;

[0052] FIG5 is a schematic diagram of the interface of the promotion module in FIG1;

[0053] FIG6 is a flow chart of a visual analysis method for checking astronomical spectrum types provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The present invention will be further described below with reference to specific embodiments and the accompanying drawings.

[0055] Figure 1 shows the structural framework of a visual analysis system for checking astronomical spectra, provided in an embodiment of the present invention. Figure 2 shows the system's workflow. The system primarily comprises a data preprocessing module 100, a selection module 200, a checking module 300, and a promotion module 400. The user first uses the data preprocessing module 100 to preprocess the reference spectrum and the spectrum to be checked, completing spectral normalization, spectral feature detection, and spectral characterization. The user then uses the selection module 200 to select a spectrum to be checked and determine the possible redshift and important spectral lines associated with that redshift. The checking module 300 then determines the redshift and type of the spectrum to be checked. Finally, the promotion module 400 generates other spectra similar to the checked spectrum. The functions and operating methods of each module are described in detail below.

[0056] Data preprocessing module 100: This module preprocesses spectral data, performing spectral normalization, spectral feature detection, and spectral characterization. The system uses a reference spectrum and the spectrum to be examined as input. The user uses data preprocessing module 100 to extract spectral features and calculate a spectral representation based on these features. Data preprocessing module 100 includes a normalization submodule 101, a spectral feature detection submodule 102, and a spectral characterization submodule 103. The following are the steps:

[0057] Normalization submodule 101: This module normalizes the spectral data to be examined, adjusting the flow rate range of each spectrum to [0, 1]. LAMOST lacks a component for acquiring corresponding photometric data, so the resulting spectral data cannot be calibrated for absolute flow rates. Instead, relative flow rate calibration is used. After calibration, the flow rate ranges of each spectrum vary significantly. To ensure comparability between spectra and facilitate subsequent processing, the user first adjusts the flow rate range of each spectrum to [0, 1] using the minimum-maximum normalization method provided by the normalization submodule.

[0058] Spectral feature detection submodule 102: Detects normalized spectral features and records their type, wavelength range, flow rate range, and protrusion value. Spectral features refer to the wavelengths corresponding to the local maximum or minimum flow rate in the spectrum, which may correspond to the wavelengths of emission lines or absorption lines, respectively. The spectral feature detection submodule uses topographic protrusion to detect the local maximum flow rate. Topographic protrusion can also be used to detect the local minimum flow rate by vertically flipping the spectrum. For each detected feature, the spectral feature detection submodule records its type (local maximum or minimum), wavelength range, flow rate range, and protrusion value.

[0059] Spectral Characterization Submodule 103: Calculates a spectral representation based on spectral features. The system uses two types of spectra: reference spectra and spectra to be inspected. Reference spectra have been inspected by experts and serve as a benchmark for subsequent inspections; spectra to be inspected are those that require inspection by the system.

[0060] The primary difference between characterizing the reference spectrum and the spectrum to be inspected lies in the redshift. The present invention directly uses the given redshift to characterize the reference spectrum. For the spectrum to be inspected, since neither the automated 1D pipeline nor the independent pipelines may provide accurate redshifts, the spectral characterization submodule 103 of the present invention improves representation confidence by selecting the n most likely redshift candidates. Spectral characterization submodule 103 calculates a representation for each candidate and combines them into the final spectral representation.

[0061] Selection Module 200: This module helps the expert user select a spectrum to be examined and identifies possible redshifts and important spectral lines at those redshifts for subsequent in-depth examination. This typically occurs at the beginning or after the expert user has examined similar spectra to an already examined spectrum. Selection Module 200 primarily makes its selection based on the following two types of information:

[0062] 1) Relationship between reference spectra and the spectrum to be checked: If the representation of a spectrum to be checked is in an area where a certain type of reference spectrum is concentrated, then the spectrum to be checked may also belong to this type.

[0063] 2) The spectrum to be checked is classified by both the 1D pipeline and the independent pipeline. If the two types are the same, then the spectrum is likely correctly classified. Otherwise, its classification may be incorrect and requires careful review.

[0064] FIG3 shows a schematic diagram of the interface of the selection module provided in a specific embodiment of the present invention, which mainly includes the following four views:

[0065] Projection view (a): The projection view shows the relationship between the reference spectrum and the spectrum to be inspected, with the reference spectrum as the background and the spectrum to be inspected as the focus. This paper uses t-SNE to first project the reference spectrum and then the spectrum to be inspected. This paper uses the OOS extension technique bi-kernel t-SNE to project the spectrum to be inspected based on the layout of the reference spectrum. The projected distribution of the reference spectrum is visualized as a contour plot, while the projection of the spectrum to be inspected is displayed as a scatter plot.

[0066] Parallel Coordinates View (b): The Parallel Coordinates view displays the parameters of the spectrum to be examined, namely the class and redshift given by the 1D pipeline and the independent pipeline.

[0067] Table view (c): The table view displays information about each spectrum to be inspected, including the category and redshift given by the spectrum, 1D pipeline, and independent pipeline, as well as the redshift and type determined by the expert. The table view gives expert users a basic understanding of these spectra and helps users view and arrange the inspection process. The table view uses a color-based representation method, namely color fields, to visualize the spectra. Color fields are a one-dimensional heat map that encodes aggregate values ​​through the saturation or brightness of the vertical stripes. The table view further uses Event Striping to highlight outliers (i.e., features) by drawing them as additional vertical stripes and superimposing them on the original stripes.

[0068] Spectral Candidates View (d): The Spectral Candidates View displays the n candidate redshifts for the spectrum selected for inspection and the important spectral lines for each candidate redshift. In addition, for each candidate redshift, the Spectral Candidates View also provides the two reference spectra that are most similar to its representation.

[0069] Inspection module 300: Mainly used to determine the redshift and type of the spectrum to be inspected. Figure 4 shows a schematic diagram of the interface of the inspection module provided in a specific embodiment of the present invention. The interface view is divided into three categories according to function: adjustment submodule 301, recommendation submodule 302, and verification submodule 303.

[0070] The adjustment submodule 301 corresponds to the horizon diagram (d) and the local flow view (f) in Figure 4, and is mainly used to adjust the redshift. Its main idea is to convert the wavelength-flow relationship of each spectral line into a redshift-flow relationship. Specifically, the wavelength range corresponding to the redshift range of each spectral line is calculated according to the formula λ=(1+z)λ0, and then the wavelength-flow point within the range is converted into a redshift-flow point (λ, z, λ0 represent the current wavelength of the spectral line, the spectral redshift, and the stationary wavelength of the spectral line, respectively). The redshift range is set according to the 1D pipeline. The flow on the boundary has no corresponding wavelength and is obtained by interpolation. This conversion eliminates the difference in the wavelength change of the spectral line during the redshift change and allows consistent horizontal movement, so that the redshift can be adjusted synchronously on multiple spectral lines.

[0071] The Horizon plot (d) displays the redshift-flux relationship, with each plot corresponding to a single spectral line. It uses a spatial segmentation technique to allow for aligned viewing of all selected spectral lines. By segmenting and layering the original redshift-flux line plot, the Horizon plot effectively compresses the vertical screen space used. The Local Flux view (f) displays the flux of each spectral line at 10 wavelengths around the current redshift, allowing for fine-tuning of the redshift.

[0072] The recommendation submodule 302 corresponds to the spectral view (a) and the overall importance view (g) in Figure 4 and is used to recommend suitable spectral lines to the user. Spectral lines are not always available. If the selected spectral line does not work, the user needs to select another spectral line. The recommendation submodule 302 provides the following two recommendation strategies:

[0073] 1) The user selects significant features in the spectrum, and the system automatically recommends spectral lines with redshifts corresponding to the selected features. Spectral view (a) displays the wavelength-flux relationship of the spectrum. Expert users can select several significant features in the spectrum, and the system automatically identifies the spectral lines. Spectral lines that meet the requirements are marked in the spectral view. After selection, the redshift is automatically adjusted, allowing experts to easily check the results.

[0074] 2) Displays the overall importance of all spectral lines at different redshifts. Redshifts with higher overall importance are likely to correspond to true redshifts. The overall importance view (g) shows the overall importance of all spectral lines at different redshifts. Each band represents a redshift, and its height indicates the overall importance. This is calculated by summing the importance of all spectral lines across the redshift range. When an expert selects a band, the redshift is adjusted, and the five most important spectral lines are selected.

[0075] Verification submodule 303 corresponds to the spectral view (a), spectral line view (b), overall importance view (g), and local spectral view (e) in Figure 4 and is used to evaluate the inspection results. Accordingly, verification submodule 303 provides the following four verification methods:

[0076] 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;

[0077] 2) Check the match between the spectral feature and the spectral line: The opacity of each spectral line in the spectral line view (b) varies depending on its match with the feature. If the opacity of all selected spectral lines is high at the current redshift, then the expert's decision is reasonable;

[0078] 3) Check the overall importance of the selected line: The overall importance view (g) can be switched to show only the overall importance of the selected line. If the redshift is in a very high band, it indicates that they may have selected the correct redshift;

[0079] 4) Check the local spectral profile of each spectral line to determine whether the redshift corresponds to a selected, false, or real but misplaced spectral line: The local spectral view (e) displays the local spectral profile of each spectral line, from which expert users can determine whether the redshift corresponds to a selected, false, or real but misplaced spectral line.

[0080] Promotion module 400: used to obtain other spectra similar to the inspected spectrum through search. After the user has inspected a spectrum, the system will extract the user's judgment criteria, that is, the selected spectral lines and their importance, to search for other similar spectra. The search uses the vectorized judgment criteria as the target pattern. It covers all spectra to be inspected. For each spectrum, the user judges whether it is similar to the corresponding vectors at the n candidate redshifts previously calculated. The user extracts the importance vector composed of the selected spectral lines from the representation of each redshift; then compares each importance value between the vector and the target pattern. This is the first screening strategy, which aims to remove vectors that have significant differences in a single spectral line. Afterwards, the user calculates the similarity between each retained vector and the target vector. The user sets a similarity threshold to filter out those vectors that do not meet the conditions again. If multiple final vectors correspond to the same spectrum, only the one with the highest similarity is retained.

[0081] Figure 5 shows a schematic diagram of the interface of the promotion module provided in a specific embodiment of the present invention. The promotion interface is updated by 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 inspected spectrum and its similar spectra. Experts can perceive the degree of similarity between the inspected spectrum and the similar spectra by their proximity, and determine the possible categories of similar spectra by their positions on the background contour map. The spectrum candidate view is updated to the similar spectrum view (b), which consists of two parts. The upper part displays the inspected spectrum and the determined type and redshift; the lower part displays similar spectra and their number, type statistics, inspection progress and other information. Since the number of similar spectra may be large, the system uses paging instead of scrolling to display them to avoid excessive rendering overhead and ensure smooth interface operation.

[0082] In the present 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.

[0083] Referring to FIG6 , the visual analysis method for checking astronomical spectrum types using the above system includes the following steps:

[0084] S1. Preprocess spectral data to complete spectral normalization, spectral feature detection and spectral characterization;

[0085] S2. Select a spectrum to be examined and determine the possible redshift and the important spectral lines at the redshift;

[0086] S3. Check the spectrum to determine the red shift and type of the spectrum;

[0087] S4. Other spectra similar to the examined spectrum are obtained by searching.

[0088] This invention provides a visual analysis system and method for checking astronomical spectral types. First, the user uses a data preprocessing module to preprocess reference and target spectra, performing spectral normalization, spectral feature detection, and spectral characterization. The selection module then selects a spectrum to be checked and determines the possible redshift and important spectral lines at that redshift. The inspection module then determines the redshift and type of the target spectrum. Finally, the generalization module generates other spectra similar to the checked spectrum. This system, based on a "select-check-generalize" workflow and the introduction of various visual guides, significantly improves the efficiency of spectral inspection while maintaining comparable accuracy to current visual inspection tools.

[0089] The above embodiments are merely illustrative of the present invention, and the present invention may also be implemented in other specific ways or in other specific forms without departing from the gist or essential characteristics of the present invention. Therefore, the described embodiments are to be considered in all respects as illustrative and not restrictive. The scope of the present invention is to be determined by the appended claims, and any variations equivalent to the intent and scope of the claims are intended to be within the scope of the present invention.

Claims

1. A visual analysis system for astronomical spectrum type inspection, characterized in that: The system comprises: Data preprocessing module: used to preprocess the reference spectrum and the spectrum to be checked. The data preprocessing module includes a normalization processing submodule, a spectrum feature detection submodule and a spectrum characterization submodule, which are respectively used to complete spectrum normalization, spectrum feature detection and spectrum characterization; Selection module: used to help users select a spectrum to be examined and determine the possible redshift and important spectral lines under the redshift; Inspection module: used to help 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 respectively used to adjust the redshift of the spectrum to be inspected, recommend appropriate spectral lines to users and evaluate the inspection results; Promotion module: It is used to help users obtain other spectra similar to the examined spectrum through search.

2. The visual analysis system for astronomical spectrum type inspection according to claim 1, characterized in that: The normalization processing submodule normalizes the spectral data to be checked by: using the minimum-maximum normalization method to adjust the flow range of each spectrum to [0, 1].

3. The visual analysis system for astronomical spectrum type inspection according to claim 2, characterized in that: The spectral feature detection submodule records the spectrum type, wavelength range, flow range and protrusion value when performing spectral feature detection, uses the terrain protrusion to detect the local maximum flow in the spectrum, and detects the local minimum flow in the spectrum by vertically flipping the spectrum and the terrain protrusion.

4. The visual analysis system for astronomical spectrum type inspection according to claim 3, characterized in that: The spectral characterization submodule uses a given redshift to characterize a reference spectrum; for a spectrum to be checked, the spectral characterization submodule selects n most likely redshift candidates, calculates a representation for each redshift candidate, and combines them together as a final spectral representation, thereby improving the representation credibility.

5. The visual analysis system for astronomical spectrum type inspection according to any one of claims 1 to 4, characterized in that: The selection module completes the selection based on the following two types of information: 1) Relationship between reference spectra and the spectrum to be checked: If the representation of a spectrum to be checked is in an area where a certain type of reference spectrum representation is concentrated, then the spectrum to be checked may also belong to this type; 2) The type of the spectrum to be checked is determined by the 1D pipeline and the independent pipeline: if the two types are the same, then the spectrum is likely to be correctly classified; otherwise, the classification may be wrong.

6. The visual analysis system for astronomical spectrum type inspection according to claim 1, characterized in that: The interface of the selection module includes: Projection view: used to show the relationship between the reference spectrum and the spectrum to be checked, with the reference spectrum as the background and the spectrum to be checked as the focus; Parallel coordinates view: used to display the parameters of the spectrum to be examined, i.e. the category and redshift given by the 1D pipeline and the independent pipeline; Table view: used to display information about each spectrum to be checked, including the spectrum, the category and redshift given by the 1D pipeline and the independent pipeline, as well as the redshift and type determined by the expert; Spectral Candidate View: It is used to display n candidate redshifts of the selected spectrum to be checked, the important spectral lines of each candidate redshift, and the two reference spectra that are most similar to it.

7. The visual analysis system for astronomical spectrum type inspection according to claim 1, characterized in that: The interface of the inspection module mainly includes: spectrum view, spectrum line view, horizon view, local spectrum view, local flow view and overall importance view, among which: The adjustment submodule includes the horizon diagram and the local flow view. The horizon diagram converts the wavelength-flow relationship of each spectral line into the redshift-flow 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-flow point in the range is converted into the redshift-flow point, where: λ, z, λ0 represent the current wavelength of the spectral line, the spectral redshift and the static wavelength of the spectral line respectively; The local flow view is used to display the flow of each spectral line at multiple wavelengths around the current redshift, which is used for fine-tuning the redshift.

8. The visual analysis system for astronomical spectrum type inspection according to claim 7, characterized in that: The recommendation submodule includes the spectral view and the overall importance view, and the recommendation strategies are: The spectrum view displays the wavelength-flow relationship of the spectrum. The user selects a significant feature in the spectrum in the spectrum view, and the system automatically recommends a redshift of the spectral line corresponding to the selected feature. The overall importance view shows the overall importance of all spectral lines at different redshifts, and redshifts with higher overall importance may correspond to true redshifts.

9. The visual analysis system for astronomical spectrum type inspection according to claim 7, characterized in that: The verification submodule includes the spectrum view, spectral line view, overall importance view and local spectrum view, and the corresponding verification methods provided are: Display the template spectrum in the spectrum view, and check the matching degree between the current spectrum and the template spectrum in the spectrum view; The opacity of each spectral line in the spectral line view varies depending on how well it matches the feature. Check the match between the spectral feature and the spectral line in the spectral line view. Check the overall importance of the selected line in the Overall Importance view; In the Local Spectrum View, examine the local spectral profile of each line to determine whether the redshift corresponds to a selected, spurious, or real but misplaced line.

10. The visual analysis system for astronomical spectrum type inspection according to claim 1, characterized in that: The screening strategy of the promotion module is: After the user has checked a spectrum, the system extracts the user's judgment criteria, i.e., the selected spectral lines and their importance; uses the vectorized judgment criteria as the target mode to search all spectra to be checked; The user determines whether each spectrum is similar to the corresponding vectors at the previously calculated n candidate redshifts, i.e., the user extracts an importance vector consisting of the selected spectral lines from the representation of each candidate redshift; and then compares each importance value between the vector and the target pattern, thereby removing vectors with significant differences on a single spectral line; After that, the similarity between each retained vector and the target vector is calculated, and the user sets a similarity threshold, and again filters out vectors whose similarity is lower than the similarity threshold; if multiple final vectors correspond to the same For the same spectra, only the vector with the highest similarity is retained.

11. The visual analysis system for astronomical spectrum type inspection according to claim 10, characterized in that: The interface of the promotion module is updated by the interface of the selection module, and the update mainly includes: Projection view: With the projection of the reference spectrum as the background and the projection of the examined spectrum and its similar spectra as the foreground, the user perceives the similarity between the examined spectrum and the similar spectrum by the proximity of the two, and determines the possible category of the similar spectrum by the position of the two on the background contour map; Similar spectrum view: replaces the original spectrum candidate view, the similar spectrum view includes an upper part and a lower part, the upper part displays the inspected spectrum and the determined type and redshift, and the lower part displays information including similar spectra and their number, type statistics and inspection progress.

12. A visual analysis method for astronomical spectrum type inspection based on the visual analysis system for astronomical spectrum type inspection according to any one of claims 1 to 11, comprising the following steps: S1. Preprocess spectral data to complete spectral normalization, spectral feature detection and spectral characterization; S2, select a spectrum to be examined and determine the possible redshift and the important spectral lines under the redshift; S3, checking the spectrum to determine the red shift and type of the spectrum; S4. Other spectra similar to the examined spectrum are obtained by searching.

Images