End-to-end pulsar candidate body identification method and device, and computer equipment

Abstract

The invention relates to an end-to-end pulsar candidate body recognition method and device and computer equipment. The method comprises the steps of obtaining an original pulsar candidate image corresponding to a pulsar candidate body signal, and performing enhancement processing on the original pulsar candidate image to obtain a pulsar candidate image; marking a sub-map in the pulsar candidate image, and constructing a target data set for identifying the pulsar signal type; constructing a positioning data set by marking the position information of the sub-map in the original pulsar candidate image; training a preset pulsar recognition model according to the positioning data set to obtain a positioning model for recognizing sub-graph position information; and obtaining a preset reasoning prompt, and training the positioning model according to the preset reasoning prompt and the target data set to obtain a pulsar recognition model for recognizing the pulsar signal type. By adopting the method, the pulsar candidate body recognition accuracy can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence image processing technology, and in particular to an end-to-end pulsar candidate identification method, apparatus, and computer equipment. Background Art

[0002] Pulsars are neutron stars that emit highly periodic pulses of radiation. By identifying and utilizing the positions and signals of pulsars, it is possible to establish high-precision cosmic navigation systems, helping spacecraft determine their position and navigation during deep space exploration, and improving navigation accuracy.

[0003] With the rapid development of deep learning technology, pulsar search techniques are also continuously improving. Applying deep learning to the identification of pulsar candidates can achieve efficient and automated screening of pulsar candidates. However, related techniques, which rely on extracting local features from candidate images for identification, cannot accurately identify pulsar candidates. Summary of the Invention

[0004] Based on this, it is necessary to provide an end-to-end pulsar candidate identification method, apparatus, computer equipment, computer-readable storage medium and computer program product that can improve the accuracy of end-to-end pulsar candidate identification in response to the above technical problems.

[0005] In a first aspect, the present application provides an end-to-end pulsar candidate identification method, comprising:

[0006] Acquiring an original pulsar candidate image corresponding to a pulsar candidate signal, and performing enhancement processing on the original pulsar candidate image to obtain a pulsar candidate image;

[0007] Annotating subimages in the pulsar candidate image to construct a target data set for identifying pulsar signal types;

[0008] Constructing a positioning data set by marking position information of sub-images in the original pulsar candidate image;

[0009] Training a preset pulsar recognition model according to the positioning data set to obtain a positioning model for identifying sub-image position information;

[0010] A preset reasoning prompt is obtained, and the positioning model is trained according to the preset reasoning prompt and the target data set to obtain a pulsar recognition model for identifying the type of pulsar signal.

[0011] In one embodiment, performing enhancement processing on the pulsar candidate image to obtain the pulsar candidate image includes:

[0012] Identifying a sub-image frequency phase image in the original pulsar candidate image;

[0013] Data enhancement processing is performed on the sub-image frequency phase image to obtain a pulsar candidate image.

[0014] In one embodiment, labeling sub-images in the original pulsar candidate image to construct a target dataset for identifying pulsar signal types includes:

[0015] Determining a sub-image frequency phase diagram and a sub-image dispersion curve diagram in the original pulsar candidate image;

[0016] Identifying a first morphological feature of the sub-image frequency-phase image, and determining first labeling information of the sub-image frequency-phase image according to an identification result of the first morphological feature;

[0017] identifying a second morphological feature of the sub-image dispersion curve graph, and determining second annotation information of the sub-image dispersion curve graph according to a recognition result of the second morphological feature and the pulsar candidate image;

[0018] A target data set for identifying pulsar signal types is constructed based on the first labeling information, the second labeling information, and the corresponding pulsar candidate images.

[0019] In one embodiment, the identifying a first morphological feature of the sub-image frequency-phase image and determining first labeling information of the sub-image frequency-phase image based on the identification result of the first morphological feature includes:

[0020] Identifying a first morphological feature of the sub-image frequency-phase image, and if a target region of a preset shape exists in the identification result of the first morphological feature, determining a phase difference corresponding to an intersection point between the target region and the horizontal and vertical coordinates of the sub-image frequency-phase image;

[0021] If the phase difference is within a preset range, determining that the first annotation information of the sub-image frequency phase diagram is a first preset label, where the first preset label is used to indicate that the sub-image frequency phase diagram satisfies the characteristics of a real pulsar signal frequency phase diagram;

[0022] If the phase difference is not within the preset range, the first annotation information is determined to be a second preset label, and the second preset label is used to indicate that the sub-image frequency phase diagram does not meet the characteristics of the real pulsar signal frequency phase diagram.

[0023] In one embodiment, identifying a second morphological feature of the sub-image dispersion curve graph, and determining second annotation information of the sub-image dispersion curve graph based on the identification result of the second morphological feature and the pulsar candidate image, includes:

[0024] identifying a second morphological feature of the sub-image dispersion curve graph; and if a preset change trend exists in the recognition result of the second morphological feature and the number of peaks of the preset change trend is a first preset value, and the target dispersion value in the candidate pulsar image is not a second preset value, determining that the second annotation information of the sub-image dispersion curve graph is a third preset label; the third preset label is used to indicate that the dispersion curve graph satisfies the characteristics of a real pulsar signal;

[0025] If there is no preset change trend in the recognition result of the second morphological feature and / or the target dispersion value is the second preset value, the second annotation information of the sub-graph dispersion curve diagram is determined to be a fourth preset label; the fourth preset label is used to characterize that the dispersion curve diagram characteristics of the real pulsar signal are not met.

[0026] In one embodiment, constructing a positioning dataset by annotating position information of sub-images in the original pulsar candidate image includes:

[0027] Marking first position information of a sub-image in the original pulsar candidate image;

[0028] performing a scale transformation process on the original pulsar candidate image to generate a plurality of resized images with different resolutions;

[0029] determining a scale ratio between each of the resized images and the original pulsar candidate image, performing mapping processing on the first position information according to the scale ratio, and determining second position information of a subimage in each of the resized images;

[0030] A positioning data set is constructed according to the first position information and the second position information.

[0031] In one embodiment, the preset reasoning prompt includes:

[0032] If the first preset label and the third preset label exist in the pulsar candidate image, determining that the pulsar candidate signal corresponding to the original pulsar candidate image is a pulsar signal;

[0033] If the first preset label and the third preset label do not exist simultaneously in the pulsar candidate image, it is determined that the pulsar candidate signal is radio frequency interference.

[0034] In one embodiment, the method further comprises:

[0035] Acquire a target pulsar candidate image corresponding to the signal of the pulsar candidate to be identified;

[0036] The image of the candidate pulsar to be identified is input into the pulsar identification model, and the signal type of the signal of the candidate pulsar to be identified is output.

[0037] In a second aspect, the present application further provides an end-to-end pulsar candidate identification device, comprising:

[0038] An image processing module is used to obtain an original pulsar candidate image corresponding to the pulsar candidate signal, and perform enhancement processing on the original pulsar candidate image to obtain a pulsar candidate image;

[0039] A first data set construction module is used to annotate sub-images in the pulsar candidate image to construct a target data set for identifying pulsar signal types;

[0040] A second data set construction module is configured to construct a positioning data set by marking position information of sub-images in the original pulsar candidate image;

[0041] A first training module is used to train a preset pulsar recognition model according to the positioning data set to obtain a positioning model for identifying sub-image position information;

[0042] The second training module is used to obtain preset reasoning prompts, train the positioning model according to the preset reasoning prompts and the target data set, and obtain a pulsar recognition model for identifying the type of pulsar signals.

[0043] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the methods described above when executing the computer program.

[0044] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of any of the above methods when executed by a processor.

[0045] In a fifth aspect, the present application also provides a computer program product, comprising a computer program, which implements the steps of any of the methods described above when executed by a processor.

[0046] The above-mentioned end-to-end pulsar candidate identification method, apparatus, computer device, computer-readable storage medium, and computer program product enhance the entire original pulsar candidate image and annotate sub-images based on the enhanced pulsar candidate image to obtain a target dataset. This approach balances the dataset, increases its diversity, and precisely annotates sub-images within the image, improving the model's ability to identify pulsars. Furthermore, the method annotates the positional information of sub-images within the entire original pulsar candidate image to obtain a positioning dataset. The positioning dataset is first used to train a preset pulsar recognition model, enabling the model to accurately identify and locate the sub-image region, improving the model's ability to locate the sub-image. The positioning model is then trained based on preset inference hints and the target dataset, enabling the model to more effectively learn pulsar characteristics and guiding the model to output recognition results based on image features. Compared to the information fragmentation problem caused by the image slicing method in related technologies, this approach further improves the comprehensiveness and accuracy of recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments of the present application or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying any creative work.

[0048] Figure 1 A diagram illustrating an application environment of an end-to-end pulsar candidate identification method according to an embodiment;

[0049] Figure 2 1 is a flow chart of an end-to-end pulsar candidate identification method according to an embodiment;

[0050] Figure 3 is a schematic diagram of an original pulsar candidate image in one embodiment;

[0051] Figure 4 204 is a flow chart of step 204 in one embodiment;

[0052] Figure 5 A schematic diagram of a flow chart of a method for constructing a positioning dataset in one embodiment;

[0053] Figure 6 is a schematic diagram of a target pulsar candidate image in one embodiment;

[0054] Figure 7 is a schematic flow chart of an end-to-end pulsar candidate identification method according to another embodiment;

[0055] Figure 8is a structural block diagram of an end-to-end pulsar candidate identification device in one embodiment;

[0056] Figure 9 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0058] With the rapid development of deep learning technology, researchers have begun exploring the use of deep learning methods to directly extract identifying features from images of pulsar candidates. For example, methods such as PICS (Pulsar Image-based Classification System), PICS-ResNet, and CCNN (Concat Convolutional Neural Network) all use convolutional neural networks to automatically extract local features from candidate images, enabling efficient and automated screening of pulsar candidates. These studies demonstrate the enormous potential of image pattern recognition technology for processing astronomical big data.

[0059] However, traditional CNN (Convolutional Neural Network)-based recognition methods rely primarily on local receptive fields, making it difficult to fully model the global context within an image. This can lead to an incomplete understanding of the overall structure of pulsar signals. These methods typically use sub-images as input, focusing solely on visual features while ignoring equally important numerical features. This, in turn, results in low accuracy for sliced-image models.

[0060] Based on this, we propose an end-to-end intelligent pulsar candidate identification method that integrates visual reasoning with a multimodal large language model. This approach integrates multimodal information, such as sub-images and numerical data, and avoids the loss of relevance caused by data segmentation. It should be noted that the end-to-end identification method can be understood as directly identifying pulsar candidate images as a whole, rather than segmenting the image into multiple sub-images for separate analysis.

[0061] The end-to-end pulsar candidate identification method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. Obtain the original pulsar candidate image corresponding to the pulsar candidate signal, enhance the original pulsar candidate image to obtain the pulsar candidate image, mark the sub-image in the pulsar candidate image, and construct a target data set for identifying the type of pulsar signal. By marking the position information of the sub-image in the original pulsar candidate image, a positioning data set is constructed. According to the positioning data set, a preset pulsar recognition model is trained to obtain a positioning model for identifying the sub-image position information; obtain a preset reasoning hint, train the positioning model according to the preset reasoning hint and the target data set, and obtain a pulsar recognition model for identifying the type of pulsar signal.

[0062] The terminal 102 may be, but is not limited to, various personal computers, laptops, smartphones, tablet computers, IoT devices, and portable wearable devices. The server 104 may be an independent physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server providing cloud computing services.

[0063] In an exemplary embodiment, Figure 2 As shown in the figure, an end-to-end pulsar candidate identification method is provided, which is applied to Figure 1 The terminal in FIG is taken as an example to illustrate the method, including the following steps 202 to 210. Among them:

[0064] Step 202: Acquire the original pulsar candidate image corresponding to the pulsar candidate signal, perform enhancement processing on the original pulsar candidate image, and obtain the pulsar candidate image.

[0065] The image acquisition method can be determined by existing methods and will not be described in detail here. The original pulsar candidate image includes multiple modal information, for example, it can include a dispersion curve diagram, a cumulative contour diagram, a frequency phase diagram, a time phase diagram, and numerical information on the original pulsar candidate image. The numerical information on the original pulsar candidate image can be, for example, the optimal dispersion value. Figure 3 As shown in the figure, the original pulsar candidate image includes the dispersion curve diagram, cumulative profile diagram, frequency phase diagram, time phase diagram and the best dispersion value. The dispersion curve diagram, cumulative profile diagram, frequency phase diagram and time phase diagram are sub-images of the original pulsar candidate image, so the corresponding ones can also be called sub-image dispersion curve diagram, sub-image cumulative profile diagram, sub-image frequency phase diagram and sub-image time phase diagram respectively. Among them, the horizontal axis "DM" in the dispersion curve diagram represents the dispersion value, and the unit is pc / cm 3 ; Vertical axis " " represents the chi-square value of dimensionality reduction; "2PulsesofBestProfile" in the cumulative profile graph represents the 2 pulses of the best profile; "Sub-band" in the frequency phase graph represents the sub-band, "Phase" represents the phase, and "Frequency" represents the frequency, in MHz; "Phase" in the time phase graph represents the phase, and "Time" represents the time. It should be noted that the specific meanings of other sub-graphs can be understood by referring to the relevant technologies in this field and are not listed here one by one.

[0066] It is understandable that image enhancement processing technology has relevant applications in image processing. However, the image enhancement processing in this embodiment takes into account the scarcity of real pulsar signals and the serious imbalance between positive and negative samples. Therefore, the original pulsar candidate image is enhanced as a whole or locally (for example, the frequency phase map is data enhanced) to better generalize to the signal position changes that may occur in the real observation data, thereby increasing the diversity of the data and improving the generalization ability of the model.

[0067] Image enhancement methods include, but are not limited to, one or more of translation, flipping, rotation, and contrast transformation. Translation involves shifting the image horizontally (on the phase axis) by a number of pixels. When translating to the right, the blank area on the left is filled with the pixel values ​​on the right. When translating to the left, the blank area on the right is filled with the pixel values ​​on the left. Flipping involves mirroring the image horizontally or vertically. Rotation involves rotating the image 180 degrees clockwise or counterclockwise around its center point. Contrast transformation involves adjusting the image contrast to brighter or darker areas using gamma transformation or histogram equalization. Optionally, training the model requires constructing a corresponding dataset, in which case multiple raw pulsar candidate images are obtained. For ease of understanding, this example uses a single raw pulsar candidate image or pulsar candidate image for illustration.

[0068] Exemplarily, a pulsar candidate dataset is obtained from a server. The dataset includes a training set Dtrain and a test set Dtest. Enhancement processing is performed on the original pulsar candidate image corresponding to each pulsar candidate signal in the training set Dtrain to obtain a pulsar candidate image. Based on these multiple pulsar candidate images, new training samples Dtraind required for the pulsar recognition model are constructed.

[0069] Step 204 : annotate the sub-images in the pulsar candidate images to construct a target data set for identifying pulsar signal types.

[0070] Image annotation involves adding labels to image data so that the model can understand the image's content. The goal of annotation is to convert image data into structured data, enabling the model to learn the relationship between images and labels. Labels can include categories, attributes, locations, and other information. For example, in image classification tasks, each image is assigned a category label; in object detection tasks, the location and category of the object are annotated.

[0071] Optionally, the annotation of subimages within the candidate pulsar images can include annotation of pulsar features, such as morphological characteristics. Morphological characteristics can include pulse period, pulse profile, and curve variation trend. The target dataset can also be considered a labeled dataset, which can be used for model training and performance evaluation. Annotation can be performed using, but is not limited to, one or more of custom annotation tools, image annotation tools, or web-based annotation tools. This implementation can be accomplished using existing methods and is not detailed here.

[0072] It is understandable that the labeled sub-graphs may be sub-graphs whose morphological distinctions meet preset conditions, and the preset conditions are determined according to actual needs or historical data.

[0073] Step 206 : constructing a positioning data set by annotating the position information of the sub-images in the original pulsar candidate image.

[0074] For example, the location information of sub-images in the original pulsar candidate image is annotated to obtain the coordinates of the sub-images in the original pulsar candidate image. The original pulsar candidate image is then rescaled to generate multiple images with different resolutions. For each rescaled image, the new coordinates of the sub-image are calculated and mapped based on its size ratio with the original image. This creates corresponding annotation information for the image sub-images of different sizes, resulting in the corresponding positioning dataset.

[0075] Step 208 : training a preset pulsar recognition model based on the positioning data set to obtain a positioning model for identifying sub-image position information.

[0076] The preset pulsar recognition model may be a base model, for example, a multimodal large language model. Training the preset pulsar recognition model may be fine-tuning the preset pulsar recognition model.

[0077] For example, using the multimodal large language model QWEN2.5-VL as a base model, a pre-set pulsar recognition model is trained based on a positioning dataset, learning the position information of sub-images within pulsar images. When the pre-set training termination conditions are met, a positioning model for identifying sub-image positions is obtained. Furthermore, verification can be performed based on the constructed positioning dataset to ensure the correctness and consistency of the sub-image coordinate positions obtained through verification of images of different sizes, thereby ensuring the reliability and practicality of the recognition results.

[0078] The multimodal large language model utilizes the Transformer architecture and can process multiple data modalities. It is suitable for tasks that require the integration of multiple data. For example, it can maintain high recognition performance under various observation conditions and data with different distributions, demonstrating generalization capabilities that are superior to traditional deep learning models. It has strong adaptability and is suitable for actual pulsar search tasks.

[0079] Step 210: Obtain a preset reasoning hint, train a positioning model according to the preset reasoning hint and the target data set, and obtain a pulsar recognition model for identifying the type of pulsar signal.

[0080] The training process utilizes a "thinking-like" reasoning approach to guide the model's output of recognition results based on image features. Preset reasoning prompts can be determined based on actual needs and serve to guide the positioning model's reasoning, rather than directly providing answers. A preset reasoning prompt might be that if the morphological features of the sub-images in a candidate pulsar image all meet those of a true pulsar signal, and the numerical information in the candidate pulsar image associated with the sub-image is not a preset value, then the signal corresponding to the candidate pulsar image is identified as a pulsar signal; otherwise, it is identified as radio frequency interference. Training the pulsar recognition model can involve fine-tuning the weights of the positioning model.

[0081] Exemplarily, a preset reasoning prompt is obtained. During the training process of a positioning model based on a target dataset, the model is guided to output recognition results based on image features using a "thinking" reasoning approach based on the preset reasoning prompt. When the preset conditions for completing positioning model training are met, a pulsar recognition model for identifying pulsar signal types is obtained. Training can include fine-tuning the weights of the positioning model to generate the pulsar recognition model. Furthermore, after training is complete, the pulsar recognition model is validated against a test dataset, and the output recognition results are analyzed to verify the accuracy and reliability of the pulsar recognition model. For example, the correct classification of pulsar signals and radio frequency interference is verified.

[0082] In this end-to-end pulsar candidate identification method, the entire original pulsar candidate image is enhanced and sub-images are annotated based on the enhanced pulsar candidate image to obtain a target dataset. This approach balances the dataset, increases its diversity, and precisely annotates sub-images within the image, improving the model's ability to identify pulsars. A localization dataset is obtained by annotating the positional information of sub-images within the entire original pulsar candidate image. This localization dataset is first used to train a preset pulsar recognition model, enabling the model to accurately identify and locate the sub-image region, improving the model's ability to locate the sub-image. The localization model is then trained based on preset inference cues and the target dataset, enabling the model to more effectively learn pulsar characteristics and guide the model to output recognition results based on image features. This approach further improves the comprehensiveness and accuracy of recognition compared to the information fragmentation problem caused by the image slicing method used in related technologies.

[0083] There are multiple sub-images in the original pulsar candidate image, but the morphological distinctions of different sub-images are different. Therefore, when performing image enhancement processing, only part of the image can be enhanced, and sub-images can be annotated only those with morphological distinctions that meet preset conditions.

[0084] In an exemplary embodiment, enhancing a pulsar candidate image to obtain the pulsar candidate image includes: identifying a sub-image frequency phase map in the original pulsar candidate image; and performing data enhancement processing on the sub-image frequency phase map to obtain the pulsar candidate image.

[0085] For example, the frequency phase map of a sub-image in the original pulsar candidate image is identified; operations such as translation, flipping, rotation, and contrast transformation are performed on the sub-image frequency phase map to obtain the pulsar candidate image. It is understandable that when the multimodal large language model QWEN2.5-VL is preset as the base model, the base model's ability to recognize dispersion curve maps meets practical requirements and can be fine-tuned based on a small amount of data. However, its ability to recognize frequency phase maps does not meet practical requirements, as the sub-image frequency phase maps in the original pulsar candidate image need to be enhanced to increase the amount of data.

[0086] This method amplifies real pulsar signals and generates new training samples based on the enhanced original pulsar candidate images. This augments the frequency-phase images to better generalize to signal position variations that may occur in real observational data. This increases data diversity and improves the model's generalization capabilities.

[0087] In an exemplary embodiment, Figure 4 As shown, step 204 includes steps 402 to 408. Among them:

[0088] Step 402: Determine the sub-image frequency phase diagram and the sub-image dispersion curve diagram in the original pulsar candidate image.

[0089] Step 404 : Identify a first morphological feature of the sub-image frequency-phase image, and determine first labeling information of the sub-image frequency-phase image according to the identification result of the first morphological feature.

[0090] Among them, the first morphological feature of identifying the sub-image frequency phase map can be to identify the image feature of the sub-image frequency phase map, whether there is a target area of ​​a preset shape, and determine the phase difference corresponding to the intersection of the target area and the horizontal and vertical coordinates of the sub-image frequency phase map.

[0091] Optionally, identifying a first morphological feature of the sub-image frequency-phase image, and determining first labeling information of the sub-image frequency-phase image according to an identification result of the first morphological feature, includes:

[0092] Identify the first morphological feature of the sub-image frequency phase diagram. If there is a target area of ​​a preset shape in the recognition result of the first morphological feature, determine the phase difference corresponding to the intersection of the target area and the horizontal and vertical coordinates of the sub-image frequency phase diagram; if the phase difference is within a preset range, determine the first annotation information of the sub-image frequency phase diagram as a first preset label, and the first preset label is used to characterize that the sub-image frequency phase diagram meets the characteristics of the real pulsar signal frequency phase diagram; if the phase difference is not within the preset range, determine the first annotation information as a second preset label, and the second preset label is used to characterize that the sub-image frequency phase diagram does not meet the characteristics of the real pulsar signal frequency phase diagram.

[0093] Frequency-phase diagrams display pulsar signals superimposed at different observation frequencies, typically presenting the relationship between frequency and phase in a two-dimensional format, with color depth representing signal amplitude. Real pulsars typically emit broadband signals, which appear as continuous bright vertical streaks (black vertical lines) across all frequency subbands in subband diagrams. The pre-defined target region can represent a broadband signal, for example, a black vertical line or a vertical band-like structure.

[0094] Among them, by judging whether the frequency phase diagram meets the characteristics of the real pulsar signal frequency phase diagram, it is divided into two categories: "yes" and "no", that is, the first preset label and the second preset label.

[0095] For example, the first morphological feature of the sub-image frequency-phase graph is identified. If a black vertical line is detected in the sub-image frequency-phase graph, and the phase difference between the vertical line and the upper and lower intersection points of the horizontal axis phase axis is within a preset range, i.e., no more than 0.4, the sub-image is classified as "yes", i.e., the first preset label. Otherwise, the sub-image frequency-phase graph is classified as "no", i.e., the second preset label.

[0096] Step 406 : Identify a second morphological feature of the sub-image dispersion curve graph, and determine second annotation information of the sub-image dispersion curve graph based on the identification result of the second morphological feature and the pulsar candidate image.

[0097] The dispersion curve represents the integral of the electron density from Earth to the pulsar and is used to determine the optimal dispersion measurement. Since real pulsar signals originate from outer space, their dispersion curves typically peak at non-zero dispersion values. The upward trend on the left side of the peak and the downward trend on the right side may indicate that the electron density of the pulsar signal first increases and then decreases as it travels through the interstellar medium. This characteristic is an important basis for verifying pulsar candidates. The second annotation information can be the degree of match between the sub-graph dispersion curve and the dispersion curve characteristics of the real pulsar signal. The identification results of the second morphological feature can include the trend of change.

[0098] Optionally, a second morphological feature of the sub-image dispersion curve diagram is identified, and second annotation information of the sub-image dispersion curve diagram is determined based on the recognition result of the second morphological feature and the pulsar candidate image, including: identifying the second morphological feature of the sub-image dispersion curve diagram, if there is a preset change trend in the recognition result of the second morphological feature and the number of peaks in the preset change trend is a first preset value, and the target dispersion value in the pulsar candidate image is not a second preset value, then determining that the second annotation information of the sub-image dispersion curve diagram is a third preset label; the third preset label is used to characterize that the characteristics of the dispersion curve diagram of a real pulsar signal are met; if there is no preset change trend and / or the target dispersion value is the second preset value in the recognition result of the second morphological feature, then determining that the second annotation information of the sub-image dispersion curve diagram is a fourth preset label; the fourth preset label is used to characterize that the characteristics of the dispersion curve diagram of a real pulsar signal are not met.

[0099] It is understandable that a monotonic increase could be due to a gradual increase in the electron density as the signal propagates. A monotonic decrease could be due to a gradual decrease in the electron density as the signal propagates, which is inconsistent with the characteristics of pulsar signals, as pulsar signals generally do not vary monotonically with frequency and peak at non-zero dispersion values.

[0100] The preset variation trend can be a dispersion curve graph with a relatively distinct peak, an upward trend on the left side of the peak, and a downward trend on the right side. The absence of the preset variation trend in the recognition result of the second morphological feature can be an overall monotonically increasing trend, a monotonically decreasing trend, or frequent fluctuations in the dispersion curve, which corresponds to the presence of zero or more relatively distinct peaks. The target dispersion value can be the optimal dispersion value of the pulsar candidate in the original pulsar candidate image. The first preset value can be, but is not limited to, 1, and the second preset value can be, but is not limited to, 0. In this embodiment, the first preset value is 1 and the second preset value is 0 for illustration.

[0101] For example, if the dispersion curve has a relatively obvious peak, with an upward trend on the left side of the peak and a downward trend on the right side; and combined with whether the best dispersion value of the pulsar candidate in the image is non-"0", if it is non-"0", it will be classified as "yes", that is, the third preset label.

[0102] If the dispersion curve shows an overall monotonically increasing or decreasing trend, or if the dispersion curve fluctuates frequently with zero or multiple relatively obvious peaks, the optimal dispersion value of the pulsar candidate in the image is "0." If any of these conditions are met, the candidate is classified as "No," which is the fourth preset label.

[0103] Step 408 : Construct a target data set for identifying pulsar signal types based on the first annotation information, the second annotation information, and the corresponding pulsar candidate images.

[0104] In this example, by meticulously annotating subimages within candidate pulsar images and constructing a target dataset, the model can be trained using this dataset to improve the model's ability to identify pulsars. This effectively integrates image information and data features, avoiding the information fragmentation associated with traditional image slicing methods and further enhancing the comprehensiveness and accuracy of recognition.

[0105] In an exemplary embodiment, a method for constructing a positioning dataset is provided, such as Figure 5 As shown, the following steps are included:

[0106] Step 502: annotate the first position information of the sub-image in the original pulsar candidate image.

[0107] The first position information can be understood as a coordinate position. The sub-graphs include a frequency phase graph and a dispersion curve graph.

[0108] Exemplarily, the coordinate positions of the frequency phase diagram and the dispersion curve diagram in the original pulsar candidate image are accurately marked manually to obtain the first position information corresponding to each of the frequency phase diagram and the dispersion curve diagram.

[0109] Step 504 : performing a scale transformation process on the original pulsar candidate image to generate a plurality of resized images with different resolutions.

[0110] The scale transformation process may be to scale the original pulsar candidate image according to a preset size ratio to obtain a plurality of resized images with different resolutions.

[0111] Step 506 : Determine the scale ratio between each resized image and the original pulsar candidate image, perform mapping processing on the first position information according to the scale ratio, and determine the second position information of the sub-image in each resized image.

[0112] Step 508: Construct a positioning data set based on the first position information and the second position information.

[0113] In the above embodiment, by constructing a positioning dataset containing images of various sizes and corresponding sub-image coordinates, the model can be adapted to the recognition of sub-image positions in images of different sizes, thereby enhancing the model's robustness to size changes.

[0114] In an exemplary embodiment, the preset inference prompt includes that if the morphological features corresponding to the first preset label and the third preset label exist in the pulsar candidate image, the pulsar candidate signal corresponding to the original pulsar candidate image is identified as a pulsar signal; if the morphological features corresponding to the first preset label and the third preset label do not exist in the pulsar candidate image at the same time, it is determined that the pulsar candidate signal is identified as radio frequency interference.

[0115] In an exemplary embodiment, the application based on the above-mentioned pulsar identification model can be to obtain the target pulsar candidate image corresponding to the signal of the pulsar candidate to be identified; input the pulsar candidate image to be identified into the pulsar identification model, and output the signal type of the pulsar candidate signal to be identified.

[0116] For example, the target pulsar candidate image corresponding to the pulsar candidate signal to be identified is obtained. First, a determination is made as to whether a black vertical line exists in the sub-image frequency phase diagram of the target pulsar candidate image, and whether the phase difference between the vertical line and the upper and lower intersections of the horizontal phase axis does not exceed 0.4. Secondly, a determination is made as to whether the sub-image dispersion curve has a relatively distinct peak, with an upward trend on the left side and a downward trend on the right side. Furthermore, the determination is made as to whether the best dispersion value of the target pulsar candidate image is non-zero. A non-zero DM value is a key characteristic of a pulsar signal. If a black vertical line exists, the phase difference does not exceed 0.4, the sub-image dispersion curve has a relatively distinct peak, with an upward trend on the left side and a downward trend on the right side, and the best dispersion value of the target pulsar candidate image is non-zero, then the signal is identified as a pulsar signal. Otherwise, it is identified as radio frequency interference.

[0117] For example, enter Figure 6 The target pulsar candidate image shown will output the following identification results:

[0118] 1. Examine the third subplot, which is a frequency-phase plot. For a real pulsar signal, you should see black vertical lines or bands running through the frequency axis. I see black vertical lines or bands in the subplot, which are characteristic of pulsar signals.

[0119] 2. Examine the fourth sub-graph, which shows a dispersion curve. Since pulsars are non-Earth signals, we expect the signal to peak at a non-zero DM value (i.e., a non-zero dispersion value). I found a clear peak, and the optimal dispersion value is non-zero, which is characteristic of pulsar signals. <Thinking>

[0120] <Answer>The signal is a pulsar signal. <Answer>

[0121] It should be noted that Figure 6 The target pulsar candidate image shown includes a frequency phase diagram, a dispersion curve diagram, and an optimal dispersion value. Figure 6 The graph type shown is the same as Figure 3 The diagrams shown are of the same type. The meaning of the parameters in the diagrams can be referred to Figure 3 It can be understood with the relevant technology in this field and will not be described here.

[0122] Based on the above model, in the case of actual model recognition, the candidate pulsar image to be identified can be directly input into the model output result without the need for labeling, segmentation and enhancement. The model will give the reasoning process, including the features of the sub-image, and the final result.

[0123] In an exemplary embodiment, Figure 7 As shown in the figure, an end-to-end pulsar candidate identification method is provided, which is applied to Figure 1 The terminal in FIG is taken as an example to illustrate the process, including the following steps 702 to 714. Among them:

[0124] Step 702: Acquire the original pulsar candidate image corresponding to the pulsar candidate signal, perform enhancement processing on the original pulsar candidate image, and obtain the pulsar candidate image.

[0125] Step 704 : annotate the sub-images in the pulsar candidate image to construct a target data set for identifying pulsar signal types.

[0126] Step 706 : constructing a positioning data set by annotating the position information of the sub-images in the original pulsar candidate image.

[0127] Step 708: Train the preset pulsar recognition model according to the positioning data set to obtain a positioning model for identifying the sub-image position information.

[0128] Step 710: Obtain a preset reasoning hint, train a positioning model according to the preset reasoning hint and the target data set, and obtain a pulsar recognition model for identifying the type of pulsar signal.

[0129] Step 712: Acquire the target pulsar candidate image corresponding to the signal of the pulsar candidate to be identified.

[0130] Step 714: Input the image of the candidate pulsar to be identified into the pulsar identification model, and output the signal type of the signal of the candidate pulsar to be identified.

[0131] It should be noted that the specific implementation of this embodiment can be achieved through the above-mentioned limited methods, which will not be elaborated here.

[0132] In the above embodiment, by enhancing the original pulsar candidate image, balancing the data set, and finely annotating the sub-images in the image, a high-quality training data set is constructed, which enables the model to more effectively learn the characteristics of pulsars and improve recognition accuracy and positioning precision; during the training process, a reasoning mechanism that simulates the thinking of scientists is introduced, so that the model can deeply analyze image features in steps and perform reasoning, greatly improving the ability to understand complex image patterns, thereby improving the accuracy and reliability of the recognition results; utilizing the multimodal large language model modeling capability of the Transformer architecture, effectively integrating image information and data features, avoiding the information fragmentation problem caused by traditional image slicing methods, and further improving the comprehensiveness and accuracy of recognition. In addition, the large model using the Transformer architecture has better learning ability and can maintain high recognition performance under various observation conditions and data with different distributions. It shows a generalization ability that is superior to traditional deep learning models, has strong adaptability, and is suitable for actual pulsar search tasks.

[0133] It should be noted that the frequency phase diagrams and dispersion curve diagrams in the original pulsar candidate image and the corresponding enhanced pulsar candidate image can be labeled identically, and the optimal dispersion values ​​in the original pulsar candidate image and the corresponding enhanced pulsar candidate image can also be consistent. In the above embodiments, the optimal dispersion value can be obtained from the original pulsar candidate image or from the corresponding enhanced pulsar candidate image. It should be understood that although the steps in the flowcharts of the above embodiments are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Furthermore, at least some of the steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages do not necessarily need to be executed at the same time, but can be executed at different times. These steps or stages do not necessarily need to be executed sequentially, but can be executed in rotation or alternation with other steps or at least a portion of steps or stages in other steps.

[0134] Based on the same inventive concept, embodiments of the present application also provide an end-to-end pulsar candidate identification device for implementing the aforementioned end-to-end pulsar candidate identification method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more of the following end-to-end pulsar candidate identification device embodiments can be found in the above-mentioned limitations of the end-to-end pulsar candidate identification method and will not be further elaborated here.

[0135] In an exemplary embodiment, Figure 8 As shown, an end-to-end pulsar candidate identification device is provided, comprising: an image processing module 802, a first data set construction module 804, a second data set construction module 806, a first training module 808 and a second training module 810, wherein:

[0136] The image processing module 802 is used to obtain the original pulsar candidate image corresponding to the pulsar candidate signal, and perform enhancement processing on the original pulsar candidate image to obtain the pulsar candidate image.

[0137] The first data set construction module 804 is used to label sub-images in the pulsar candidate image and construct a target data set for identifying the type of pulsar signal.

[0138] The second data set construction module 806 is configured to construct a positioning data set by annotating position information of sub-images in the original pulsar candidate image.

[0139] The first training module 808 is used to train a preset pulsar recognition model according to the positioning data set to obtain a positioning model for identifying sub-image position information.

[0140] The second training module 810 is used to obtain preset reasoning hints, train the positioning model according to the preset reasoning hints and the target data set, and obtain a pulsar recognition model for identifying the type of pulsar signals.

[0141] The above-mentioned end-to-end pulsar candidate identification device enhances the entire original pulsar candidate image and annotates sub-images based on the enhanced pulsar candidate image to obtain a target dataset. This method balances the dataset, increases its diversity, and finely annotates sub-images within the image, improving the model's ability to recognize pulsars. A positioning dataset is obtained by annotating the position information of sub-images within the entire original pulsar candidate image. The positioning dataset is first used to train a preset pulsar recognition model, enabling the model to accurately identify and frame the sub-image area, improving the model's ability to locate the sub-image. The positioning model is then trained based on preset inference prompts and the target dataset, enabling the model to more effectively learn the characteristics of pulsars and guiding the model to output recognition results based on image features. Compared to the information fragmentation problem caused by the image slicing method in related technologies, this method further improves the comprehensiveness and accuracy of recognition.

[0142] In an exemplary embodiment, the image processing module 802 is configured to identify a sub-image frequency phase image in an original pulsar candidate image; and perform data enhancement processing on the sub-image frequency phase image to obtain a pulsar candidate image.

[0143] In an exemplary embodiment, the first data set construction module 804 is configured to determine a sub-image frequency phase map and a sub-image dispersion curve map in the original pulsar candidate image;

[0144] Identifying a first morphological feature of the sub-image frequency-phase image, and determining first labeling information of the sub-image frequency-phase image based on an identification result of the first morphological feature;

[0145] identifying a second morphological feature of the sub-image dispersion curve graph, and determining second annotation information of the sub-image dispersion curve graph based on the identification result of the second morphological feature and the pulsar candidate image;

[0146] A target data set for identifying pulsar signal types is constructed based on the first labeling information, the second labeling information, and the corresponding pulsar candidate images.

[0147] In an exemplary embodiment, the first data set construction module 804 is configured to identify a first morphological feature of the sub-image frequency phase image, and if a target region of a preset shape exists in the recognition result of the first morphological feature, determine a phase difference corresponding to an intersection point between the target region and the horizontal and vertical coordinates of the sub-image frequency phase image;

[0148] If the phase difference is within a preset range, determining the first annotation information of the sub-image frequency phase diagram as a first preset label, where the first preset label is used to indicate that the sub-image frequency phase diagram satisfies the characteristics of a real pulsar signal frequency phase diagram;

[0149] If the phase difference is not within the preset range, the first annotation information is determined to be a second preset label, and the second preset label is used to indicate that the sub-image frequency phase diagram does not meet the characteristics of the real pulsar signal frequency phase diagram.

[0150] In an exemplary embodiment, the first data set construction module 804 is configured to identify a second morphological feature of the sub-image dispersion curve graph. If a preset change trend exists in the identification result of the second morphological feature and the number of the preset change trends is a first preset value, and the target dispersion value in the original pulsar candidate image is not the second preset value, then determine that the second annotation information of the sub-image dispersion curve graph is a third preset label; the third preset label is used to indicate that the dispersion curve graph satisfies the characteristics of a real pulsar signal.

[0151] If there is no preset change trend in the recognition result of the second morphological feature, and / or the target dispersion value is the second preset value, then the second annotation information of the sub-graph dispersion curve diagram is determined to be the fourth preset label; the fourth preset label is used to characterize that the dispersion curve diagram characteristics of the real pulsar signal are not met.

[0152] In an exemplary embodiment, the second data set construction module 806 is configured to annotate first position information of a sub-image in the original pulsar candidate image;

[0153] Performing a scale transformation on the original pulsar candidate image to generate multiple resized images with different resolutions;

[0154] determining a scale ratio between each resized image and the original pulsar candidate image, performing mapping processing on the first position information according to the scale ratio, and determining second position information of the sub-image in each resized image;

[0155] A positioning data set is constructed according to the first position information and the second position information.

[0156] In an exemplary embodiment, the preset reasoning prompts include:

[0157] If the morphological features corresponding to the first preset label and the third preset label exist in the pulsar candidate image, the pulsar candidate signal corresponding to the original pulsar candidate image is identified as a pulsar signal;

[0158] If the morphological features corresponding to the first preset label and the third preset label do not exist in the pulsar candidate image at the same time, it is determined that the pulsar candidate signal is identified as radio frequency interference.

[0159] In an exemplary embodiment, the above-mentioned end-to-end pulsar candidate identification apparatus further includes an identification module, the identification module being configured to obtain a target pulsar candidate image corresponding to a signal of the pulsar candidate to be identified;

[0160] The image of the candidate pulsar to be identified is input into the pulsar identification model, and the signal type of the signal of the candidate pulsar to be identified is output.

[0161] Each module in the above-mentioned end-to-end pulsar candidate identification apparatus can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in the form of software in a memory in the computer device, so that the processor can call and execute the corresponding operations of each module.

[0162] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 9 As shown. The computer device includes a processor, memory, an input / output interface, a communication interface, a display unit, and an input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with external terminals via wired or wireless means. The wireless means can be implemented via Wi-Fi, mobile cellular networks, near-field communication (NFC), or other technologies. When executed by the processor, the computer program implements an end-to-end pulsar candidate identification method. The display unit of the computer device is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device casing, or an external keyboard, touchpad or mouse.

[0163] Those skilled in the art will understand that Figure 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0164] In one embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0165] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0166] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0167] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0168] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the various embodiments provided herein may be, but are not limited to, general-purpose processors, central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), programmable logic devices (PLDs), quantum computing-based data processing logic devices, artificial intelligence (AI) processors, and the like.

[0169] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0170] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. An end-to-end pulsar candidate identification method, characterized in that: The method comprises: Acquiring an original pulsar candidate image corresponding to a pulsar candidate signal, and performing enhancement processing on the original pulsar candidate image to obtain a pulsar candidate image; Annotating subimages in the pulsar candidate image to construct a target data set for identifying pulsar signal types; Constructing a positioning data set by marking position information of sub-images in the original pulsar candidate image; Training a preset pulsar recognition model according to the positioning data set to obtain a positioning model for identifying sub-image position information; A preset reasoning prompt is obtained, and the positioning model is trained according to the preset reasoning prompt and the target data set to obtain a pulsar recognition model for identifying the type of pulsar signal.

2. The method according to claim 1, characterized in that Performing enhancement processing on the pulsar candidate image to obtain the pulsar candidate image includes: Identifying a sub-image frequency phase image in the original pulsar candidate image; Data enhancement processing is performed on the sub-image frequency phase image to obtain a pulsar candidate image.

3. The method according to claim 1, characterized in that The step of labeling sub-images in the original pulsar candidate image to construct a target data set for identifying pulsar signal types includes: Determining a sub-image frequency phase diagram and a sub-image dispersion curve diagram in the original pulsar candidate image; Identifying a first morphological feature of the sub-image frequency-phase image, and determining first labeling information of the sub-image frequency-phase image according to an identification result of the first morphological feature; identifying a second morphological feature of the sub-image dispersion curve graph, and determining second annotation information of the sub-image dispersion curve graph according to an identification result of the second morphological feature and the pulsar candidate image; A target data set for identifying pulsar signal types is constructed based on the first labeling information, the second labeling information, and the corresponding pulsar candidate images.

4. The method according to claim 3, characterized in that The identifying a first morphological feature of the sub-image frequency-phase image and determining first labeling information of the sub-image frequency-phase image according to an identification result of the first morphological feature includes: Identifying a first morphological feature of the sub-image frequency-phase image, and if a target region of a preset shape exists in the identification result of the first morphological feature, determining a phase difference corresponding to an intersection point between the target region and the horizontal and vertical coordinates of the sub-image frequency-phase image; If the phase difference is within a preset range, determining that the first annotation information of the sub-image frequency phase diagram is a first preset label, where the first preset label is used to indicate that the sub-image frequency phase diagram satisfies the characteristics of a real pulsar signal frequency phase diagram; If the phase difference is not within the preset range, the first annotation information is determined to be a second preset label, and the second preset label is used to indicate that the sub-image frequency phase diagram does not meet the characteristics of the real pulsar signal frequency phase diagram.

5. The method according to claim 4, characterized in that The identifying a second morphological feature of the sub-image dispersion curve graph, and determining second labeling information of the sub-image dispersion curve graph according to the identification result of the second morphological feature and the pulsar candidate image, includes: identifying a second morphological feature of the sub-image dispersion curve graph; and if a preset change trend exists in the recognition result of the second morphological feature and the number of peaks in the preset change trend is a first preset value, and the target dispersion value in the pulsar candidate image is not a second preset value, determining that the second annotation information of the sub-image dispersion curve graph is a third preset label; the third preset label is used to indicate that the dispersion curve graph satisfies the characteristics of a real pulsar signal; If the preset change trend does not exist in the recognition result of the second morphological feature, and / or the target dispersion value is the second preset value, then the second annotation information of the sub-graph dispersion curve diagram is determined to be a fourth preset label; the fourth preset label is used to characterize that the dispersion curve diagram characteristics of the real pulsar signal are not met.

6. The method according to claim 1, wherein The step of constructing a positioning data set by marking position information of sub-images in the original pulsar candidate image includes: Marking first position information of a sub-image in the original pulsar candidate image; performing a scale transformation process on the original pulsar candidate image to generate a plurality of resized images with different resolutions; determining a scale ratio between each of the resized images and the original pulsar candidate image, performing mapping processing on the first position information according to the scale ratio, and determining second position information of a subimage in each of the resized images; A positioning data set is constructed according to the first position information and the second position information.

7. The method according to claim 5, characterized in that The preset reasoning prompts include: If the morphological features corresponding to the first preset label and the third preset label exist in the pulsar candidate image, the pulsar candidate signal corresponding to the original pulsar candidate image is identified as a pulsar signal; If the morphological features corresponding to the first preset label and the third preset label do not exist in the pulsar candidate image at the same time, it is determined that the pulsar candidate signal is identified as radio frequency interference.

8. The method according to any one of claims 1 to 7, characterized in that The method further comprises: Acquire a target pulsar candidate image corresponding to the signal of the pulsar candidate to be identified; The image of the candidate pulsar to be identified is input into the pulsar identification model, and the signal type of the signal of the candidate pulsar to be identified is output.

9. An end-to-end pulsar candidate identification device, characterized in that: The device comprises: An image processing module is used to obtain an original pulsar candidate image corresponding to the pulsar candidate signal, and perform enhancement processing on the original pulsar candidate image to obtain a pulsar candidate image; A first data set construction module is used to annotate sub-images in the pulsar candidate image to construct a target data set for identifying pulsar signal types; A second data set construction module is configured to construct a positioning data set by marking position information of sub-images in the original pulsar candidate image; A first training module is used to train a preset pulsar recognition model according to the positioning data set to obtain a positioning model for identifying sub-image position information; The second training module is used to obtain preset reasoning prompts, train the positioning model according to the preset reasoning prompts and the target data set, and obtain a pulsar recognition model for identifying the type of pulsar signals.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

12. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images