CN116304624A - Quick radio storm searching method based on oblique line detection and curve fitting dispersion elimination - Google Patents
Quick radio storm searching method based on oblique line detection and curve fitting dispersion elimination Download PDFInfo
- Publication number
- CN116304624A CN116304624A CN202310212412.0A CN202310212412A CN116304624A CN 116304624 A CN116304624 A CN 116304624A CN 202310212412 A CN202310212412 A CN 202310212412A CN 116304624 A CN116304624 A CN 116304624A
- Authority
- CN
- China
- Prior art keywords
- frb
- result
- method based
- detection
- dispersion
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 40
- 239000006185 dispersion Substances 0.000 title claims abstract description 28
- 238000000034 method Methods 0.000 title claims abstract description 25
- 230000008030 elimination Effects 0.000 title claims abstract description 10
- 238000003379 elimination reaction Methods 0.000 title claims abstract description 10
- 230000011218 segmentation Effects 0.000 claims abstract description 27
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 20
- 238000000605 extraction Methods 0.000 claims abstract description 17
- 238000012937 correction Methods 0.000 claims abstract description 14
- 230000014509 gene expression Effects 0.000 claims description 34
- 238000010586 diagram Methods 0.000 claims description 14
- 238000005070 sampling Methods 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 6
- 230000000877 morphologic effect Effects 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 5
- 238000012935 Averaging Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 abstract description 2
- 238000007670 refining Methods 0.000 abstract 1
- 238000013135 deep learning Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 4
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000005855 radiation Effects 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000002360 explosive Substances 0.000 description 1
- 230000004907 flux Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Abstract
A fast radio storm searching method based on oblique line detection and curve fitting dispersion elimination comprises the following steps: denoising by multi-convolution kernel oblique line detection, semantic level segmentation by utilizing a U-shaped network, FRB detection, skeleton extraction, RANSAC fitting, primary correction and residual correction. The invention uses multi-convolution kernel oblique line detection to remove noise, enhances a rapid radio storm signal, uses a U-shaped network to divide FRB, and carries out FRB detection on the transverse and longitudinal relative lengths according to the division result. And refining the segmentation result by using a skeleton extraction algorithm, then fitting FRB by using a RANSAC algorithm, and finally carrying out primary correction and residual correction on the original image, and calculating physical parameters such as a dispersion value. The method can well strengthen FRB signals, inhibit strong noise signals, solve the problem that marked data sets are absent in astronomical data, and solve the problem of huge calculation power consumption in the achromatizing stage of the traditional algorithm.
Description
Technical Field
The invention relates to the field of radioastronomy, in particular to a quick radiostorm searching method based on oblique line detection and curve fitting dispersion elimination.
Background
Fast Radio Burst (FRB) is an explosive, broadband, highly dispersive, pulsed Radio radiation astronomical phenomenon of only a few milliseconds in duration, with instantaneous radiation flux up to tens of central sryl (Jy). In 2007, lorimer first discovered this astronomical phenomenon when analyzing historical data of the Australian Parkes astronomical station pulsar patrol. Up to now many properties of FRB remain unknown. Thus, a study based on a large number of FRB observations is needed to answer questions about its origin and emission mechanism. At present, more and more telescopes in the world begin to search for FRB, and domestic radio telescopes represented by FAST also develop FRB searching and observing, and produce important scientific results. With the development of FRB observation and research, the technical bottleneck of FRB search of massive observation data is gradually revealed: the disposable and transient characteristics of the FRB have high requirements on the real-time searching precision and speed of the FRB. Therefore, high-rate and high-precision FRB detection search techniques are critical to achieving selective real-time voltage data dumping.
Currently, these monopulse events are searched by automated, high performance software based on achromatic theory, such as HEIMDALL, FDMT, bonsai, presto. Such software is capable of meeting processing speed requirements for offline processing and lower data rates. Screening is performed manually due to Radio Frequency Interference (RFI), system gain variation, or other factors. However, as the amount of FRB observations increases, traditional manual screening approaches have been elusive. In recent years, due to the development of computer technology and GPU, deep learning has achieved excellent results in terms of signal classification, pattern recognition, etc. in all fields of data science. Deep learning has been successfully applied in the field of radioastronomy, such as the identification and classification of pulsar candidates. Deep learning requires a very large data set training network to achieve high performance, but so far the number of FRBs that have been detected is only two digits, not enough to build a meaningful training set for deep learning.
Disclosure of Invention
In order to overcome the problems, the invention provides a fast radiostorm search method based on oblique line detection and curve fitting dispersion elimination.
The technical scheme adopted by the invention is as follows: the fast radio storm searching method based on oblique line detection and curve fitting dispersion elimination comprises the following steps:
(1) Denoising by utilizing multi-core oblique line detection; performing sliding window operation on the time-frequency diagram of the FRB, and performing channel dimension splicing on the original diagram, the denoising result diagram and the expansion result to obtain an input diagram of the neural network;
(2) Carrying out semantic level segmentation by utilizing a U-shaped network; binary segmentation is carried out on the FRB time-frequency image by utilizing a U-shaped network;
(3) FRB detection; the binary segmentation result is respectively projected to the transverse direction w and the longitudinal direction h, the percentage ratio of the projected pixel quantity to the transverse direction w and the longitudinal direction h is calculated, and if the ratio is larger than delta, the ratio is judged as an FRB candidate;
(4) Extracting a framework; processing the predicted FRB candidate result by using a skeleton extraction algorithm to obtain a refined skeleton of the FRB;
(5) Fitting the FRB curve;
(6) Primary correction; respectively correcting the original image, the binary segmentation result and the skeleton extraction result for the first time by using a fitting expression;
(7) Correcting residual errors; and respectively carrying out secondary correction on the original image, the binary segmentation result and the skeleton extraction result by using a new expression.
Further, the step 1 includes the steps of:
1a) Generating multiple slopes; the FRB signal characteristics conform to a known expression, an expression is created by using a dispersion value, and segmentation is carried out according to frequency to obtain a plurality of different slopes;
1b) Constructing a diagonal detection convolution kernel; generating convolution kernels with different sizes for each slope obtained in step 1 a;
1c) Denoising and signal enhancement are carried out by utilizing a convolution kernel; performing sliding window operation on the time-frequency diagram of the FRB by adopting the convolution check generated in the step 1b, performing segmented convolution on different frequency segments, performing splicing and adding operations, generating a denoising result, and performing morphological expansion on the denoising result; and finally, splicing the original image, the denoising result image and the expansion result in the channel dimension to obtain an input image of the neural network.
Further, in the step 1a, the FRB corresponds to the following expression:
in the formula, v lo And v hi The lowest frequency and the highest frequency, respectively, DM is the dispersion value and Δt is the time difference.
Further, in the step 1c, a sliding window operation is performed on the acquired time-frequency data, specifically, 2S is used as a window, and the step length is 1S.
Further, the step 2 includes the steps of:
2a) Constructing a large number of pseudo samples by using FRB signal characteristic expressions, wherein the labels are FRB segmentation results;
2b) Constructing a U-shaped network, wherein an encoder is a ResNet-34 model pre-trained on an ImageNet, training is carried out by utilizing the data set obtained in the step 2a, and the loss function is Tversky loss and binary cross entropy loss;
2c) And (3) performing end-to-end binary segmentation on the result obtained in the step (2 b).
Further, the definition of the loss function in the step 2b is as follows:
Loss_fun＝T(A，B)+BCE(A，B) (2)
wherein T (A, B) is Tversky loss:
wherein A represents output, B is label, in T (A, B), A-B represents FP false positive, B-A represents FN false negative, alphase:Sub>A and betase:Sub>A control false negative and false positive respectively, and BCE (A, B) is binary cross entropy loss.
Further, in step 2b, the trade-off between alpha and beta control false positives and false negatives is adjusted; the values of α and β are α=0.3, and β=0.7.
Further, the step 5 includes the steps of:
5a) Sequentially shielding the left part of noise points and the right part of noise points step by using a RANSAC algorithm, fitting the FRB curve, and generating two fitting expressions; averaging the two fitting expressions to obtain a final expression;
5b) The DM value is calculated from the final fitting parameters of the RANSAC algorithm.
Further, the RANSAC algorithm flow in steps 5a, 5b is as follows:
(5.1) taking into account a model with a minimum sampling set potential of n (n is the minimum number of samples required for initializing model parameters) and a sample set P, wherein the number of samples # of the set P (P) > n, randomly extracting a subset S of P containing n samples from P to initialize the model M;
and (5.2) a sample set with an error from the model M of less than a certain set threshold t in the remainder set sc=p\s and S constitute S. S is considered as the Set of interior points, which constitute a consistent Set of S (Consensus Set);
(5.3) if # (S) is greater than or equal to N, considering that the correct model parameters are obtained, and recalculating a new model M by using a least square method and the like by utilizing the set S (interior points); and randomly extracting new S again, and repeating the process.
And (5.4) after a certain sampling time is completed, if the consistent set is not found, the algorithm fails, otherwise, the maximum consistent set obtained after sampling is selected to judge the inner point and the outer point, and the algorithm is ended.
Further, the step 7 includes the steps of:
7a) Collecting pixel positions of a neighborhood of the correction result by using the corrected skeleton extraction result, and calculating to obtain deviation of each row, wherein the missing part is processed by using linear interpolation, and a new expression is obtained;
7b) And respectively carrying out secondary correction on the original image, the binary segmentation result and the skeleton extraction result by using a new expression.
The principle of the invention is as follows: carrying out sliding window operation on the acquired time frequency data (2S is used as a window, and the step length is 1S); denoising and signal enhancement are carried out by using a diagonal detection algorithm; binary segmentation is carried out on the FRB time-frequency image by utilizing a U-shaped network; FRB detection; fitting an FRB expression by using the binary segmentation result; finally, dispersion elimination is performed by using the expression. Since the pulse signal of the FRB presents a quadratic curve shape with a simple structure in a time-frequency domain, the FRB pulse sample can be established through simulation by fewer parameters. A fast radio storm searching method based on oblique line detection and curve fitting dispersion elimination is established by deep learning, and the bottleneck in precision and speed faced in the existing FRB searching work is solved.
The beneficial effects of the invention are as follows:
firstly, the invention provides a diagonal detection method, which can directly detect FRB from time-frequency data and uses a curve fitting strategy to eliminate chromatic dispersion, thereby solving the problem of huge calculation power consumption in the chromatic dispersion stage of the traditional algorithm and having great advantage in FRB searching efficiency.
Secondly, the oblique line detection method provided by the invention can well enhance the FRB signal and inhibit the strong noise signal in the weak FRB signal scene.
Thirdly, the problem that marked data sets are absent in astronomical data is solved. The invention solves the problem of dependence of the deep learning model on the sample by utilizing the strategy of constructing the pseudo sample.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a schematic flow chart of the present invention;
FIG. 3 is a schematic diagram of multi-convolution kernel diagonal detection;
fig. 4 is a dummy FRB sample;
FIG. 5 shows the normal FRB detection and achromatic results;
FIG. 6 shows the result of weak FRB detection and achromatic;
fig. 7 shows the result of FRB detection and achromatizing with frequency drift.
Detailed Description
The following description of the embodiments of the present invention will be made more apparent and fully hereinafter with reference to the accompanying drawings, in which some, but not all embodiments of the invention are shown. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In the description of the present invention, it should be noted that, as the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," "outer," and the like are used for convenience in describing the present invention and simplifying the description based on the azimuth or positional relationship shown in the drawings, it should not be construed as limiting the present invention, but rather should indicate or imply that the devices or elements referred to must have a specific azimuth, be constructed and operated in a specific azimuth. Furthermore, the terms "first," "second," "third," and the like, as used herein, are used for descriptive purposes only and are not to be construed as indicating or implying any relative importance.
In the description of the present invention, it should be noted that unless explicitly stated and limited otherwise, the terms "mounted," "connected," and "connected" should be construed broadly, and may be, for example, fixedly connected, detachably connected, or integrally connected; can be mechanically or electrically connected; can be directly connected or indirectly connected through an intermediate medium, and can be communication between two elements. The specific meaning of the above terms in the present invention will be understood in specific cases by those of ordinary skill in the art.
Example 1
Referring to the drawings, a fast radio storm search method based on nuclear diagonal detection and curve fitting dispersion elimination comprises the following steps:
and step 1, multi-core oblique line detection denoising.
The FRB expression is used to construct a multi-slope, multi-scale convolution kernel. And performing sliding window operation on the FRB time-frequency diagram, performing segmented convolution on the FRB time-frequency sliding window diagram by utilizing convolution check, and performing splicing, adding and adding operations to generate a denoising and signal enhancement result. Meanwhile, splicing the original image, the denoising result image and the morphological expansion image of the denoising result image in channel dimension, wherein a red channel is the original image, a green channel is the denoising result image, and a blue channel is the morphological expansion image, so as to finally obtain an input image of the neural network; the method comprises the following steps:
(1a) Generating multiple slopes
From experience, it is known that FRB conforms to the expression,
wherein v is lo And v hi The lowest frequency and the highest frequency, respectively, DM is the dispersion value (Dispersion Measure, DM), Δt is the time difference. And selecting a common DM value, generating an expression, and segmenting according to the frequency to obtain a plurality of different slopes.
(1b) Construction of a diagonal detection convolution kernel
For each slope generated in 1a, a different size convolution kernel is generated.
(1c) Denoising and signal enhancement using convolution kernels
And (3) carrying out sliding window operation (2S is used as a window, the step length is 1S) on the time-frequency diagram of the FRB by adopting the convolution kernel generated in the step 1a, carrying out sectional convolution on different frequency sections, carrying out splicing and adding operation, generating a denoising result, and carrying out morphological expansion on the denoising result. And finally, splicing the original image, the denoising result image and the expansion result in the channel dimension to obtain an input image of the neural network.
And 2, carrying out semantic level segmentation by utilizing a U-shaped network.
(2a) A large number of pseudo samples are constructed by using the FRB expression, and the labels are the division results of the FRB and are used as training sets.
(2b) Constructing a U-shaped network, selecting a composite loss function for training the U-shaped network,
loss fun1 ＝BCE(A，B) (4)
loss fun3 ＝T(A，B) (5)
wherein BCE (a, B) represents a binary cross entropy loss, T (a, B) is Tversky loss:
a represents output, B is Iabel, in T (A, B), A-B means FP (false positive) and B-A means FN (false negative), and alphase:Sub>A and betase:Sub>A control false negative and false positive, respectively. Because of the imbalance of positive and negative samples, we need to adjust α and β to control the trade-off between false positives and false negatives, α=0.3 and β=0.7 being common values; BCE (a, B) is a binary cross entropy loss.
T (a, B) may adjust the superparameter to mitigate sample imbalance. Predicting real data by using the model so as to obtain a binary segmentation result of FRB;
FRB detection.
The binary segmentation result is respectively projected to the transverse direction w and the longitudinal direction h, the percentage ratio of the projected pixel quantity to the transverse direction w and the longitudinal direction h is calculated, and if the ratio is larger than delta, the ratio is judged as an FRB candidate;
and 4, skeleton extraction.
Processing the prediction result by using a skeleton extraction algorithm in OpenCV to obtain a refined skeleton of the FRB;
and 5, fitting by using a RANSAC algorithm.
(5a) Utilizing a random sampling consistency (Random Sample Consensus, RANSAC) algorithm to sequentially and gradually shade left part noise points and right part noise points, fitting an FRB curve to generate two fitting expressions, and averaging the two fitting expressions to obtain a final expression;
(5b) Calculating a DM value according to the final fitting parameters of the RANSAC;
and 6, correcting once.
Respectively correcting the original image, the binary segmentation result and the skeleton extraction result for the first time by using a fitting expression;
and 7, correcting residual errors.
(7a) And (3) collecting pixel positions of a neighborhood of the correction result by using the corrected skeleton extraction result obtained in the step (4), and calculating to obtain the deviation of each row, wherein the missing part is processed by using linear interpolation, and a new expression is obtained.
(7b) And (3) respectively carrying out primary correction on the original image, the binary segmentation result and the skeleton extraction result by using a new expression.
Example two
In comparison with the first embodiment, the definition of the loss function in the step 2b is as follows:
Loss_fun＝T(A，B)+BCE(A，B) (2)
wherein T (A, B) is Tversky loss:
wherein A represents output, B is Iabel, in T (A, B), A-B represents FP false positive, B-A represents FN false negative, alphase:Sub>A and betase:Sub>A control false negative and false positive respectively, and BCE (A, B) is binary cross entropy loss.
The embodiments described in the present specification are merely examples of implementation forms of the inventive concept, and the scope of protection of the present invention should not be construed as being limited to the specific forms set forth in the embodiments, and the scope of protection of the present invention and equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.
Claims (10)
1. The fast radio storm searching method based on oblique line detection and curve fitting dispersion elimination is characterized by comprising the following steps:
(1) Denoising by utilizing multi-core oblique line detection; performing sliding window operation on the time-frequency diagram of the FRB, and performing channel dimension splicing on the original diagram, the denoising result diagram and the expansion result to obtain an input diagram of the neural network;
(2) Carrying out semantic level segmentation by utilizing a U-shaped network; binary segmentation is carried out on the FRB time-frequency image by utilizing a U-shaped network;
(3) FRB detection; the binary segmentation result is respectively projected to the transverse direction w and the longitudinal direction h, the percentage ratio of the projected pixel quantity to the transverse direction w and the longitudinal direction h is calculated, and if the ratio is larger than delta, the ratio is judged as an FRB candidate;
(4) Extracting a framework; processing the predicted FRB candidate result by using a skeleton extraction algorithm to obtain a refined skeleton of the FRB;
(5) Fitting the FRB curve;
(6) Primary correction; respectively correcting the original image, the binary segmentation result and the skeleton extraction result for the first time by using a fitting expression;
(7) Correcting residual errors; and respectively carrying out secondary correction on the original image, the binary segmentation result and the skeleton extraction result by using a new expression.
2. The fast radio storm search method based on diagonal detection and curve fitting dispersion of claim 1, wherein: the step 1 comprises the following steps:
1a) Generating multiple slopes; the FRB signal characteristics conform to a known expression, an expression is created by using a dispersion value, and segmentation is carried out according to frequency to obtain a plurality of different slopes;
1b) Constructing a diagonal detection convolution kernel; generating convolution kernels with different sizes for each slope obtained in step 1 a;
1c) Denoising and signal enhancement are carried out by utilizing a convolution kernel; performing sliding window operation on the time-frequency diagram of the FRB by adopting the convolution check generated in the step 1b, performing segmented convolution on different frequency segments, performing splicing and adding operations, generating a denoising result, and performing morphological expansion on the denoising result; and finally, splicing the original image, the denoising result image and the expansion result in the channel dimension to obtain an input image of the neural network.
3. The fast radio storm search method based on diagonal detection and curve fitting dispersion as claimed in claim 2, wherein: in the step 1a, the FRB corresponds to the following expression:
in the formula, v lo And v hi The lowest frequency and the highest frequency, respectively, DM is the dispersion value and Δt is the time difference.
4. The fast radio storm search method based on diagonal detection and curve fitting dispersion as claimed in claim 2, wherein: in the step 1c, a sliding window operation is performed on the acquired time-frequency data, specifically, 2S is used as a window, and the step length is 1S.
5. A fast radiostorm search method based on slash detection and curve fitting dispersion as claimed in claim 3 wherein: the step 2 comprises the following steps:
2a) Constructing a large number of pseudo samples by using FRB signal characteristic expressions, wherein the labels are FRB segmentation results;
2b) Constructing a U-shaped network, wherein an encoder is a ResNet-34 model pre-trained on an ImageNet, training is carried out by utilizing the data set obtained in the step 2a, and the loss function is Tverskylos and binary cross entropy loss;
2c) And (3) performing end-to-end binary segmentation on the result obtained in the step (2 b).
6. The fast radio storm search method based on diagonal detection and curve fitting dispersion of claim 5, wherein: the definition of the loss function in step 2b is as follows:
Loss_fun＝T(A，B)+BCE(A，B) (2)
wherein T (A, B) is Tversky loss:
wherein A represents output, B is label, in T (A, B), A-B represents FP false positive, B-A represents FN false negative, alphase:Sub>A and betase:Sub>A control false negative and false positive respectively, and BCE (A, B) is binary cross entropy loss.
7. The fast radio storm search method based on diagonal detection and curve fitting dispersion of claim 6, wherein: in step 2b, the trade-off between alpha and beta control false positives and false negatives is adjusted; the values of α and β are α=0.3, and β=0.7.
8. The fast radiostorm search method based on diagonal detection and curve fitting dispersion of claim 7 wherein: said step 5 comprises the steps of:
5a) Sequentially shielding the left part of noise points and the right part of noise points step by using a RANSAC algorithm, fitting the FRB curve, and generating two fitting expressions; averaging the two fitting expressions to obtain a final expression;
5b) The DM value is calculated from the final fitting parameters of the RANSAC algorithm.
9. The fast radiostorm search method based on diagonal detection and curve fitting dispersion of claim 8, wherein: the RANSAC algorithm flow in steps 5a, 5b is as follows:
(5.1) taking into account a model with a minimum sampling set potential of n (n is the minimum number of samples required for initializing model parameters) and a sample set P, wherein the number of samples # of the set P (P) > n, randomly extracting a subset S of P containing n samples from P to initialize the model M;
and (5.2) a sample set with an error from the model M of less than a certain set threshold t in the remainder set sc=p\s and S constitute S. S is considered as the Set of interior points, which constitute a consistent Set of S (Consensus Set);
(5.3) if # (S) is greater than or equal to N, considering that the correct model parameters are obtained, and recalculating a new model M by using a least square method and the like by utilizing the set S (interior points); and randomly extracting new S again, and repeating the process.
And (5.4) after a certain sampling time is completed, if the consistent set is not found, the algorithm fails, otherwise, the maximum consistent set obtained after sampling is selected to judge the inner point and the outer point, and the algorithm is ended.
10. The fast radiostorm search method based on diagonal detection and curve fitting dispersion of claim 9, wherein: the step 7 comprises the following steps:
7a) Collecting pixel positions of a neighborhood of the correction result by using the corrected skeleton extraction result, and calculating to obtain deviation of each row, wherein the missing part is processed by using linear interpolation, and a new expression is obtained;
7b) And respectively carrying out secondary correction on the original image, the binary segmentation result and the skeleton extraction result by using a new expression.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310212412.0A CN116304624B (en) | 2023-02-27 | 2023-02-27 | Quick radio storm searching method based on oblique line detection and curve fitting dispersion elimination |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310212412.0A CN116304624B (en) | 2023-02-27 | 2023-02-27 | Quick radio storm searching method based on oblique line detection and curve fitting dispersion elimination |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116304624A true CN116304624A (en) | 2023-06-23 |
| CN116304624B CN116304624B (en) | 2024-01-30 |
Family
ID=86826821
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310212412.0A Active CN116304624B (en) | 2023-02-27 | 2023-02-27 | Quick radio storm searching method based on oblique line detection and curve fitting dispersion elimination |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116304624B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117556266A (en) * | 2024-01-11 | 2024-02-13 | 之江实验室 | Model training method, signal identification method and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020086707A1 (en) * | 2000-11-15 | 2002-07-04 | Struhsaker Paul F. | Wireless communication system using block filtering and fast equalization-demodulation and method of operation |
| CN110427878A (en) * | 2019-07-31 | 2019-11-08 | 中国科学院新疆天文台 | A kind of sudden and violent signal recognition method of Rapid Radio and system |
| CN115499092A (en) * | 2022-07-28 | 2022-12-20 | 南阳理工学院 | Astronomical radio transient signal searching method, system, device and readable storage medium |
-
2023
- 2023-02-27 CN CN202310212412.0A patent/CN116304624B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020086707A1 (en) * | 2000-11-15 | 2002-07-04 | Struhsaker Paul F. | Wireless communication system using block filtering and fast equalization-demodulation and method of operation |
| CN110427878A (en) * | 2019-07-31 | 2019-11-08 | 中国科学院新疆天文台 | A kind of sudden and violent signal recognition method of Rapid Radio and system |
| CN115499092A (en) * | 2022-07-28 | 2022-12-20 | 南阳理工学院 | Astronomical radio transient signal searching method, system, device and readable storage medium |
Non-Patent Citations (2)
| Title |
|---|
| LIAM CONNOR ET AL.: "Applying Deep Learning to Fast Radio Burst Classification", 《THE ASTRONOMICAL JOURNAL》 * |
| 许余云;李;刘志杰;王晨;王培;张蕾;潘之辰;: "人工智能在脉冲星候选体筛选中的应用", 天文学进展, no. 03 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117556266A (en) * | 2024-01-11 | 2024-02-13 | 之江实验室 | Model training method, signal identification method and device |
| CN117556266B (en) * | 2024-01-11 | 2024-03-22 | 之江实验室 | Signal identification model training method, signal identification method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116304624B (en) | 2024-01-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109784283B (en) | Remote sensing image target extraction method based on scene recognition task | |
| CN112308133A (en) | Modulation identification method based on convolutional neural network | |
| US20220309772A1 (en) | Human activity recognition fusion method and system for ecological conservation redline | |
| CN116304624B (en) | Quick radio storm searching method based on oblique line detection and curve fitting dispersion elimination | |
| CN111460936A (en) | Remote sensing image building extraction method, system and electronic equipment based on U-Net network | |
| CN110147760B (en) | Novel efficient electric energy quality disturbance image feature extraction and identification method | |
| Tian et al. | Multiscale building extraction with refined attention pyramid networks | |
| CN114221790A (en) | BGP (Border gateway protocol) anomaly detection method and system based on graph attention network | |
| CN112989995B (en) | Text detection method and device and electronic equipment | |
| CN110363163B (en) | SAR target image generation method with controllable azimuth angle | |
| CN114092815B (en) | Remote sensing intelligent extraction method for large-range photovoltaic power generation facility | |
| CN112036239A (en) | Radar signal working mode identification method and system based on deep learning network | |
| CN115797216A (en) | Inscription character restoration model and restoration method based on self-coding network | |
| CN109685823B (en) | Target tracking method based on deep forest | |
| CN109145738B (en) | Dynamic video segmentation method based on weighted non-convex regularization and iterative re-constrained low-rank representation | |
| CN111079608A (en) | Quick radio storm real-time searching method | |
| CN112906800B (en) | Image group self-adaptive collaborative saliency detection method | |
| CN111104850A (en) | Remote sensing image building automatic extraction method and system based on residual error network | |
| CN112751633B (en) | Broadband spectrum detection method based on multi-scale window sliding | |
| CN110517272B (en) | Deep learning-based blood cell segmentation method | |
| CN112861670B (en) | Transmission line hardware detection method and system | |
| CN112014821B (en) | Unknown vehicle target identification method based on radar broadband characteristics | |
| CN112434685B (en) | Remote sensing image information fine extraction method based on earth surface elements | |
| CN114970601A (en) | Power equipment partial discharge type identification method, equipment and storage medium | |
| CN109212501B (en) | Radar high-resolution range profile target identification method based on local embedding |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |