Multi-scale signal detection and extraction method and device

Abstract

The invention provides a multi-scale signal detection and extraction method and device, and belongs to the technical field of communication, and the method comprises the steps: carrying out the image segmentation through obtaining a time-frequency image of a full-band short-wave signal, and obtaining a segmented sub-image set; inputting the cut sub-image set into a preset first target detection model for large-scale coarse detection, and outputting a first detection result; based on the first detection result, narrow-band signals in a target frequency band in the full-frequency-band short-wave signals are extracted, and narrow-band signal samples are obtained; and inputting the narrowband signal sample into a preset second target detection model for small-scale fine detection, outputting a second detection result, and merging the first detection result and the second detection result to obtain a target detection result. According to the invention, the deep learning model is adopted for preliminary identification on a large scale, and the signal detection efficiency is improved. And finally, a high-resolution narrow-band signal sample is used for accurate identification at a small-scale level, and the response capability in a complex signal environment is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a multi-scale signal detection and extraction method and device. Background Art

[0002] The field of full-band shortwave signal detection is facing a severe challenge: accurately identifying and extracting weak or noise-buried target signals in a complex and changing electromagnetic environment. In order to meet the detection and processing speed of wide-band signals, traditional detection systems use low-frequency resolution time-frequency images as input. This strategy sacrifices the requirement for high-precision signal detection, especially in low-intensity or cluttered background conditions, resulting in an increase in the probability of false detection. The limitation of current technology is that it is difficult to achieve highly accurate recognition while maintaining real-time performance. Summary of the invention

[0003] The present invention provides a multi-scale signal detection and extraction method and device to solve the defect of the prior art that it is difficult to achieve highly accurate recognition while maintaining real-time performance. By constructing a full-band shortwave signal detection and extraction framework that integrates fast and accurate performance, and combining the advantages of multi-scale recognition algorithms, the limitations of traditional methods in processing weak signals are overcome.

[0004] In a first aspect, the present invention provides a multi-scale signal detection and extraction method, comprising:

[0005] Obtain the time-frequency image of the full-band shortwave signal to perform image segmentation to obtain a segmentation sub-atlas;

[0006] Inputting the slice sub-atlas into a preset first target detection model, and outputting a first detection result, wherein the first detection result is signal detection frame result information corresponding to the signal in each slice sub-atlas;

[0007] Based on the first detection result, extracting the narrowband signal in the target frequency band from the full-band shortwave signal to obtain a narrowband signal sample;

[0008] The narrowband signal sample is input into a preset second target detection model, a second detection result is output, and the first detection result and the second detection result are merged to obtain the detection result of the target, wherein the second detection result is the signal detection box result information corresponding to each narrowband signal.

[0009] According to the multi-scale signal detection and extraction method, based on the first detection result, extracting the narrowband signal in the target frequency band of the full-band shortwave signal to obtain a narrowband signal sample includes:

[0010] Integrate the preliminary detection results of the slice subgraphs with the same index on the time axis, merge the same signal, and obtain the merged detection results;

[0011] Based on the combined detection result, the narrowband signal in the target frequency band of the full-band shortwave signal is extracted to obtain a narrowband signal sample.

[0012] According to the multi-scale signal detection and extraction method, the time-frequency image of the full-band shortwave signal is obtained and image segmentation is performed to obtain a segmentation sub-atlas, including:

[0013] The time-frequency image of the full-band shortwave signal is obtained by segmenting on the time and frequency axes using an overlapping segmentation strategy to obtain a segmentation subatlas.

[0014] According to the multi-scale signal detection and extraction method, the step of acquiring the time-frequency image of the full-band shortwave signal includes:

[0015] Obtain a broadband complex signal based on the full-band shortwave signal, and frame the broadband complex signal to obtain framed broadband complex signal data;

[0016] The framed broadband complex signal data is subjected to Fourier transformation to obtain broadband complex signal frequency domain data, and the broadband complex signal frequency domain data is accumulated to obtain a time-frequency image.

[0017] According to the multi-scale signal detection and extraction method, the preliminary detection results of the slice subgraphs with the same index on the time axis are integrated, and the same signal is merged to obtain the merged detection result, including:

[0018] The detection results of the sliced ​​subgraphs with the same index on the time axis are merged, the same signal is determined based on the intersection-and-union ratio, and the overlapping parts of the boundaries of the same signal are integrated to obtain the merged detection result.

[0019] According to the multi-scale signal detection and extraction method, the first detection result and the second detection result are combined to obtain a detection result of the target, including:

[0020] If the second detection result is a signal type, and no signal is detected within the preset time index before the signal frequency point, it means that the signal appears for the first time, and the signal is added to the detection result signal list;

[0021] If the second detection result is a background noise type, and no signal is detected within a preset time index before the signal frequency point, it is considered that the rough detection is wrong, the signal extraction is terminated, and the extracted signal information is discarded;

[0022] If the second detection result is a signal type, and a signal is detected within a preset time index before the frequency point, the detection is merged with the previous detection result, the parameters of the detection signal at the same frequency point in the detection result signal list are kept unchanged, and the signal is continued to be extracted;

[0023] If the second detection result is a background noise type, and a signal is detected within a preset time index before the frequency point, it is considered that the signal has stopped sending, the signal is stopped from being extracted, and the signal list deletes the signal.

[0024] According to the multi-scale signal detection and extraction method, the signal is not detected within a preset time index before the frequency point, including:

[0025] Calculate the spectrum intersection and combination ratio with each signal in the detection result signal list, the intersection and combination ratio is less than the preset threshold, or the intersection and combination ratio is greater than the preset threshold, but the signal type is different from the signal that appeared within the preset time index before the frequency point;

[0026] The signal is detected within a preset time index before the frequency point, including:

[0027] The spectrum intersection and union ratio is calculated with each signal in the detection result signal list. The intersection and union ratio is greater than the preset threshold, and the signal type is the same as the signal that appeared within the preset time index before the frequency point.

[0028] In a second aspect, the present invention further provides a multi-scale signal detection and extraction device, comprising:

[0029] A segmentation unit is used to obtain the time-frequency image of the full-band shortwave signal to segment the image and obtain a segmentation sub-atlas;

[0030] A first detection unit, configured to input the slice sub-atlas into a preset first target detection model, and output a first detection result, wherein the first detection result is signal detection frame result information corresponding to a signal in each slice sub-atlas;

[0031] A processing unit, configured to extract a narrowband signal within a target frequency band from the full-band shortwave signal based on the first detection result to obtain a narrowband signal sample;

[0032] The second detection unit is used to input the narrowband signal sample into a preset second target detection model, output a second detection result, merge the first detection result and the second detection result to obtain the detection result of the target, wherein the second detection result is the signal detection box result information corresponding to each narrowband signal.

[0033] In a third aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of any of the multi-scale signal detection and extraction methods described above are implemented.

[0034] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the multi-scale signal detection and extraction methods described above.

[0035] The multi-scale signal detection and extraction method and device provided by the present invention obtains the time-frequency image of the full-band shortwave signal for image segmentation to obtain a segmentation sub-atlas; the segmentation sub-atlas is input into a preset first target detection model for large-scale coarse detection, and a first detection result is output; based on the first detection result, the narrowband signal in the target frequency band of the full-band shortwave signal is extracted to obtain a narrowband signal sample; the narrowband signal sample is input into a preset second target detection model for small-scale fine detection, and a second detection result is output, and the first detection result and the second detection result are merged to obtain the detection result of the target. A deep learning model is used for preliminary recognition on a large scale to improve the signal detection efficiency. Finally, high-resolution narrowband signal samples are used for accurate recognition at a small scale level to enhance the response capability in a complex signal environment. The present invention overcomes the limitations of traditional methods in processing weak signals by combining the advantages of multi-scale recognition algorithms. The scheme of the present invention can provide strong support for frequency resource management, interference monitoring and security protection in the field of communications. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0037] Figure 1 It is a schematic diagram of the process of the multi-scale signal detection and extraction method provided by the present invention;

[0038] Figure 2 It is a schematic diagram of the broadband signal time-frequency image segmentation process provided by the present invention;

[0039] Figure 3 It is a schematic diagram of the detection information of the molecular graphs indexed by merging the same time axis provided by the present invention;

[0040] Figure 4 It is a schematic diagram of judging the validity of a detection signal frame provided by the present invention;

[0041] 5 is a schematic diagram of maintaining a detection list and updating or merging signals according to small-scale detection results provided by the present invention;

[0042] Figure 6 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] It should be noted that, in the description of the embodiments of the present invention, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "include one..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0045] Combine the following Figure 1-Figure 6 The multi-scale signal detection and extraction method and device provided by the embodiments of the present invention are described.

[0046] Figure 1 is a flow chart of the multi-scale signal detection and extraction method provided by the present invention, such as Figure 1 As shown, including but not limited to the following steps (STEP, referred to as S):

[0047] S102: Obtaining a time-frequency image of a full-band shortwave signal to perform image segmentation to obtain a segmentation subatlas.

[0048] S104: inputting the slice sub-atlas into a preset first target detection model, and outputting a first detection result, wherein the first detection result corresponds to a signal in each slice sub-atlas;

[0049] S106: Based on the preliminary detection result, extract the narrowband signal in the target frequency band from the full-band shortwave signal to obtain a narrowband signal sample;

[0050] S108: Input the narrowband signal sample into a preset second target detection model, output a second detection result, and merge the first detection result and the second detection result to obtain a detection result of the target.

[0051] An optional multi-scale signal detection and extraction method provided in the present application can be used to accurately detect full-band shortwave signals and identify target signals that are weak or buried in noise.

[0052] In the full-band shortwave signal detection and recognition of the present embodiment, the original full-band broadband signal is first converted into frequency domain data, and a broadband time-frequency image is formed by accumulating data from different time segments. The time-frequency image is then decomposed into multiple sub-images, and a preset first target detection model is used for detection to obtain a first detection result; the detection results corresponding to the multiple sub-images are then merged; based on the signal center frequency point and the preset bandwidth obtained by the detection result, narrowband signal samples are extracted; a high-resolution narrowband signal time-frequency diagram is generated using the extracted narrowband signal samples, and accurate detection and recognition are performed based on the preset second target detection model.

[0053] The preset first target detection model may be a deep learning detection model pre-trained based on the sliced ​​sub-atlas samples and the signal detection frame result information labels corresponding to the sliced ​​sub-atlas, which is a large-scale coarse detection. The preset second target detection model may be a deep learning detection model or a deep learning recognition model pre-trained based on the narrowband signal samples and the signal detection frame result information labels corresponding to the narrowband signal samples, which is a small-scale fine detection.

[0054] Based on the contents of the above embodiments, as an optional embodiment, the multi-scale signal detection and extraction method provided by the present invention obtains the time-frequency image of the full-band shortwave signal and performs image segmentation to obtain a segmentation sub-atlas, including: obtaining the time-frequency image of the full-band shortwave signal and using an overlapping segmentation strategy to perform segmentation on the time and frequency axes to obtain a segmentation sub-atlas.

[0055] Optionally, the step of acquiring the time-frequency image of the full-band shortwave signal includes:

[0056] S11: obtaining a broadband complex signal based on the full-band shortwave signal, and framing the broadband complex signal to obtain framed broadband complex signal data;

[0057] S12: Perform Fourier transform on the framed broadband complex signal data to obtain broadband complex signal frequency domain data, and accumulate the broadband complex signal frequency domain data to obtain a time-frequency image.

[0058] In this embodiment, a broadband complex signal is obtained from a signal receiving device. X , and the broadband complex signal is framed to obtain the framed broadband complex signal data X i , i = 1, 2, ..., assuming that the sampling rate of the broadband complex signal is f s0 , the coarse detection accuracy is set to f p Hz, each frame signal X i Length L f Depend on Calculated;

[0059] For each frame of broadband complex signal X i, i=1,2,… respectively perform Fourier transform to obtain the broadband complex signal frequency domain data Y i ,i=1,2,…。

[0060] Furthermore, for the complex signal frequency domain data Y i , i=1,2,…are accumulated to form a time-frequency image I, and the time-frequency image I is segmented to be suitable as an input for deep neural network image detection.

[0061] To avoid the impact of new boundaries on signal detection after image segmentation, the number of time overlap points N can be set. ot , the number of frequency overlap points N of , time segment size N t , time segment size N f , the time-frequency image I is segmented to form a segmented time-frequency image I ij , where i=1,2,… represents the image time axis index, represents the image frequency axis index, where is the maximum index of the image frequency axis. If the width of the segmented image is less than N f , fill the insufficient part of the segmented image with 0, Figure 2 The schematic diagram of the broadband signal time-frequency image segmentation process provided in this embodiment, the image segmentation process, such as Figure 2 shown.

[0062] Based on the content of the above embodiment, as an optional embodiment, the multi-scale signal detection and extraction method provided by the present invention extracts the narrowband signal in the target frequency band of the full-band shortwave signal based on the preliminary detection result, including:

[0063] S21: Integrate the preliminary detection results of the slice subgraphs with the same index on the time axis, merge the same signal, and obtain the merged detection result;

[0064] S21: Based on the combined detection result, extract the narrowband signal in the target frequency band of the full-band shortwave signal to obtain a narrowband signal sample.

[0065] Optionally, the step of integrating the preliminary detection results of the slice subgraphs with the same index on the time axis and merging the same signal to obtain the merged detection result includes:

[0066] The detection results of the sliced ​​subgraphs with the same index on the time axis are merged, the same signal is determined based on the intersection-and-union ratio, and the overlapping parts of the boundaries of the same signal are integrated to obtain the merged detection result.

[0067] In this embodiment, Figure 3A schematic diagram of merging detection information of the same time axis index cut molecular graph provided in this embodiment, such as Figure 3 As shown, for the same time axis index cut subgraph Based on the information of the identified signal detection frames, the merging process is performed. Since the signal may cross the sub-image boundary, causing detection errors, each sub-image sets an effective detection boundary and discards the detection frame whose center is outside the effective detection boundary. Figure 4 The schematic diagram of judging the validity of the detection signal frame provided in this embodiment is as follows: Figure 4 Then, the detection results of the same signal from different sub-images with the same time index are merged.

[0068] Assume that the effective detection boundary length is set to N e , where N e ≤N of To prevent the signal from being missed due to jitter in the detection boundary box, if x c <N e or x c >N f -N e , that is, the center of the detection frame is outside the effective detection boundary, then the detection frame is considered invalid. c ,y c ) ijk , mapping it to the undivided time-frequency image position (x ′ c ,y ′ c ) il , where i = 1, 2, ... represents the image time axis index, Represents the image frequency axis index, k represents the kth valid detection frame in the time-frequency segmentation subgraph ij, and l represents the valid detection frame index of the unsegmented time-frequency image at the same time index.

[0069] By calculating the undivided time-frequency image signal detection frame (x ′ c ,y ′ c ) il Intersection and Union IOU Determine whether it is the same signal and use the preset threshold T If the intersection-and-union ratio (IOU) of any two signals is ≥ T, the two signals are merged, and the merged detection boundary takes the smallest rectangular detection frame that can cover the union of the two detection frames, and (x″ c ,y″ c ) in , (w″,h″) in .

[0070] Based on the content of the above embodiment, as an optional embodiment, in the multi-scale signal detection and extraction method provided by the present invention, the first detection result and the second detection result are merged to obtain the detection result of the target, including:

[0071] If the second detection result is a signal type, and no signal is detected within the preset time index before the signal frequency point, it means that the signal appears for the first time, and the signal is added to the detection result signal list;

[0072] If the second detection result is a background noise type, and no signal is detected within a preset time index before the signal frequency point, it is considered that the rough detection is wrong, the signal extraction is terminated, and the extracted signal information is discarded;

[0073] If the second detection result is a signal type, and a signal is detected within a preset time index before the frequency point, the detection is merged with the previous detection result, the parameters of the detection signal at the same frequency point in the detection result signal list are kept unchanged, and the signal is continued to be extracted;

[0074] If the second detection result is a background noise type, and no signal is detected within a preset time index before the frequency point, it is considered that the signal has stopped sending, the signal is stopped from being extracted, and the signal list deletes the signal.

[0075] In this embodiment, FIG5 is a schematic diagram of maintaining a detection list provided in this embodiment, and updating or merging signals according to small-scale detection results. As shown in FIG5, the detection result signal list is maintained and the detection results are merged. Since the preset second target detection model in small-scale precision detection has a higher frequency accuracy, the signal type detection result can generally be based on the small-scale detection result. The following is an example of the merging process of the detection results divided into several cases:

[0076] 1) If the recognition result of the preset second target detection model of the small-scale precision detection is a signal type, and no signal is detected within the M time indexes before the frequency point, it means that the signal appears for the first time. The signal is added to the signal list and the precision detection signal center frequency point f is used. n ′ Bandwidth B ′ n and signal type c ′ n Mark signal.

[0077] 2) If the recognition result of the preset second target detection model of the small-scale fine detection is the background noise type, and no signal is detected within the M time indexes before the frequency point, it is considered that the coarse detection is wrong, the signal extraction is terminated, and the extracted signal information is discarded.

[0078] 3) If the recognition result of the preset second target detection model of small-scale precision detection is a signal type, and the signal is detected within M time indexes before the frequency point, the detection is merged with the previous detection result, and the detection signal parameters of the same frequency point in the signal list (such as the detection signal center frequency, bandwidth, signal type) are kept unchanged, and the signal continues to be extracted.

[0079] 4) If the recognition result of the preset second target detection model of small-scale precision detection is the background noise type, and no signal is detected within the M time indexes before the frequency point, it is considered that the signal has stopped sending, the signal is stopped from being extracted, and the signal list deletes the signal.

[0080] Furthermore, the fact that no signal is detected within a preset time index before the frequency point includes:

[0081] Calculate the spectrum intersection and combination ratio with each signal in the detection result signal list, the intersection and combination ratio is less than the preset threshold, or the intersection and combination ratio is greater than the preset threshold, but the signal type is different from the signal that appeared within the preset time index before the frequency point;

[0082] The signal is detected within a preset time index before the frequency point, including:

[0083] The spectrum intersection and union ratio is calculated with each signal in the detection result signal list. The intersection and union ratio is greater than the preset threshold, and the signal type is the same as the signal that appeared within the preset time index before the frequency point.

[0084] In this embodiment, if no signal is detected within the M time indexes before the frequency point, the spectrum intersection and combination ratio is calculated with each signal in the detection list, and the intersection and combination ratio is less than the threshold, or the intersection and combination ratio is greater than the threshold, but the signal type c ′ n Different from the previous signal.

[0085] The signal detected within the M time indexes before the frequency point is the spectrum intersection and union ratio calculated with each signal in the detection list. The intersection and union ratio is greater than the threshold, and the signal type c ′ n The same signal as before.

[0086] In summary, the present invention proposes a multi-scale signal detection and extraction method, which quickly excludes non-relevant signal areas in the initial large-scale detection, and then conducts in-depth analysis at a high-resolution small-scale level to ensure accurate capture of the target signal.

[0087] The key to the present invention is to use advanced artificial intelligence technology and multi-scale analysis strategies to optimize the automatic detection and extraction process of full-band shortwave signals. The system first converts the original broadband complex signal into a frequency domain representation and forms a broadband time-frequency image as the basis for subsequent processing. Subsequently, a deep learning model is used for preliminary recognition on a large scale to improve signal detection efficiency. Finally, high-resolution signal samples are used at a small scale for accurate recognition to enhance the response capability in complex signal environments.

[0088] In summary, the present invention aims to build a fast and accurate full-band shortwave signal detection and extraction framework, which overcomes the limitations of traditional methods in processing weak signals by combining the advantages of multi-scale recognition algorithms. The innovative solution of the present invention can provide strong support for frequency resource management, interference monitoring and security protection in the field of communications.

[0089] The present invention also provides a multi-scale signal detection and extraction device, which includes: a graph optimization model construction module, a cost function construction module and a unified benchmark construction module.

[0090] A segmentation unit is used to obtain the time-frequency image of the full-band shortwave signal to segment the image and obtain a segmentation sub-atlas;

[0091] A first detection unit, configured to input the slice sub-atlas into a preset first target detection model, and output a first detection result, wherein the first detection result is signal detection frame result information corresponding to a signal in each slice sub-atlas;

[0092] A processing unit, configured to extract a narrowband signal within a target frequency band from the full-band shortwave signal based on the first detection result to obtain a narrowband signal sample;

[0093] The second detection unit is used to input the narrowband signal sample into a preset second target detection model, output a second detection result, merge the first detection result and the second detection result to obtain the detection result of the target, wherein the second detection result is the signal detection box result information corresponding to each narrowband signal.

[0094] It should be noted that the multi-scale signal detection and extraction device provided in the embodiment of the present invention can execute the multi-scale signal detection and extraction method described in any of the above embodiments during specific operation, which will not be elaborated in this embodiment.

[0095] Figure 6 is a schematic diagram of the structure of the electronic device provided by the present invention, such as Figure 6As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the multi-scale signal detection and extraction method, which includes: obtaining the time-frequency image of the full-band shortwave signal for image segmentation to obtain a segmentation sub-atlas; inputting the segmentation sub-atlas into a preset first target detection model and outputting a first detection result; based on the first detection result, extracting the narrowband signal in the target frequency band of the full-band shortwave signal to obtain a narrowband signal sample; inputting the narrowband signal sample into a preset second target detection model and outputting a second detection result, merging the first detection result and the second detection result to obtain a target detection result.

[0096] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0097] On the other hand, the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the multi-scale signal detection and extraction method provided by the above embodiments, and the method includes: acquiring a time-frequency image of a full-band shortwave signal for image segmentation to obtain a segmentation sub-atlas; inputting the segmentation sub-atlas into a preset first target detection model, and outputting a first detection result, wherein the first detection result is signal detection box result information corresponding to the signal in each segmentation sub-atlas; based on the first detection result, extracting a narrowband signal in a target frequency band in the full-band shortwave signal to obtain a narrowband signal sample; inputting the narrowband signal sample into a preset second target detection model, and outputting a second detection result, and merging the first detection result and the second detection result to obtain a detection result of the target, wherein the second detection result is signal detection box result information corresponding to each narrowband signal.

[0098] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the multi-scale signal detection and extraction method provided in the above-mentioned embodiments, the method comprising: acquiring a time-frequency image of a full-band shortwave signal for image segmentation to obtain a segmentation sub-atlas; inputting the segmentation sub-atlas into a preset first target detection model, and outputting a first detection result, wherein the first detection result is signal detection box result information corresponding to the signal in each segmentation sub-atlas; based on the first detection result, extracting a narrowband signal in a target frequency band in the full-band shortwave signal to obtain a narrowband signal sample; inputting the narrowband signal sample into a preset second target detection model, and outputting a second detection result, and merging the first detection result and the second detection result to obtain a detection result of the target, wherein the second detection result is signal detection box result information corresponding to each narrowband signal.

[0099] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0100] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0101] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-scale signal detection and extraction method, characterized in that: include: Obtain the time-frequency image of the full-band shortwave signal to perform image segmentation to obtain a segmentation sub-atlas; Inputting the slice sub-atlas into a preset first target detection model, and outputting a first detection result, wherein the first detection result is signal detection frame result information corresponding to the signal in each slice sub-atlas; Based on the first detection result, extracting the narrowband signal in the target frequency band from the full-band shortwave signal to obtain a narrowband signal sample; The narrowband signal sample is input into a preset second target detection model, a second detection result is output, and the first detection result and the second detection result are merged to obtain the detection result of the target; wherein the second detection result is the signal detection box result information corresponding to each narrowband signal.

2. The multi-scale signal detection and extraction method according to claim 1, characterized in that: The extracting, based on the first detection result, a narrowband signal in a target frequency band from the full-band shortwave signal to obtain a narrowband signal sample comprises: Integrate the preliminary detection results of the slice subgraphs with the same index on the time axis, merge the same signal, and obtain the merged detection results; Based on the combined detection result, the narrowband signal in the target frequency band of the full-band shortwave signal is extracted to obtain a narrowband signal sample.

3. The multi-scale signal detection and extraction method according to claim 1, characterized in that: The step of acquiring the time-frequency image of the full-band shortwave signal and performing image segmentation to obtain a segmentation sub-atlas comprises: The time-frequency image of the full-band shortwave signal is obtained by segmenting on the time and frequency axes using an overlapping segmentation strategy to obtain a segmentation subatlas.

4. The multi-scale signal detection and extraction method according to claim 3, characterized in that: The step of acquiring the time-frequency image of the full-band shortwave signal comprises: Obtain a broadband complex signal based on the full-band shortwave signal, and frame the broadband complex signal to obtain framed broadband complex signal data; The framed broadband complex signal data is subjected to Fourier transformation to obtain broadband complex signal frequency domain data, and the broadband complex signal frequency domain data is accumulated to obtain a time-frequency image.

5. The multi-scale signal detection and extraction method according to claim 2, characterized in that: The preliminary detection results of the slice subgraphs with the same index on the time axis are integrated, and the same signal is merged to obtain the merged detection results, including: The detection results of the sliced ​​subgraphs with the same index on the time axis are merged, the same signal is determined based on the intersection-and-union ratio, and the overlapping parts of the boundaries of the same signal are integrated to obtain the merged detection result.

6. The multi-scale signal detection and extraction method according to claim 1, characterized in that: Combining the first detection result and the second detection result to obtain a detection result of the target includes: If the second detection result is a signal type, and no signal is detected within the preset time index before the signal frequency point, it means that the signal appears for the first time, and the signal is added to the detection result signal list; If the second detection result is a background noise type, and no signal is detected within a preset time index before the signal frequency point, it is considered that the rough detection is wrong, the signal extraction is terminated, and the extracted signal information is discarded; If the second detection result is a signal type, and a signal is detected within a preset time index before the frequency point, the detection is merged with the previous detection result, the parameters of the detection signal at the same frequency point in the detection result signal list are kept unchanged, and the signal is continued to be extracted; If the second detection result is a background noise type, and a signal is detected within a preset time index before the frequency point, it is considered that the signal has stopped sending, the signal is stopped from being extracted, and the signal list deletes the signal.

7. The multi-scale signal detection and extraction method according to claim 6, characterized in that: The signal is not detected within a preset time index before the frequency point, including: Calculate the spectrum intersection and combination ratio with each signal in the detection result signal list, the intersection and combination ratio is less than the preset threshold, or the intersection and combination ratio is greater than the preset threshold, but the signal type is different from the signal that appears within the preset time index before the frequency point; The signal is detected within a preset time index before the frequency point, including: The spectrum intersection and union ratio is calculated with each signal in the detection result signal list. The intersection and union ratio is greater than the preset threshold, and the signal type is the same as the signal that appeared within the preset time index before the frequency point.

8. A multi-scale signal detection and extraction device, characterized in that: The device comprises: A segmentation unit is used to obtain the time-frequency image of the full-band shortwave signal to segment the image and obtain a segmentation sub-atlas; A first detection unit, configured to input the slice sub-atlas into a preset first target detection model, and output a first detection result, wherein the first detection result is signal detection frame result information corresponding to a signal in each slice sub-atlas; A processing unit, configured to extract a narrowband signal within a target frequency band from the full-band shortwave signal based on the first detection result to obtain a narrowband signal sample; The second detection unit is used to input the narrowband signal sample into a preset second target detection model, output a second detection result, merge the first detection result and the second detection result to obtain the detection result of the target, wherein the second detection result is the signal detection box result information corresponding to each narrowband signal.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the multi-scale signal detection and extraction method according to any one of claims 1 to 7 are implemented.

10. A non-transitory 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 multi-scale signal detection and extraction method according to any one of claims 1 to 7 are implemented.

Images