State intelligent perception method and system based on synthetic aperture radar multi-source data

Abstract

The present invention relates to the field of data processing technology, and in particular to a state intelligent perception method and system based on synthetic aperture radar multi-source data, including: obtaining several band monitoring data of SAR; obtaining the interference degree of the characteristic signal matrix of each band monitoring data; obtaining the denoised characteristic signal matrix of each band monitoring data according to the noise area evaluation degree of each signal characteristic window in the characteristic signal matrix of each band monitoring data; obtaining the regional state of each element corresponding to the geographical location in the denoised characteristic signal matrix according to the denoised characteristic signal matrix. The present invention improves the accuracy of the state perception of geographical information in the radar irradiation area.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a state intelligent perception method and system based on synthetic aperture radar multi-source data. Background Art

[0002] Synthetic Aperture Radar (SAR) is an active radar technology that obtains radar data of the target area from multiple perspectives and synthesizes it using signal processing technology to obtain high-resolution, all-weather, and all-day surface target images. SAR's multi-source data collects scattering characteristic information of the surface at different frequencies, and can obtain different geographic collection information based on the identification characteristics of different frequency bands. However, the reflection characteristics between different bands are actually different, so there are differences in the expression forms of the collected data of different bands. When the neural network identifies the matrix of collected data of different bands, there will be errors in the collection matrix due to the interference of the characteristics of different bands themselves on different areas of the surface. Therefore, it is necessary to remove the noise of the collected data, which helps to improve the accuracy of the data system in analyzing the surface status through SAR monitoring data.

[0003] Existing algorithms often use SPP-Net (spatial pyramid pooling) to extract features from SAR multi-source data. Since the SPP layer can process feature windows of different sizes, the noise data generated in some frequency bands will cause significant fluctuations in the SAR multi-source data, resulting in interference in the identification of key target areas in the radar illumination area, making the perception of the geographic information status in the radar illumination area inaccurate. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a state intelligent perception method and system based on synthetic aperture radar multi-source data.

[0005] An embodiment of the present invention provides a state intelligent perception method based on synthetic aperture radar multi-source data, the method comprising the following steps:

[0006] Acquire several band monitoring data of SAR; each band monitoring data corresponds to a data matrix, the data matrix contains multiple elements, and each element corresponds to a geographical location;

[0007] According to the data matrix of each band monitoring data, the characteristic signal matrix of each band monitoring data is obtained, and all signal characteristic windows of the characteristic signal matrix of each band monitoring data are obtained; according to the fluctuation of each element in the characteristic signal matrix of each band monitoring data, the interference degree of the characteristic signal matrix of each band monitoring data is obtained;

[0008] Obtain the obvious signal elements of each signal feature window in the characteristic signal matrix of each band monitoring data; obtain the noise area evaluation factor of each signal feature window according to the interference degree and the obvious signal elements of each signal feature window; clear and filter all signal feature windows according to the noise area evaluation factor to obtain the denoised characteristic signal matrix of each band monitoring data;

[0009] According to the denoised feature signal matrix, the regional status of the geographical location corresponding to each element in the denoised feature signal matrix is ​​obtained.

[0010] Preferably, the method of obtaining the characteristic signal matrix of each band monitoring data according to the data matrix of each band monitoring data and obtaining all signal characteristic windows of the characteristic signal matrix of each band monitoring data includes the following specific methods:

[0011] A parameter K is preset. For the r-th band monitoring data of SAR, the data matrix of the r-th band monitoring data is input into the convolution layer of the CNN neural network, and the characteristic signal matrix of the r-th band monitoring data is output. K windows are selected from the characteristic signal matrix of the r-th band monitoring data through a selective search algorithm, and the set composed of K windows is used as the signal feature window set of the characteristic signal matrix of the r-th band monitoring data.

[0012] Preferably, the specific formula for obtaining the interference degree of the characteristic signal matrix of each band monitoring data according to the fluctuation of each element in the characteristic signal matrix of each band monitoring data is:

[0013]

[0014] Where A represents the interference degree of the characteristic signal matrix of any band monitoring data; n represents the total number of all elements in the characteristic signal matrix of the band monitoring data; t i Represents the i-th element in the characteristic signal matrix of the band monitoring data; represents the average value of all elements in the characteristic signal matrix of the band monitoring data; σ represents the standard deviation of all elements in the characteristic signal matrix of the band monitoring data.

[0015] Preferably, the specific method of obtaining the obvious signal elements of each signal feature window in the characteristic signal matrix of each band monitoring data includes:

[0016] For the w-th signal feature window of the characteristic signal matrix of the r-th band monitoring data, all elements in the w-th signal feature window are traversed row by row, and all elements are extracted in sequence to form a first sequence, all elements in the w-th signal feature window are traversed column by column, and all elements are extracted in sequence to form a second sequence, the first sequence and the second sequence are merged and used as the feature sequence of the w-th signal feature window; the peak detection algorithm is used to obtain all maximum elements of the feature sequence of the w-th signal feature window, and all maximum elements are used as obvious signal elements of the w-th signal feature window.

[0017] Preferably, the method of obtaining the noise area evaluation factor of each signal feature window according to the interference degree and the obvious signal elements of each signal feature window includes the following specific methods:

[0018] For any obvious signal element in the w-th signal feature window, obtain the adjustable window after each increase of the obvious signal element and the total number of iterative increases of the adjustable window;

[0019] Obtain the noise area evaluation degree of each signal feature window according to the interference degree and the adjustable window after each increase of the obvious signal elements in each signal feature window and the total number of iterative increases of the adjustable window;

[0020] The noise region evaluation degrees of all signal feature windows are obtained, and each noise region evaluation degree after linear normalization of all noise region evaluation degrees is recorded as a noise region evaluation factor.

[0021] Preferably, the method of obtaining the adjustable window after each increase of the obvious signal element and the total number of iterative increases of the adjustable window includes the following specific methods:

[0022] An adjustable window is constructed with the obvious signal element as the center point, the initial window side length of the adjustable window is preset to 1, and the side length of the adjustable window is iteratively increased with a step size of 1 until the adjustable window contacts other obvious signal elements of the w-th signal feature window or the edge of the w-th signal feature window during the increasing process, then the iterative increase is stopped, and the adjustable window after each increase of the obvious signal element and the total number of iterative increases of the adjustable window are obtained.

[0023] Preferably, the specific formula for obtaining the noise area evaluation degree of each signal feature window according to the interference degree and the adjustable window after each increase of the obvious signal elements in each signal feature window and the total number of iterative increases of the adjustable window is:

[0024]

[0025] In the formula, B r,wIndicates the noise area evaluation degree of the wth signal feature window in the characteristic signal matrix of the rth band monitoring data; n r Represents the total number of all elements in the characteristic signal matrix of the r-th band monitoring data; n r,w represents the total number of all elements in the wth signal feature window in the characteristic signal matrix of the rth band monitoring data; m r,w Represents the total number of all obvious signal elements in the w-th signal feature window in the characteristic signal matrix of the r-th band monitoring data; N represents the average value of the Euclidean distance between all obvious signal elements in the w-th signal feature window in the characteristic signal matrix of the r-th band monitoring data; r,w,q The total number of iterative increases of the adjustable window of the qth obvious signal element of the wth signal feature window in the characteristic signal matrix of the rth band monitoring data; It represents the average value of all elements in the adjustable window after the jth increase of the qth obvious signal element in the wth signal feature window in the characteristic signal matrix of the rth band monitoring data; A represents the average value of all elements in the adjustable window after the qth obvious signal element in the wth signal feature window in the characteristic signal matrix of the rth band monitoring data is increased for the j-1th time; r A represents the interference degree of the characteristic signal matrix of the monitoring data of the rth band; r,w It represents the interference degree of the w-th signal feature window in the characteristic signal matrix of the r-th band monitoring data; || represents the absolute value; ε represents the preset hyperparameter.

[0026] Preferably, the method of clearing and screening all signal feature windows according to the noise area evaluation factor to obtain the denoised feature signal matrix of each band monitoring data includes the following specific methods:

[0027] A noise threshold t is preset. If the noise area evaluation factor of the w-th signal feature window in the characteristic signal matrix of the r-th band monitoring data is greater than or equal to the noise threshold T, the w-th signal feature window is recorded as the noise window; after all noise windows in the characteristic signal matrix of the r-th band monitoring data are removed, the result is recorded as the denoised characteristic signal matrix of the r-th band monitoring data.

[0028] Preferably, the method of obtaining the regional status of the geographical location corresponding to each element in the denoised characteristic signal matrix according to the denoised characteristic signal matrix includes the following specific methods:

[0029] The denoised characteristic signal matrix of each band monitoring data is recorded as the signal matrix to be detected. After the neural network training is completed, all signal feature windows in the signal matrix to be detected are input into the trained neural network to obtain the categories of all signal feature windows in the signal matrix to be detected.

[0030] For any signal feature window, if the output is a normal signal window category, the geographical location area corresponding to all elements in the signal feature window is a normal area; if the output is a mildly abnormal signal window category, the geographical location area corresponding to all elements in the signal feature window is a mildly abnormal area; if the output is a moderately abnormal signal window category, the geographical location area corresponding to all elements in the signal feature window is a moderately abnormal area; if the output is a severely abnormal signal window category, the geographical location area corresponding to all elements in the signal feature window is a severely abnormal area.

[0031] The present invention also proposes a state intelligent perception system based on synthetic aperture radar multi-source data, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the system implements any one of the steps of the state intelligent perception method based on synthetic aperture radar multi-source data.

[0032] The beneficial effects of the technical solution of the present invention are as follows: the present invention screens all signal feature windows according to the noise area evaluation factor and removes the noise window to obtain the denoised feature signal matrix of each band monitoring data; according to the denoised feature signal matrix, the regional state of the geographical location corresponding to each element in the denoised feature signal matrix is ​​obtained, so that the noise window is adaptively removed based on the difference of several band monitoring data of SAR and the signal significance in the signal feature window, thereby avoiding the interference of noise data on the identification of key target areas in the radar irradiation area, thereby improving the accuracy of the perception of the geographical information status in the radar irradiation area. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in 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 only 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.

[0034] Figure 1 It is a flow chart of the steps of the state intelligent perception method based on synthetic aperture radar multi-source data of the present invention;

[0035] Figure 2 Flowchart for obtaining the denoised feature signal matrix. DETAILED DESCRIPTION

[0036] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the state intelligent perception method and system based on synthetic aperture radar multi-source data proposed by the present invention, its specific implementation method, structure, features and effects, in conjunction with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0037] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0038] The specific scheme of the state intelligent perception method and system based on synthetic aperture radar multi-source data provided by the present invention is described in detail below with reference to the accompanying drawings.

[0039] See also Figure 1 , which shows a flowchart of a state intelligent perception method based on synthetic aperture radar multi-source data provided by an embodiment of the present invention, the method comprising the following steps:

[0040] Step S001: Acquire monitoring data of several bands of SAR.

[0041] It should be noted that SAR's multi-source data can provide more comprehensive and multi-angle information in synthetic aperture radar, thereby enhancing data processing and target recognition capabilities.

[0042] Specifically, we first need to collect monitoring data of several SAR bands. The specific process is as follows:

[0043] The SAR equipment is installed in the preselected collection area. The antenna orientation, beam width and working mode are installed and set according to the specifications of the SAR equipment and the instruction manual of the SAR equipment, so as to obtain several band monitoring data of the SAR to be processed. The several band monitoring data of the SAR to be processed are preprocessed through the correction of the pulse repetition frequency and the conversion of the geographic coordinate system to obtain several band monitoring data of the SAR.

[0044] It should be noted that, since the monitoring is carried out relative to the same pre-selected collection area, the matrix size of the monitoring data of each band is the same, and the geographical target position corresponding to each element is also the same; each band monitoring data corresponds to a data matrix, and each element corresponds to a geographical location.

[0045] So far, several band monitoring data of SAR are obtained through the above method.

[0046] Step S002: Obtain the interference degree of the characteristic signal matrix of each band monitoring data.

[0047] A parameter k is preset, wherein this embodiment is described by taking K=1500 as an example, and this embodiment is not specifically limited, wherein K is determined according to specific implementation conditions.

[0048] Specifically, for the r-th band monitoring data of SAR, the data matrix of the r-th band monitoring data is input into the convolutional layer of the CNN (Convolutional Neural Networks) neural network, and the characteristic signal matrix of the r-th band monitoring data is output. K windows are selected from the characteristic signal matrix of the r-th band monitoring data through a selective search algorithm, and the set of K windows is used as a signal feature window set of the characteristic signal matrix of the r-th band monitoring data.

[0049] Among them, the process of outputting a matrix through a CNN neural grid, namely, a feature information matrix, is well-known content, and the selective search algorithm is a well-known technology, which will not be described in detail in this embodiment.

[0050] It should be noted that the more uneven the signal distribution in the signal feature window is, the more locations with obvious data changes exist in the characteristic signal matrix of the band monitoring data. Therefore, the fluctuation amplitude of the signal value of each element in the characteristic signal matrix of the band monitoring data and the frequency of occurrence of the signal value at the location with obvious fluctuation are analyzed to determine the extreme value signal situation in the characteristic signal matrix of the rth band monitoring data, so as to facilitate the subsequent comprehensive analysis of the extreme value signal differences of multiple bands to analyze the noise interference situation of the current band.

[0051] Specifically, the interference degree of the characteristic signal matrix of each band monitoring data is obtained according to the fluctuation of each element in the characteristic signal matrix of each band monitoring data.

[0052] As an example, the calculation method for obtaining the interference degree of the characteristic signal matrix of each band monitoring data is as follows:

[0053]

[0054] Where A represents the interference degree of the characteristic signal matrix of any band monitoring data; n represents the total number of all elements in the characteristic signal matrix of the band monitoring data; t i Represents the i-th element in the characteristic signal matrix of the band monitoring data; represents the average value of all elements in the characteristic signal matrix of the band monitoring data; σ represents the standard deviation of all elements in the characteristic signal matrix of the band monitoring data.

[0055] It should be noted that the larger the interference degree value is, the easier it is for the current band monitoring data represented by the characteristic signal matrix of the current band monitoring data to collect detailed information within the monitoring range, and therefore it is easier to produce an obvious signal position in the characteristic signal matrix of the current band monitoring data. Therefore, the signal difference in the characteristic signal matrix of the current band monitoring data is more obvious. Therefore, when screening for signal value peak decay at obvious signal positions in the characteristic signal matrix of the current band monitoring data, the signal position screening criteria in the characteristic signal matrix of the current band monitoring data should be more relaxed to avoid incorrect identification of the noise signal value position.

[0056] So far, the interference degree of the characteristic signal matrix for obtaining monitoring data of each band is obtained through the above method.

[0057] Step S003: Obtain the denoised characteristic signal matrix of each band monitoring data according to the noise area evaluation degree of each signal characteristic window in the characteristic signal matrix of each band monitoring data.

[0058] It should be noted that since the selective search algorithm will select the elements of the obvious signal position in the matrix and the elements around it, the information value in the current window is judged by the high-value geographic information position within the current detection range represented by the elements of the obvious signal position. When only the obvious signal position appears in the window, it means that the geographic information characteristics within the monitoring range represented by the current obvious signal position are reflected by only one element position, which means that the current obvious signal element position is more likely to be a noise position. Therefore, a comprehensive judgment is made based on the distribution characteristics of the obvious signal position in the current window and the attenuation difference between the signal value of the obvious signal position and the surrounding signal values.

[0059] Specifically, for the w-th signal feature window of the characteristic signal matrix of the r-th band monitoring data, all elements in the w-th signal feature window are traversed row by row, and all elements are extracted in sequence to form a first sequence, all elements in the w-th signal feature window are traversed column by column, and all elements are extracted in sequence to form a second sequence, the first sequence and the second sequence are merged and used as the feature sequence of the w-th signal feature window; the peak detection algorithm is used to obtain all maximum elements of the feature sequence of the w-th signal feature window, and all maximum elements are used as obvious signal elements of the w-th signal feature window.

[0060] For any obvious signal element in the w-th signal feature window, an adjustable window is constructed with the obvious signal element as the center point. In this embodiment, the initial window side length of the adjustable window is 1, and the adjustable window side length is iteratively increased with a step size of 1 until the adjustable window contacts other obvious signal elements of the w-th signal feature window or the edge of the w-th signal feature window during the increase process. Then, the iterative increase is stopped, and the adjustable window after each increase of the obvious signal element and the total number of iterative increases of the adjustable window are obtained.

[0061] According to the interference degree of the characteristic signal matrix of each band monitoring data and the obvious signal elements of each signal characteristic window, the noise area evaluation factor of each signal characteristic window in the characteristic signal matrix of each band monitoring data is obtained.

[0062] As an example, the calculation method for obtaining the noise area evaluation degree of each signal feature window in the characteristic signal matrix of each band monitoring data is as follows:

[0063]

[0064] In the formula, B r,w Indicates the noise area evaluation degree of the wth signal feature window in the characteristic signal matrix of the rth band monitoring data; n r Represents the total number of all elements in the characteristic signal matrix of the r-th band monitoring data; n r,w represents the total number of all elements in the wth signal feature window in the characteristic signal matrix of the rth band monitoring data; m r,w Represents the total number of all obvious signal elements in the w-th signal feature window in the characteristic signal matrix of the r-th band monitoring data; N represents the average value of the Euclidean distance between all obvious signal elements in the w-th signal feature window in the characteristic signal matrix of the r-th band monitoring data; r,w,q The total number of iterative increases of the adjustable window of the qth obvious signal element of the wth signal feature window in the characteristic signal matrix of the rth band monitoring data; It represents the average value of all elements in the adjustable window after the jth increase of the qth obvious signal element in the wth signal feature window in the characteristic signal matrix of the rth band monitoring data; A represents the average value of all elements in the adjustable window after the qth obvious signal element in the wth signal feature window in the characteristic signal matrix of the rth band monitoring data is increased for the j-1th time; r A represents the interference degree of the characteristic signal matrix of the monitoring data of the rth band; r,wIt represents the interference degree of the w-th signal feature window in the characteristic signal matrix of the r-th band monitoring data; || represents the absolute value; ε represents the preset hyperparameter, and this implementation presets ε=1 to prevent the denominator from being 0.

[0065] It should be noted that the method for obtaining the interference degree of the signal characteristic window refers to the method for obtaining the interference degree of the characteristic signal matrix of the reference band monitoring data.

[0066] It represents the size ratio of the w-th signal feature window. The larger the value is, the smaller the element range is. Therefore, it is more clear that the obvious signal position of the w-th signal feature window represents the more outlier data points, which contains less data information within the monitoring range. It is more unfavorable to distinguish the geographic information characteristics within the current monitoring range through the data information in the w-th signal feature window, and it is more of a noise.

[0067] The larger it is, the denser the distribution of obvious signal elements in the w-th signal feature window is, which means that the w-th signal feature window may contain continuous geographic feature information. Therefore, it can form continuously distributed obvious signal elements, which is more conducive to identifying the geographic information features in the current window, and the possibility of being noise is smaller.

[0068] It indicates the signal attenuation degree within each step range of the obvious signal element in the w-th signal feature window. The larger the value is, the higher the frequency of the obvious signal element in the w-th signal feature window. At the same time, with the iterative increase of a single obvious signal element, the rapid attenuation change of the element is more reflected, which further reflects that the obvious signal element output by the convolution layer represents the possibility of an isolated noise position in the w-th signal feature window, and the larger the noise window is.

[0069] Furthermore, the noise area evaluation degrees of all signal feature windows in the characteristic signal matrix of the r-th band monitoring data are obtained, and each noise area evaluation degree after linear normalization of all noise area evaluation degrees is recorded as a noise area evaluation factor.

[0070] A noise threshold T is preset, wherein this embodiment is described by taking T=0.85 as an example, and this embodiment is not specifically limited, wherein T is determined according to specific implementation conditions.

[0071] If the noise area evaluation factor of the w-th signal feature window in the characteristic signal matrix of the r-th band monitoring data is greater than or equal to the noise threshold T, the w-th signal feature window is recorded as the noise window; after removing all noise windows in the characteristic signal matrix of the r-th band monitoring data, it is recorded as the denoised characteristic signal matrix of the r-th band monitoring data, and then the denoised characteristic signal matrix of each band monitoring data is obtained.

[0072] So far, the denoised characteristic signal matrix of each band monitoring data is obtained through the above method; the specific process of obtaining the denoised characteristic signal matrix is ​​as follows Figure 2 .

[0073] Step S004: According to the denoised characteristic signal matrix, the regional status of the geographical location corresponding to each element in the denoised characteristic signal matrix is ​​obtained.

[0074] Specifically, the denoised characteristic signal matrix of each band monitoring data is recorded as the signal matrix to be detected; a large number of signal feature windows in the denoised characteristic signal matrix of the band monitoring data of normal collection areas are obtained and recorded as normal signal windows; a large number of signal feature windows in the denoised characteristic signal matrix of the band monitoring data of mildly abnormal collection areas are obtained and recorded as mildly abnormal signal windows; a large number of signal feature windows in the denoised characteristic signal matrix of the band monitoring data of moderately abnormal collection areas are obtained and recorded as moderately abnormal signal windows; a large number of signal feature windows in the denoised characteristic signal matrix of the band monitoring data of severely abnormal collection areas are obtained and recorded as severely abnormal signal windows; a large number of normal signal windows, mildly abnormal signal windows, moderately abnormal signal windows and severely abnormal signal windows are taken as data sets, and the category information in the data sets is manually annotated.

[0075] Among them, the normal signal window category is marked as 0, the mild abnormal signal window category is 1, the moderate abnormal signal window category is 2, and the severe abnormal signal window category is 3. The cross entropy loss function is used to supervise the training. After the neural network training is completed, all signal feature windows in the signal matrix to be detected are input into the fully connected layer of the trained neural network to obtain the categories of all signal feature windows in the signal matrix to be detected.

[0076] For any signal feature window, if the output is a normal signal window category, the geographical location area corresponding to all elements in the signal feature window is a normal area; if the output is a mildly abnormal signal window category, the geographical location area corresponding to all elements in the signal feature window is a mildly abnormal area; if the output is a moderately abnormal signal window category, the geographical location area corresponding to all elements in the signal feature window is a moderately abnormal area; if the output is a severely abnormal signal window category, the geographical location area corresponding to all elements in the signal feature window is a severely abnormal area.

[0077] Among them, the CNN neural network is used in this embodiment, and other neural network models can be used in other embodiments, which is not limited in this embodiment.

[0078] Through the above steps, the state intelligent perception method based on synthetic aperture radar multi-source data is completed.

[0079] The present invention also proposes a state intelligent perception system based on synthetic aperture radar multi-source data, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the state intelligent perception method based on synthetic aperture radar multi-source data in steps S001 to S004 are implemented.

[0080] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A state intelligent perception method based on synthetic aperture radar multi-source data, characterized in that: The method comprises the following steps: Acquire several band monitoring data of SAR; each band monitoring data corresponds to a data matrix, the data matrix contains multiple elements, and each element corresponds to a geographical location; According to the data matrix of each band monitoring data, the characteristic signal matrix of each band monitoring data is obtained, and all signal characteristic windows of the characteristic signal matrix of each band monitoring data are obtained; according to the fluctuation of each element in the characteristic signal matrix of each band monitoring data, the interference degree of the characteristic signal matrix of each band monitoring data is obtained; Obtain the obvious signal elements of each signal feature window in the characteristic signal matrix of each band monitoring data; obtain the noise area evaluation factor of each signal feature window according to the interference degree and the obvious signal elements of each signal feature window; clear and filter all signal feature windows according to the noise area evaluation factor to obtain the denoised characteristic signal matrix of each band monitoring data; According to the denoised feature signal matrix, the regional status of the geographical location corresponding to each element in the denoised feature signal matrix is ​​obtained.

2. The state intelligent perception method based on synthetic aperture radar multi-source data according to claim 1 is characterized in that: The specific method of obtaining the characteristic signal matrix of each band monitoring data according to the data matrix of each band monitoring data and obtaining all signal characteristic windows of the characteristic signal matrix of each band monitoring data is as follows: A parameter K is preset. For the r-th band monitoring data of SAR, the data matrix of the r-th band monitoring data is input into the convolution layer of the CNN neural network, and the characteristic signal matrix of the r-th band monitoring data is output. K windows are selected from the characteristic signal matrix of the r-th band monitoring data through a selective search algorithm, and the set composed of K windows is used as the signal feature window set of the characteristic signal matrix of the r-th band monitoring data.

3. The state intelligent perception method based on synthetic aperture radar multi-source data according to claim 1 is characterized in that: The specific formula for obtaining the interference degree of the characteristic signal matrix of each band monitoring data according to the fluctuation of each element in the characteristic signal matrix of each band monitoring data is: Where A represents the interference degree of the characteristic signal matrix of any band monitoring data; n represents the total number of all elements in the characteristic signal matrix of the band monitoring data; t i Represents the i-th element in the characteristic signal matrix of the band monitoring data; represents the average value of all elements in the characteristic signal matrix of the band monitoring data; ρ represents the standard deviation of all elements in the characteristic signal matrix of the band monitoring data.

4. The state intelligent perception method based on synthetic aperture radar multi-source data according to claim 1 is characterized in that: The specific method of obtaining the obvious signal elements of each signal feature window in the characteristic signal matrix of each band monitoring data includes: For the w-th signal feature window of the characteristic signal matrix of the r-th band monitoring data, all elements in the w-th signal feature window are traversed row by row, and all elements are extracted in sequence to form a first sequence, all elements in the w-th signal feature window are traversed column by column, and all elements are extracted in sequence to form a second sequence, the first sequence and the second sequence are merged and used as the feature sequence of the w-th signal feature window; the peak detection algorithm is used to obtain all maximum elements of the feature sequence of the w-th signal feature window, and all maximum elements are used as obvious signal elements of the w-th signal feature window.

5. The state intelligent perception method based on synthetic aperture radar multi-source data according to claim 1 is characterized in that: The specific method of obtaining the noise area evaluation factor of each signal feature window according to the interference degree and the obvious signal elements of each signal feature window is as follows: For any obvious signal element in the w-th signal feature window, obtain the adjustable window after each increase of the obvious signal element and the total number of iterative increases of the adjustable window; Obtain the noise area evaluation degree of each signal feature window according to the interference degree and the adjustable window after each increase of the obvious signal elements in each signal feature window and the total number of iterative increases of the adjustable window; The noise region evaluation degrees of all signal feature windows are obtained, and each noise region evaluation degree after linear normalization of all noise region evaluation degrees is recorded as a noise region evaluation factor.

6. The state intelligent perception method based on synthetic aperture radar multi-source data according to claim 5 is characterized in that: The specific method of obtaining the adjustable window after each increase of the obvious signal element and the total number of iterative increases of the adjustable window includes: An adjustable window is constructed with the obvious signal element as the center point, the initial window side length of the adjustable window is preset to 1, and the side length of the adjustable window is iteratively increased with a step size of 1 until the adjustable window contacts other obvious signal elements of the w-th signal feature window or the edge of the w-th signal feature window during the increasing process, then the iterative increase is stopped, and the adjustable window after each increase of the obvious signal element and the total number of iterative increases of the adjustable window are obtained.

7. The state intelligent perception method based on synthetic aperture radar multi-source data according to claim 5 is characterized in that: The specific formula for obtaining the noise area evaluation degree of each signal feature window according to the interference degree and the adjustable window after each increase of the obvious signal elements in each signal feature window and the total number of iterative increases of the adjustable window is: In the formula, B r,w Indicates the noise area evaluation degree of the wth signal feature window in the characteristic signal matrix of the rth band monitoring data; n r Represents the total number of all elements in the characteristic signal matrix of the r-th band monitoring data; n r,w represents the total number of all elements in the wth signal feature window in the characteristic signal matrix of the rth band monitoring data; m r,w Represents the total number of all obvious signal elements in the w-th signal feature window in the characteristic signal matrix of the r-th band monitoring data; N represents the average value of the Euclidean distance between all obvious signal elements in the w-th signal feature window in the characteristic signal matrix of the r-th band monitoring data; r,w,q The total number of iterative increases of the adjustable window of the qth obvious signal element of the wth signal feature window in the characteristic signal matrix of the rth band monitoring data; It represents the average value of all elements in the adjustable window after the jth increase of the qth obvious signal element in the wth signal feature window in the characteristic signal matrix of the rth band monitoring data; A represents the average value of all elements in the adjustable window after the qth obvious signal element in the wth signal feature window in the characteristic signal matrix of the rth band monitoring data is increased for the j-1th time; r A represents the interference degree of the characteristic signal matrix of the monitoring data of the rth band; r,w It represents the interference degree of the w-th signal feature window in the characteristic signal matrix of the r-th band monitoring data; | | represents the absolute value; ε represents the preset hyperparameter.

8. The state intelligent perception method based on synthetic aperture radar multi-source data according to claim 1 is characterized in that: The method of clearing and screening all signal feature windows according to the noise area evaluation factor to obtain the denoised feature signal matrix of each band monitoring data includes the following specific methods: A noise threshold T is preset. If the noise area evaluation factor of the wth signal feature window in the characteristic signal matrix of the rth band monitoring data is greater than or equal to the noise threshold T, the wth signal feature window is recorded as the noise window; the rth signal feature window is recorded as the noise window. After all noise windows are removed from the characteristic signal matrix of the monitoring data of the rth band, it is recorded as the denoised characteristic signal matrix of the monitoring data of the rth band.

9. The state intelligent perception method based on synthetic aperture radar multi-source data according to claim 1 is characterized in that: The specific method of obtaining the regional status of the geographical location corresponding to each element in the denoised characteristic signal matrix according to the denoised characteristic signal matrix includes: The denoised characteristic signal matrix of each band monitoring data is recorded as the signal matrix to be detected. After the neural network training is completed, all signal feature windows in the signal matrix to be detected are input into the trained neural network to obtain the categories of all signal feature windows in the signal matrix to be detected. For any signal feature window, if the output is a normal signal window category, the geographical location area corresponding to all elements in the signal feature window is a normal area; If the output is a mild abnormal signal window category, the geographical location area corresponding to all elements in the signal feature window is a mild abnormal area; if the output is a moderate abnormal signal window category, the geographical location area corresponding to all elements in the signal feature window is a moderate abnormal area; if the output is a severe abnormal signal window category, the geographical location area corresponding to all elements in the signal feature window is a severe abnormal area.

10. A state intelligent perception system based on synthetic aperture radar multi-source data, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the state intelligent perception method based on synthetic aperture radar multi-source data as described in any one of claims 1 to 9 are implemented.

Images