Time-frequency matrix denoising method based on local adaptive threshold algorithm and iterative algorithm

Abstract

The present invention discloses a time-frequency matrix denoising method based on a local adaptive threshold algorithm and an iterative algorithm, comprising the steps of acquiring an image and performing a short-time Fourier transform to obtain a time-frequency matrix and then calculating a signal-to-noise ratio; when the signal-to-noise ratio is greater than a preset value, calculating an initial value of an adaptive factor and using it as the current adaptive factor, calculating a current local adaptive threshold, using the initial value of the adaptive factor to perform denoising on the time-frequency matrix, calculating the sum of the elements in the current time-frequency matrix, and determining the corresponding current time-frequency matrix as the time-frequency matrix after denoising if a first preset condition is met; when the signal-to-noise ratio is less than a preset value, calculating an initial value of a noise threshold and using it as the current noise threshold, updating the current noise threshold based on elements in the time-frequency matrix that are greater than or less than the current noise threshold, and performing denoising on the time-frequency matrix using the updated noise threshold if a second preset condition is met. The present invention can effectively remove noise and achieve good denoising effect.

Description

Technical Field

[0001] The present invention belongs to the technical field of unmanned aerial vehicles, and in particular relates to a time-frequency matrix denoising method based on a local adaptive threshold algorithm and an iterative algorithm. Background Art

[0002] Because drones do not require pilots and are inexpensive, easy to operate, small in size, and convenient for performing different tasks, they are widely used in military and civilian fields, such as catering services, aerial photography, news reporting, and environmental mapping.

[0003] "Illegal flying" refers to flights conducted without a private pilot's license or without legally registered aircraft. While relevant drone management policies have been gradually introduced, addressing this issue still requires the support of relevant detection and interception technologies. The research and application of drone detection, tracking, and jamming technologies can effectively detect illegal drones, severing their communication links and preventing them from continuing remote control flight. This has important practical implications for protecting personal privacy and public safety.

[0004] Obtaining clear time-frequency maps is the first step in processing UAV remote sensing signals. However, due to the limitations of time-frequency resolution and noise, obtaining clear time-frequency maps in low signal-to-noise ratio (SNR) conditions is challenging. Therefore, at the outset of processing, a certain degree of preprocessing of the received signal is required to remove noise and obtain a relatively clear time-frequency map. Prior art methods for denoising time-frequency maps include the large-scale, small-scale removal method, logarithmic statistics, and morphological denoising. However, these methods have limitations and drawbacks. Specifically, the large-scale, small-scale removal method generally achieves good results after a secondary selection process, with relatively low computational effort. However, in complex channel environments, multiple selection processes are required, increasing the computational effort. Furthermore, the global distribution of noise in the time-frequency space is completely different from the local distribution of the signal. While the logarithmic statistics method can leverage the amplitude characteristics of both the noise and the signal, it does not fully exploit the local characteristics of the signal. Summary of the Invention

[0005] In order to solve the above problems existing in the prior art, the present invention provides a time-frequency matrix denoising method based on a local adaptive threshold algorithm and an iterative algorithm. The technical problem to be solved by the present invention is achieved through the following technical solutions:

[0006] The present invention provides a time-frequency matrix denoising method based on a local adaptive threshold algorithm and an iterative algorithm, comprising:

[0007] Obtain the image collected by the drone and perform short-time Fourier transform to obtain the time-frequency matrix;

[0008] Calculating the signal-to-noise ratio of the time-frequency matrix according to the average value of all elements in the time-frequency matrix;

[0009] When the signal-to-noise ratio is greater than a preset value, the initial value of the adaptive factor is calculated, and the initial value of the adaptive factor is used as the current adaptive factor. The current local adaptive threshold is calculated according to the current adaptive factor, and the time-frequency matrix is further denoised using the current local adaptive threshold. The sum of the elements in the current time-frequency matrix is calculated. k ;

[0010] If the sum of the elements s k If the first precondition is met, the sum of the elements s k The corresponding current time-frequency matrix is determined as the time-frequency matrix after denoising;

[0011] When the signal-to-noise ratio is less than a preset value, the initial value of the noise threshold is calculated, the initial value of the noise threshold is used as the current noise threshold, and the number of elements N in the time-frequency matrix that are greater than the current noise threshold is counted with the current noise threshold as the boundary. s And the number of elements N that are smaller than the current noise threshold n , further updating the current noise threshold according to elements greater than the current noise threshold and elements less than the current noise threshold;

[0012] If the updated noise threshold ε k If the second preset condition is met, the updated noise threshold ε is used. k Performing denoising processing on the time-frequency matrix.

[0013] In one embodiment of the present invention, when the signal-to-noise ratio is greater than a preset value, the initial value of the adaptive factor is calculated, the initial value of the adaptive factor is used as the current adaptive factor, the current local adaptive threshold is calculated according to the current adaptive factor, and the time-frequency matrix is further denoised using the current local adaptive threshold, and the sum of the elements in the current time-frequency matrix is calculated. k The steps include:

[0014] Determine the maximum value M of the elements in the time-frequency matrix and calculate the average value m of all elements;

[0015] Calculating an initial value of an adaptive factor using the maximum value M and the average value m;

[0016] Taking the initial value of the adaptive factor as the current adaptive factor, and calculating the current local adaptive threshold according to the current adaptive factor;

[0017] Performing denoising on the time-frequency matrix using the current local adaptive threshold, and setting elements in the time-frequency matrix that are smaller than the current local adaptive threshold to 0;

[0018] Calculate the sum s of the elements in the current time-frequency matrix after denoising k , k represents the current iteration number.

[0019] In one embodiment of the present invention, if the sum of the elements s k If the first precondition is met, the sum of the elements s k Before the step of determining the corresponding current time-frequency matrix as the time-frequency matrix after denoising, the method further includes:

[0020] Detection k -s k-1 Is it greater than s k-1 -s k-2 ;

[0021] The detection k -s k-1 Is it greater than s k-1 -s k-2 After the steps, it also includes:

[0022] If s k -s k-1 Less than s k-1 -s k-2 , then adjust the current adaptive factor according to the preset step size, and return to the step of calculating the current local adaptive threshold according to the current adaptive factor.

[0023] In one embodiment of the present invention, the step of calculating the initial value of the adaptive factor using the maximum value M and the average value m includes:

[0024] The ratio of the average value m to the maximum value M is used as the initial value of the adaptive factor.

[0025] In one embodiment of the present invention, the step of using the initial value of the adaptive factor as the current adaptive factor and calculating the current local adaptive threshold according to the current adaptive factor includes:

[0026] Using the initial value of the adaptive factor as the current adaptive factor;

[0027] Select the maximum and minimum values of the elements in the i-th row of the time-frequency matrix and calculate the sum of the two T 0,i ;

[0028] The sum of the two T 0,i Multiply it with the current adaptive factor to get the current local adaptive threshold of the i-th row of the time-frequency matrix.

[0029] In one embodiment of the present invention, when the signal-to-noise ratio is less than a preset value, the initial value of the noise threshold is calculated, the initial value of the noise threshold is used as the current noise threshold, and the number of elements N in the time-frequency matrix that are greater than the current noise threshold is counted with the current noise threshold as the boundary. s And the number of elements N that are smaller than the current noise threshold n , further updating the current noise threshold according to elements greater than the current noise threshold and elements less than the current noise threshold, comprising:

[0030] Determine the maximum value M of the elements in the time-frequency matrix and calculate the average value m of all elements;

[0031] Calculating an initial value of a noise threshold using the maximum value M and the average value m;

[0032] The initial value of the noise threshold is used as the current noise threshold, and the number of elements N in the time-frequency matrix that are greater than the current noise threshold is counted with the current noise threshold as the boundary. s And the number of elements N that are smaller than the current noise threshold n ;

[0033] respectively calculating a first mean value of elements greater than a current noise threshold and a second mean value of elements less than the current noise threshold in the time-frequency matrix;

[0034] Calculate the average of the first mean and the second mean to obtain the updated noise threshold ε k .

[0035] In one embodiment of the present invention, the step of calculating the initial value of the noise threshold using the maximum value M and the average value m includes:

[0036] The average of the maximum value M and the average value m is used as the initial value of the noise threshold.

[0037] In one embodiment of the present invention, if the updated noise threshold ε k If the second preset condition is met, the updated noise threshold ε is used. k Before the step of performing denoising on the time-frequency matrix, the method further includes:

[0038] Detection of ε k Is it equal to ε k-1 ;

[0039] The detection ε k Is it equal to ε k-1 After the steps, it also includes:

[0040] If ε k Not equal to ε k-1 , the updated noise threshold εk As the current noise threshold, and return to the current noise threshold as the boundary, counting the number of elements N in the time-frequency matrix that are greater than the current noise threshold s And the number of elements N that are smaller than the current noise threshold n steps.

[0041] In one embodiment of the present invention, the preset value is 10.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] The present invention provides a time-frequency matrix denoising method based on a local adaptive threshold algorithm and an iterative algorithm. On the one hand, when the signal-to-noise ratio of the time-frequency graph is high, considering that the amplitude difference between the signal and noise is large and easy to distinguish, and the signal after short-time Fourier transform is concentrated in a fixed time-frequency domain, while the noise is distributed in the entire time-frequency plane, the above-mentioned difference can be utilized to eliminate noise by finding a suitable local adaptive threshold. On the other hand, in the case of low signal-to-noise ratio, the time-frequency matrix is divided into a signal part and a noise part based on a noise threshold, and then the noise threshold is updated according to the elements of the two parts until the noise threshold no longer changes, and then the time-frequency graph is denoised using the noise threshold. The present invention uses different algorithms to process time-frequency graphs with different signal-to-noise ratios, which can effectively eliminate noise and obtain good denoising effect.

[0044] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a flow chart of a time-frequency matrix denoising method based on a local adaptive threshold algorithm and an iterative algorithm provided by an embodiment of the present invention;

[0046] Figure 2 is a schematic diagram of a time-frequency matrix provided by an embodiment of the present invention;

[0047] Figure 3a This is a time-frequency diagram under low signal-to-noise ratio conditions provided by an embodiment of the present invention;

[0048] Figure 3b The embodiment of the present invention provides Figure 3a Time-frequency diagram after denoising;

[0049] Figure 4a This is a time-frequency diagram under a high signal-to-noise ratio condition provided by an embodiment of the present invention;

[0050] Figure 4b The embodiment of the present invention provides Figure 4a Time-frequency diagram after denoising;

[0051] Figure 5a is another time-frequency diagram provided by an embodiment of the present invention;

[0052] Figure 5b The embodiment of the present invention provides Figure 5a 3D time-frequency diagram of

[0053] Figure 5c The embodiment of the present invention provides Figure 5a Time-frequency diagram after denoising;

[0054] Figure 5d The embodiment of the present invention provides Figure 5a Three-dimensional time-frequency diagram after denoising. DETAILED DESCRIPTION

[0055] The present invention will be further described in detail below with reference to specific examples, but the embodiments of the present invention are not limited thereto.

[0056] Figure 1 This is a flow chart of a time-frequency matrix denoising method based on a local adaptive threshold algorithm and an iterative algorithm provided by an embodiment of the present invention. Figure 1 As shown, an embodiment of the present invention provides a time-frequency matrix denoising method based on a local adaptive threshold algorithm and an iterative algorithm, comprising:

[0057] S1, obtain the image collected by the UAV and perform short-time Fourier transform to obtain the time-frequency matrix;

[0058] S2, calculating the signal-to-noise ratio of the time-frequency matrix based on the average value of all elements in the time-frequency matrix;

[0059] S3. When the signal-to-noise ratio is greater than the preset value, the initial value of the adaptive factor is calculated, and the initial value of the adaptive factor is used as the current adaptive factor. The current local adaptive threshold is calculated based on the current adaptive factor, and the current local adaptive threshold is further used to denoise the time-frequency matrix, and the sum of the elements in the current time-frequency matrix is calculated. k ;

[0060] S4. If the sum of the elements s k If the first precondition is met, the sum of the elements s k The corresponding current time-frequency matrix is determined as the time-frequency matrix after denoising;

[0061] S5. When the signal-to-noise ratio is less than the preset value, calculate the initial value of the noise threshold, use the initial value of the noise threshold as the current noise threshold, and use the current noise threshold as the boundary to count the number of elements N in the time-frequency matrix that are greater than the current noise threshold. s And the number of elements N that are smaller than the current noise threshold n, further updating the current noise threshold according to the elements greater than the current noise threshold and the elements less than the current noise threshold;

[0062] S6, if the updated noise threshold ε k If the second preset condition is met, the updated noise threshold ε is used. k Perform denoising on the time-frequency matrix.

[0063] In this embodiment, a short-time Fourier transform (STFT) is first performed on the image collected by the drone, and then the signal-to-noise ratio of the obtained time-frequency graph is calculated. Different methods are used to denoise the time-frequency graph according to the signal-to-noise ratio.

[0064] Specifically, when the signal-to-noise ratio is greater than a preset value, it indicates that the signal-to-noise ratio of the time-frequency graph is high. A local adaptive threshold can be set to divide the time-frequency matrix into two parts: noise and signal. Figure 2 is a schematic diagram of the time-frequency matrix provided by an embodiment of the present invention, such as Figure 2 As shown in the figure, the greater the difference in amplitude between the noise part and the signal part, the greater the difference between the noise and the signal, thus achieving noise and signal segmentation. It should be understood that since the signal is concentrated in a fixed time-frequency domain after the short-time Fourier transform, while noise such as Gaussian white noise is distributed in the entire time-frequency plane after the short-time Fourier transform, it is necessary to fully utilize the difference in time-frequency aggregation between the frequency hopping signal and Gaussian white noise after the short-time Fourier transform in the case of high signal-to-noise ratio, and eliminate the Gaussian white noise by setting a local adaptive threshold.

[0065] When the signal-to-noise ratio is low, the distance between the receiver and the transmitter is relatively far and the ambient noise is also relatively large. Therefore, this embodiment divides the time-frequency matrix into two parts by setting the initial value of the noise threshold, one part is the signal part and the other part is the noise part. The noise threshold is then updated for the signal part to perform denoising. This process is repeated until the noise threshold no longer changes, which is the final noise threshold used for denoising.

[0066] It should be noted that, in the above-mentioned time-frequency matrix denoising method, the preset value for judging the signal-to-noise ratio can be flexibly set according to actual needs. In this embodiment, the preset value is set to 10.

[0067] Optionally, when the signal-to-noise ratio is greater than a preset value, the initial value of the adaptive factor is calculated in step S2, and the initial value of the adaptive factor is used as the current adaptive factor. The current local adaptive threshold is calculated according to the current adaptive factor, and the time-frequency matrix is further denoised using the current local adaptive threshold. The sum of the elements in the current time-frequency matrix is calculated as s k The steps include:

[0068] S201, determining the maximum value M of the elements in the time-frequency matrix and calculating the average value m of all elements;

[0069] S202, calculating the initial value of the adaptive factor using the maximum value M and the average value m;

[0070] S203, taking the initial value of the adaptive factor as the current adaptive factor, and calculating the current local adaptive threshold according to the current adaptive factor;

[0071] S204, using the current local adaptive threshold to perform denoising on the time-frequency matrix, and setting the elements in the time-frequency matrix that are smaller than the current local adaptive threshold to 0;

[0072] S205, calculating the sum s of the elements in the current time-frequency matrix after denoising k , k represents the current iteration number.

[0073] Specifically, the elements with larger amplitudes in the time-frequency matrix are usually located in the time-frequency domain corresponding to the signal. Therefore, an initial value can be set for the adaptive factor μ. In this embodiment, the ratio of the average value m of each element in the time-frequency matrix to the maximum value M of the element is used as the initial value of the adaptive factor. The initial value of the adaptive factor is then used as the current adaptive factor. The maximum and minimum values of the elements in the i-th row of the time-frequency matrix are then selected and their sum T is calculated. 0,i , further add the sum of the two T 0,i Multiply it with the current adaptive factor to get the current local adaptive threshold of the i-th row of the time-frequency matrix.

[0074] For example, the formula is as follows:

[0075] Threshold k =μ k T 0,i

[0076] Among them, T 0,i ={max[STFT(i,·)]+min[STFT(i,·)]},μ k Indicates the current adaptive factor of the kth iteration, Threshold k represents the current local adaptive threshold of the kth iteration, max[STFT(i,·)] and min[STFT(i,·)] represent the maximum and minimum values of the elements in the i-th row of the time-frequency matrix, respectively.

[0077] Furthermore, in the above steps S204 to S205, the current local adaptive threshold Threshold is used. k Denoise the time-frequency matrix, that is, remove the noise in the time-frequency matrix that is less than Threshold kThe elements are set to 0 and greater than or equal to Threshold k The elements of are not processed, and then the sum of the elements in the current time-frequency matrix after denoising is calculated. k .

[0078] In the above step S4, if the sum of the elements s k If the first precondition is met, the sum of the elements s k Before the step of determining the corresponding current time-frequency matrix as the time-frequency matrix after denoising, the method further includes:

[0079] Detection k -s k-1 Is it greater than s k-1 -s k-2 ;

[0080] The above tests k -s k-1 Is it greater than s k-1 -s k-2 After the steps, it also includes:

[0081] If s k -s k-1 Less than s k-1 -s k-2 , the current adaptive factor is adjusted according to the preset step size, and the step of calculating the current local adaptive threshold based on the current adaptive factor is returned.

[0082] In this embodiment, the current adaptive factor μ k ∈(μ0,1), μ0 is the initial value of the adaptive factor, and the preset step size is set to 0.04.

[0083] It should be understood that in the first round of iteration, after the time-frequency matrix is denoised using the current local adaptive threshold Threshold1, most of the fixed-frequency interference and noise can be eliminated. At this time, the sum of the elements of the time-frequency matrix will be greatly reduced. As the number of iterations increases, the current adaptive factor continues to increase, and the sum of the elements in the current time-frequency matrix s k The speed of decrease gradually slows down. If s k -s k-1 Greater than s k-1 -s k-2 , which means that the optimal value of the current adaptive factor is obtained. The sum of the elements in this round of iteration is s k The corresponding current time-frequency matrix is the time-frequency matrix after denoising.

[0084] Optionally, when the signal-to-noise ratio is less than a preset value, the initial value of the noise threshold is calculated in step S5, and the initial value of the noise threshold is used as the current noise threshold. The number of elements N in the time-frequency matrix that are greater than the current noise threshold is counted with the current noise threshold as the boundary. s And the number of elements N that are smaller than the current noise threshold n , further updating the current noise threshold according to the elements greater than the current noise threshold and the elements less than the current noise threshold, including:

[0085] S501, determining the maximum value M of the elements in the time-frequency matrix and calculating the average value m of all elements;

[0086] S502, calculating an initial value of the noise threshold using the maximum value M and the average value m;

[0087] S503: Taking the initial value of the noise threshold as the current noise threshold, and taking the current noise threshold as the boundary, counting the number of elements N in the time-frequency matrix that are greater than the current noise threshold s And the number of elements N that are smaller than the current noise threshold n ;

[0088] S504, respectively calculating a first mean value of elements greater than a current noise threshold and a second mean value of elements less than the current noise threshold in the time-frequency matrix;

[0089] S505: Calculate the average of the first mean and the second mean to obtain the updated noise threshold ε k .

[0090] In this embodiment, the average value m of all elements and the maximum value M of the elements in the time-frequency matrix are used as the initial value of the noise threshold, that is, ε0 = (M + m) / 2. Then, with ε0 as the boundary, the number of elements N in the time-frequency matrix that are greater than the current noise threshold is counted. s And the number of elements N that are smaller than the current noise threshold n , and calculate the average values of the two parts of elements respectively to get the first average value X s1 and the second mean value X n1 . Further, the first average value X s1 With the second mean X n1 The mean of the two is used as the updated noise threshold ε k .

[0091] Optionally, if the updated noise threshold ε k If the second preset condition is met, the updated noise threshold ε is used. k Before the step of denoising the time-frequency matrix, the following steps are also included:

[0092] Detection of ε k Is it equal to εk-1 ;

[0093] The above detection ε k Is it equal to ε k-1 After the steps, it also includes:

[0094] If ε k Not equal to ε k-1 , the updated noise threshold ε k As the current noise threshold, and return the number of elements N in the time-frequency matrix that are greater than the current noise threshold, with the current noise threshold as the boundary. s And the number of elements N that are smaller than the current noise threshold n steps.

[0095] The time-frequency matrix denoising method based on the local adaptive threshold algorithm and the iterative algorithm provided by the present invention is further illustrated below through simulation experiments.

[0096] Simulation conditions: the normalized frequency set of the hopping frequency is 0.025-0.20, the frequency interval is 0.025, there are 8 hopping signals in total, the hopping period is 5ms, the short-time Fourier transform uses a Hamming window with a length of 507, and the sampling frequency is set to 0.1MHz.

[0097] (1) Denoising using a local adaptive threshold algorithm under high signal-to-noise ratio conditions

[0098] In order to verify that the local adaptive threshold method is not suitable for low signal-to-noise ratio conditions, this embodiment calculates the time-frequency diagrams when SNR=-5dB and SNR=5dB, as well as the time-frequency diagram after denoising by the local adaptive threshold algorithm.

[0099] Figure 3a is a time-frequency diagram under low signal-to-noise ratio conditions provided by an embodiment of the present invention, Figure 3b The embodiment of the present invention provides Figure 3a The time-frequency diagram after denoising. Figure 3a-3b As shown in FIG, in the case of low signal-to-noise ratio (SNR=-5dB), since the amplitude of some noise is greater than the signal amplitude, only part of the noise can be removed by using the local adaptive threshold, and the denoising result is not ideal.

[0100] Figure 4a is a time-frequency diagram under a high signal-to-noise ratio condition provided by an embodiment of the present invention, Figure 4b The embodiment of the present invention provides Figure 4a The time-frequency diagram after denoising. Figure 4a-4bAs shown in the figure, under high signal-to-noise ratio conditions (SNR=5dB), the amplitude difference between the signal and the noise is easy to distinguish, so most of the noise can be removed by the local adaptive threshold algorithm. In addition, the denoising algorithm based on the local adaptive threshold has a small amount of calculation and is relatively fast, and is more suitable for denoising under high signal-to-noise ratio conditions.

[0101] (2) Denoising using iterative algorithms under low signal-to-noise ratio conditions

[0102] Figure 5a is another time-frequency diagram provided by an embodiment of the present invention, Figure 5b The embodiment of the present invention provides Figure 5a The three-dimensional time-frequency diagram of Figure 5c The embodiment of the present invention provides Figure 5a The time-frequency diagram after denoising. Figure 5d The embodiment of the present invention provides Figure 5a The three-dimensional time-frequency diagram after denoising. Figures 5a-5d As shown in FIG, in the case of low signal-to-noise ratio (SNR=-5dB), the iterative algorithm can better remove the influence of noise, can better eliminate noise, and the extracted time-frequency surface is also smoother.

[0103] It can be seen from the above embodiments that the beneficial effects of the present invention are:

[0104] The present invention provides a time-frequency matrix denoising method based on a local adaptive threshold algorithm and an iterative algorithm. On the one hand, when the signal-to-noise ratio of the time-frequency graph is high, considering that the amplitude difference between the signal and noise is large and easy to distinguish, and the signal after short-time Fourier transform is concentrated in a fixed time-frequency domain, while the noise is distributed in the entire time-frequency plane, the above-mentioned difference can be utilized to eliminate noise by finding a suitable local adaptive threshold. On the other hand, in the case of low signal-to-noise ratio, the time-frequency matrix is divided into a signal part and a noise part based on a noise threshold, and then the noise threshold is updated according to the elements of the two parts until the noise threshold no longer changes, and then the time-frequency graph is denoised using the noise threshold. The present invention uses different algorithms to process time-frequency graphs with different signal-to-noise ratios, which can effectively eliminate noise and obtain good denoising effect.

[0105] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Therefore, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0106] Descriptions with reference to the terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" mean that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any suitable manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification.

[0107] Although the present application is described herein in conjunction with various embodiments, in the process of implementing the claimed application, those skilled in the art can understand and implement other variations of the disclosed embodiments by reviewing the drawings, the disclosure, and the appended claims.

[0108] The above is a further detailed description of the present invention in conjunction with specific preferred embodiments, and the specific implementation of the present invention should not be considered to be limited to these descriptions. For those skilled in the art of the present invention, without departing from the concept of the present invention, several simple deductions or substitutions can be made, which should be considered to fall within the scope of protection of the present invention.

Claims

1. A time-frequency matrix denoising method based on a local adaptive threshold algorithm and an iterative algorithm, characterized in that: include: Obtain the image collected by the drone and perform short-time Fourier transform to obtain the time-frequency matrix; Calculating the signal-to-noise ratio of the time-frequency matrix according to the average value of all elements in the time-frequency matrix; When the signal-to-noise ratio is greater than a preset value, an initial value of the adaptive factor is calculated, the initial value of the adaptive factor is used as the current adaptive factor, a current local adaptive threshold is calculated according to the current adaptive factor, and the time-frequency matrix is further denoised using the current local adaptive threshold, and the sum of the elements in the current time-frequency matrix is calculated. ; If the sum of the elements If the first precondition is met, the sum of the elements The corresponding current time-frequency matrix is determined as the time-frequency matrix after denoising; When the signal-to-noise ratio is less than a preset value, the initial value of the noise threshold is calculated, the initial value of the noise threshold is used as the current noise threshold, and the number of elements in the time-frequency matrix that are greater than the current noise threshold is counted with the current noise threshold as the boundary. and the number of elements smaller than the current noise threshold , further updating the current noise threshold according to elements greater than the current noise threshold and elements less than the current noise threshold; If the updated noise threshold If the second preset condition is met, the updated noise threshold is used. Performing denoising on the time-frequency matrix; When the signal-to-noise ratio is less than a preset value, the initial value of the noise threshold is calculated, the initial value of the noise threshold is used as the current noise threshold, and the number of elements in the time-frequency matrix that are greater than the current noise threshold is counted with the current noise threshold as the boundary. and the number of elements smaller than the current noise threshold , further updating the current noise threshold according to elements greater than the current noise threshold and elements less than the current noise threshold, comprising: Determine the maximum value of the elements in the time-frequency matrix And calculate the average of all elements ; Using the maximum value and the average value Calculate the initial value of the noise threshold; The initial value of the noise threshold is used as the current noise threshold, and the number of elements in the time-frequency matrix that are greater than the current noise threshold is counted with the current noise threshold as the boundary. and the number of elements smaller than the current noise threshold ; respectively calculating a first mean value of elements greater than a current noise threshold and a second mean value of elements less than the current noise threshold in the time-frequency matrix; Calculate the average of the first mean and the second mean to obtain the updated noise threshold .

2. The time-frequency matrix denoising method based on local adaptive threshold algorithm and iterative algorithm according to claim 1 is characterized in that: When the signal-to-noise ratio is greater than a preset value, an initial value of the adaptive factor is calculated, the initial value of the adaptive factor is used as the current adaptive factor, a current local adaptive threshold is calculated according to the current adaptive factor, and the time-frequency matrix is further denoised using the current local adaptive threshold, and the sum of the elements in the current time-frequency matrix is calculated. The steps include: Determine the maximum value of the elements in the time-frequency matrix And calculate the average of all elements ; Using the maximum value and the average value Calculate the initial value of the adaptive factor; Taking the initial value of the adaptive factor as the current adaptive factor, and calculating the current local adaptive threshold according to the current adaptive factor; Performing denoising on the time-frequency matrix using the current local adaptive threshold, and setting elements in the time-frequency matrix that are smaller than the current local adaptive threshold to 0; Calculate the sum of the elements in the current time-frequency matrix after denoising , Indicates the current iteration number.

3. The time-frequency matrix denoising method based on local adaptive threshold algorithm and iterative algorithm according to claim 2 is characterized in that: If the sum of the elements If the first precondition is met, the sum of the elements Before the step of determining the corresponding current time-frequency matrix as the time-frequency matrix after denoising, the method further includes: Detection Is it greater than ; The detection Is it greater than After the steps, it also includes: like Less than , then adjust the current adaptive factor according to the preset step size, and return to the step of calculating the current local adaptive threshold according to the current adaptive factor.

4. The time-frequency matrix denoising method based on local adaptive threshold algorithm and iterative algorithm according to claim 2 is characterized in that: Using the maximum value and the average value The steps for calculating the initial value of the adaptive factor include: The average value With the maximum value The ratio of is taken as the initial value of the adaptive factor.

5. The time-frequency matrix denoising method based on local adaptive threshold algorithm and iterative algorithm according to claim 2 is characterized in that: The step of taking the initial value of the adaptive factor as the current adaptive factor and calculating the current local adaptive threshold according to the current adaptive factor comprises: Using the initial value of the adaptive factor as the current adaptive factor; Select the time-frequency matrix i Calculate the maximum and minimum values of the elements in the row and their sum ; the sum of the two Multiplying it with the current adaptive factor, we get the time-frequency matrix i The current local adaptive threshold for the row.

6. The time-frequency matrix denoising method based on local adaptive threshold algorithm and iterative algorithm according to claim 1 is characterized in that: Using the maximum value and the average value The steps of calculating the initial value of the noise threshold include: The maximum value With the average The mean of is taken as the initial value of the noise threshold.

7. The time-frequency matrix denoising method based on local adaptive threshold algorithm and iterative algorithm according to claim 1 is characterized in that: If the updated noise threshold If the second preset condition is met, the updated noise threshold is used. Before the step of performing denoising on the time-frequency matrix, the method further includes: Detection Is it equal to ; The detection Is it equal to After the steps, it also includes: like Not equal to , the updated noise threshold As the current noise threshold, and return to the current noise threshold as the boundary, counting the number of elements in the time-frequency matrix that are greater than the current noise threshold and the number of elements smaller than the current noise threshold steps.

8. The time-frequency matrix denoising method based on local adaptive threshold algorithm and iterative algorithm according to claim 1 is characterized in that: The preset value is 10.

Images