A feature selection and verification method for radar electromagnetic simulation waveforms with added weights

Abstract

The present invention discloses a feature selection verification method for radar electromagnetic simulation waveforms with added weights. The present invention assigns weights to low-frequency components. When calculating the difference in the DC component of the radar electromagnetic simulation waveform, it can effectively improve the failure of positive and negative alternating data, fully consider the difference between the DC component and the low-frequency component, and is more suitable for data analysis with a large proportion of DC components. By assigning weights to the first-order derivatives of high-frequency components, the present invention can reflect the changing characteristics of the data and describe the details of the data more specifically. The present invention increases the weight of the first-order derivative of the low-frequency component, and can be suitable for feature selection verification calculations of data with a small proportion of high-frequency components but a relatively large proportion of low-frequency components.

Description

Technical Field

[0001] The invention belongs to the field of radar electromagnetic simulation, and in particular relates to a feature selection and verification method for a weighted radar electromagnetic simulation waveform. Background Art

[0002] As simulation methods play an increasingly important role in electromagnetic analysis, prediction, and design, electromagnetic simulation algorithms and models continue to emerge. Only by ensuring the validity of the simulation algorithm or model can the simulation results have practical application value. Feature selection validation (FSV) has emerged as a necessary step. This method preprocesses the raw data into DC, low-frequency, and high-frequency components. It then calculates the amplitude difference measure (ADM) using a formula. The DC component of the ADM is called the ODM, and δ is the denominator in the ODM calculation. The ADM is composed of the ODM and the low-frequency component. The feature difference measure (FDM) is then calculated using a formula. The FDM consists of FDM1, FDM2, and FDM3. The ADM and FDM are then combined to form the global difference measure (GDM). The values of ADM, GDM, and FDM are referred to as FSV values. Using a conversion table, the corresponding distribution histograms can be clearly obtained, which can then be converted into a natural language description of the waveform fit quality. The results obtained by the feature selection validation method should not differ significantly from those obtained by expert visual inspection, otherwise they will lose their significance.

[0003] However, current feature selection verification methods fail when processing certain special data. The algorithm and parameters of the method itself need further demonstration, and the performance of the method needs to be improved. When using standard feature selection verification methods for data containing alternating positive and negative values, or radar data containing transient components, they fail. This is because current feature selection verification methods are still in the development stage, and their consideration of each component is too simplistic. This leads to the following problems when processing radar waveform signals:

[0004] 1. When calculating the amplitude difference metric (ADM), the DC component and the low-frequency component are calculated separately without considering the impact of the two on each other;

[0005] 2. When calculating FDM3() in the characteristic difference measure (FDM), the description of the details is not detailed enough, and the coupling effect of the first-order derivative and the second-order derivative of the high-frequency component in the radar waveform is not fully considered;

[0006] 3. When calculating the final result of the global difference measure (GDM), the result is not ideal for radar waveforms with a large proportion of low-frequency components.

[0007] Radar electromagnetic simulations often use transient excitation sources, resulting in results that often exhibit alternating positive and negative behavior and contain transient components. To address these failure scenarios, appropriate weights can be set to allow feature selection verification methods to consider more complex situations, expanding their scope of application. Summary of the Invention

[0008] The purpose of the present invention is to overcome the above-mentioned shortcomings and provide a feature selection and verification method for radar electromagnetic simulation waveforms with added weights, which can effectively improve the failure of alternating positive and negative data and fully consider the difference between the DC component and the low-frequency component.

[0009] In order to achieve the above object, the present invention comprises the following steps:

[0010] Obtain radar electromagnetic simulation waveforms and select positive and negative alternating radar signal waveforms to be evaluated;

[0011] Perform Fourier decomposition on the positive and negative alternating radar signal waveform to be evaluated to DC to obtain low-frequency components and high-frequency components;

[0012] Assign weights to the low-frequency components, calculate the difference in the DC component of the radar electromagnetic simulation waveform, and then calculate the amplitude difference point by point;

[0013] Assign weights to the first-order derivatives of the high-frequency components, calculate the difference of the second-order derivatives of the high-frequency components of the radar electromagnetic simulation waveform, and then calculate the characteristic difference point by point;

[0014] Modify the weight of the first-order derivative of the low-frequency component and calculate the global difference of the radar electromagnetic simulation waveform point by point;

[0015] Feature selection verification is performed on data with a low-frequency component ratio higher than the threshold, and the verification is finally completed.

[0016] The method for calculating the difference in the DC component of the radar electromagnetic simulation waveform is as follows:

[0017]

[0018] Where DC1(i) is the DC component of the actual waveform, DC2(i) is the DC component of the simulated waveform, Lo1(i) is the low-frequency component of the actual waveform, Lo2(i) is the low-frequency component of the simulated waveform, and N is the total number of data.

[0019] The method for calculating the characteristic difference of the second-order derivative of the high-frequency component after adding the weight is as follows:

[0020]

[0021] Where Hi′1() is the first-order derivative of the high-frequency component of the actual waveform, Hi′2() is the first-order derivative of the high-frequency component of the simulated waveform, Hi"1(i) is the second-order derivative of the high-frequency component of the actual waveform, Hi"2() is the second-order derivative of the high-frequency component of the simulated waveform, and N is the total number of data.

[0022] The global difference includes a low-frequency component FDM1() and a high-frequency component. The high-frequency component is the sum of a characteristic difference FDM2() of the first-order derivative of the high-frequency component and a characteristic difference FDM3() of the second-order derivative of the high-frequency component after adding a weight.

[0023] The calculation method of the characteristic difference FDM(x) is as follows:

[0024] FDM(x)=2(0.9|FDM1(x)|+0.1(|FDM2(x)|+|FDM3(x)|))

[0025] Among them, FDM1(x) is the characteristic difference of the first-order derivative of the low-frequency component, FDM2(x) is the characteristic difference of the first-order derivative of the high-frequency component, and FDM3(x) is the characteristic difference of the second-order derivative of the high-frequency component after adding weights.

[0026] The calculation method of the global difference GDM(x) is as follows:

[0027]

[0028] Among them, ADM(x) is the amplitude difference.

[0029] Compared with the prior art, the present invention assigns weights to low-frequency components. When calculating the difference in the DC component of the radar electromagnetic simulation waveform, it can effectively improve the failure of alternating positive and negative data, fully consider the difference between the DC component and the low-frequency component, and is more suitable for data analysis with a large proportion of DC components. By assigning weights to the first-order derivatives of high-frequency components, the present invention can reflect the changing characteristics of the data and describe the details of the data more specifically. The present invention modifies the weight of the first-order derivative of the low-frequency component, which can be suitable for feature selection and verification calculations of data with a small proportion of high-frequency components but a relatively large proportion of low-frequency components. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 is a flow chart of the present invention;

[0031] Figure 2 This is a radar case in the embodiment;

[0032] Figure 3 is the ADM histogram in the embodiment;

[0033] Figure 4 is the FDM histogram in the embodiment;

[0034] Figure 5 is the GDM histogram in the embodiment. DETAILED DESCRIPTION

[0035] The present invention will be further described below with reference to the accompanying drawings.

[0036] See also Figure 1 , the present invention comprises the following steps:

[0037] Step 1: Obtain radar electromagnetic simulation waveform and select the positive and negative alternating radar signal waveform to be evaluated;

[0038] Step 2: Perform Fourier decomposition on the waveform to DC to obtain low-frequency components and high-frequency components;

[0039] Step three: assign weights to the low-frequency components, calculate the difference ODM of the DC component of the radar electromagnetic simulation waveform, and then calculate the amplitude difference ADM point by point. When calculating the difference ODM of the DC component of the radar electromagnetic simulation waveform, calculate δ based on the DC component accounting for 80% and the low-frequency component accounting for 20%. After adding the low-frequency component, δ can act as a stabilizing factor, and a certain weight distribution is performed on the DC component and the low-frequency component, which can effectively improve the failure of alternating positive and negative data, fully consider the difference between the DC component and the low-frequency component, and is more suitable for data analysis with a large proportion of DC components. For the improvement of the difference ODM of the DC component, the formula for calculating the difference δ of the DC component of the radar electromagnetic simulation waveform is as follows:

[0040]

[0041] Where DC1(i) is the DC component of the actual waveform, DC2(i) is the DC component of the simulated waveform, Lo1(i) is the low-frequency component of the actual waveform, Lo2(i) is the low-frequency component of the simulated waveform, N is the total number of data, and 0.2 is the weight assigned to the low-frequency component.

[0042] The weights assigned to the first-order derivatives of the high-frequency components are used to calculate the differences in the second-order derivatives of the high-frequency components of the radar electromagnetic simulation waveform. The characteristic difference FDM3 of the second-order derivatives of the high-frequency components after weighting is then calculated point by point. When calculating the characteristic difference FDM3 of the second-order derivatives of the high-frequency components after weighting, the second-order derivatives of the high-frequency components after weighting are calculated based on an 80% share of the second-order derivatives and a 20% share of the first-order derivatives. This allows for a more detailed description of the data's changing characteristics and a more specific description of the data's details. The formula for calculating the characteristic difference FDM3(x) of the second-order derivatives of the high-frequency components after weighting is as follows:

[0043]

[0044] Where Hi′1(i) is the first-order derivative of the high-frequency component of the actual waveform, Hi′2(i) is the first-order derivative of the high-frequency component of the simulated waveform, Hi"1(i) is the second-order derivative of the high-frequency component of the actual waveform, Hi"2(i) is the second-order derivative of the high-frequency component of the simulated waveform, N is the total number of data, and 0.2 is the weight assigned to the first-order derivative of the high-frequency component.

[0045] Modify the weight of the first-order derivative of the low-frequency component and calculate the global difference of the radar electromagnetic simulation waveform point by point; the global difference includes the low-frequency component FDM1(x) and the high-frequency component. The high-frequency component is the sum of the characteristic difference FDM2(x) of the first-order derivative of the high-frequency component and the characteristic difference FDM3(x) of the second-order derivative of the high-frequency component after adding the weight. Increase the weight of the low-frequency component FDM1(x) to 90%, and set the weight of the high-frequency component FDM2(x)+FDM3(x) to 10%. This is more suitable for feature selection verification calculations on data with a small proportion of high-frequency components but a relatively large proportion of low-frequency components, and can achieve good results. The formula for calculating the characteristic difference FDM is as follows:

[0046] FDM(x)=2(0.9|FDM1(x)|+0.1(|FDM2(x)|+|FDM3(x)|))

[0047] Among them, FDM1(x) is the characteristic difference of the first-order derivative of the low-frequency component, FDM2(x) is the characteristic difference of the first-order derivative of the high-frequency component, and FDM3(x) is the characteristic difference of the second-order derivative of the high-frequency component after adding weights.

[0048] The calculation method of the global difference GDM(x) is as follows:

[0049]

[0050] Among them, ADM(x) is the amplitude difference.

[0051] Step 4: Perform feature selection verification on the data whose low-frequency component ratio is higher than the threshold, and finally complete the verification.

[0052] Example:

[0053] This method adds weights to the traditional feature selection method, making it suitable for different scenarios and more effective for specific applications. By performing feature selection verification on the case curve, the corresponding amplitude difference (ADM), feature difference (FDM3), and global difference (GDM) are obtained.

[0054] As shown in Table 1, the results of the expert visual assessment of the two data sets are as follows: the two data sets have good consistency in terms of overall data values, but there are certain differences in the details of the data. The final evaluation of the waveforms is that the data are basically the same, with an evaluation level of 3.

[0055] However, the alternating positive and negative values of the two sets of data affected the results of the traditional feature selection verification method. The resulting amplitude difference (ADM) result had an evaluation level of 4, and the feature difference (FDM3) result of the second-order derivative of the high-frequency component after adding weights had an evaluation level of 5, which was significantly different from the expert visual evaluation results.

[0056] Based on the Fourier decomposition of the two sets of data, it was found that the DC and low-frequency components of the two sets of data accounted for a larger proportion, while the high-frequency components accounted for a smaller proportion. Therefore, by changing the above data weight ratios for testing, an evaluation level of 3 was obtained, which was close to the expert visual inspection results.

[0057] Table 1 Conversion scale of feature selection validation methods

[0058]

[0059] Step 1: If Figure 2 The two groups of radar detection electromagnetic simulation waveforms shown are: one group is the theoretically derived waveform, represented by a solid line; the other group is the actually measured waveform, represented by a dotted line. The waveforms are alternating positive and negative. The DC component, low-frequency component and high-frequency component of the two groups of waveforms are extracted.

[0060] Step 2: The DC component and low-frequency component extracted in this embodiment are calculated by ADM with the DC component accounting for 80% and the low-frequency component accounting for 20%. The result is as follows: Figure 3 As shown, the horizontal axis of the histogram represents the evaluation level, and the vertical axis represents the proportion of the level. This fully considers the impact of the low-frequency component on ADM, and the previous failure situation will not occur. In the obtained histogram, the evaluation level is 3, which is consistent with the expert visual inspection result.

[0061] Step 3: The second-order derivative of the high-frequency component extracted from the case is weighted at 80%, and the first-order derivative of the high-frequency component is weighted at 20%. The FDM3 calculation of the second-order derivative of the high-frequency component after adding weights is obtained. The result is as follows Figure 4 As shown, the horizontal axis of the histogram represents the evaluation level, and the vertical axis represents the proportion of the level. Compared with the traditional feature selection verification method, it describes the detailed changes of the data in more detail.

[0062] Step 4: Calculate the feature difference FDM by taking the low-frequency component extracted from the case at a weight of 90% and the high-frequency component at a weight of 10%, and then obtain the final global difference GDM result according to the formula. Figure 5As shown, the horizontal axis of the histogram represents the evaluation level, and the vertical axis represents the proportion of the level. The low-frequency component of the original data accounts for a relatively large proportion. According to this weight, the final global difference GDM result can be more consistent with the characteristics of this set of data, so that the result obtained is more credible.

[0063] The waveform evaluation result after adding the weight is evaluation level 3, which is consistent with the evaluation level 3 of expert visual inspection. It can effectively solve the problem of failure of feature selection verification method when evaluating alternating positive and negative radar electromagnetic simulation waveforms.

[0064] The above further describes the present invention in detail with reference to specific embodiments of the present invention. The contents described above are all explanations of the present invention, but these descriptions cannot be understood as limiting the scope of the present invention. The scope of protection of the present invention is defined by the appended claims, and any changes based on the claims of the present invention are within the scope of protection of the present invention.

Claims

1. A feature selection and verification method for radar electromagnetic simulation waveforms with added weights, characterized in that: The following steps are involved: Obtain radar electromagnetic simulation waveforms and select positive and negative alternating radar signal waveforms to be evaluated; Perform Fourier decomposition on the positive and negative alternating radar signal waveform to be evaluated to DC to obtain low-frequency components and high-frequency components; Assign weights to the low-frequency components, calculate the difference in the DC component of the radar electromagnetic simulation waveform, and then calculate the amplitude difference point by point. The method for calculating the difference in the DC component of the radar electromagnetic simulation waveform is as follows: in, is the DC component of the actual waveform, is the DC component of the simulation waveform, is the low-frequency component of the actual waveform, is the low-frequency component of the simulation waveform, and N is the total number of data; Assign weights to the first-order derivatives of the high-frequency components, calculate the difference of the second-order derivatives of the high-frequency components of the radar electromagnetic simulation waveform, and then calculate the characteristic difference point by point. The method for calculating the characteristic difference of the second-order derivatives of the high-frequency components after adding weights is as follows: in, is the first-order derivative of the high-frequency component of the actual waveform, is the first-order derivative of the high-frequency component of the simulation waveform, is the second-order derivative of the high-frequency component of the actual waveform, is the second-order derivative of the high-frequency component of the simulation waveform, and N is the total number of data; Modify the weight of the first-order derivative of the low-frequency component and calculate the global difference of the radar electromagnetic simulation waveform point by point; the global difference includes the low-frequency component and high-frequency components, where the high-frequency components are the characteristic difference of the first-order derivative of the high-frequency components The characteristic difference between the second-order derivative of the high-frequency component and the weighted high-frequency component of and; Feature selection verification is performed on data with a low-frequency component ratio higher than the threshold, and the verification is finally completed.

2. The feature selection and verification method for a weighted radar electromagnetic simulation waveform according to claim 1, characterized in that: Feature Difference The calculation method is as follows: in, is the characteristic difference of the first-order derivative of the low-frequency component, is the characteristic difference of the first-order derivative of the high-frequency component, It is the characteristic difference of the second-order derivative of the high-frequency component after adding the weight.

3. The feature selection and verification method for weighted radar electromagnetic simulation waveform according to claim 2, characterized in that: Global difference The calculation method is as follows: Among them, ADM(x) is the amplitude difference.

Images