A spectral smoothing method applicable to infrared spectral data

Abstract

The present invention discloses a spectral smoothing method applicable to infrared spectral data. The steps of the method are as follows: (1) Determine whether the infrared spectral data is sampled at equal intervals in the spectral dimension; (2) If the infrared spectral data is non-uniformly sampled, perform uniform resampling on the infrared spectral data; (3) Perform mean filtering on the uniformly sampled infrared spectral data; (4) Calculate the spectral similarity before and after mean filtering; (5) Determine whether the spectral similarity is within the threshold range. If it is not within the threshold range, continue to perform mean filtering; if it is within the threshold range, output the smoothed infrared spectral data. The method of the present invention has the characteristics of good accuracy, easy operation, and easy determination of parameters.

Description

Technical Field

[0001] The present invention relates to the technical field of earth observation, and particularly to a spectral smoothing method applicable to infrared spectral data, which is used to remove noise from infrared spectral data and obtain a more accurate spectral trend line. Background Art

[0002] Infrared spectra are mainly formed by the vibration of substance molecules and can accurately reflect the absorption characteristics of substance components. Using this principle, infrared spectral technology is widely applied to various substance identification scenarios. In the field of earth observation, infrared hyperspectral remote sensing technology has been applied to fields including resource exploration, surface environment monitoring, and atmospheric environment monitoring, and excellent results have been achieved.

[0003] Whether it is a non-imaging infrared spectrometer or an infrared hyperspectral imager, they both have a relatively complex structure and a long internal radiation transmission optical path. Affected by components such as optics, detectors, and readout circuits, the output infrared spectral data inevitably contains various noises with different characteristics. Currently, the methods for spectral denoising mainly include two categories: frequency filtering methods and spatial domain filtering methods. These methods require many parameters to be set artificially. When the parameters are set properly, good denoising effects can be achieved. The accuracy of parameter setting directly affects the denoising effect. The molecular absorption in the infrared spectral band is mainly band absorption, and the absorption width is relatively large. Based on this, the present invention proposes a spectral smoothing method applicable to infrared spectral data that combines continuous smoothing with spectral similarity constraints. Summary of the Invention

[0004] Aiming at the existing technical gaps and deficiencies, the technical problem to be solved by the present invention is to provide a denoising method applicable to infrared spectral data with good accuracy and easy parameter setting.

[0005] To solve the above technical problem, a spectral smoothing method applicable to infrared spectral data provided by the present invention is characterized in that:

[0006] (1) Judge whether the infrared spectral data is sampled at equal intervals in the spectral dimension. Calculate the sampling interval of the infrared spectral data

[0007] Δλ k =λ k+1 -λ k

[0008] where λ k is the k-th sampling wavelength, k = 1, 2, 3,..., N k and N k is the number of sampling wavelengths; Δλ k is the sampling interval. Calculate the average value and standard deviation of the sampling interval of the infrared spectral data

[0009]

[0010]

[0011] Among them, is the average value of the sampling interval, and σ Δλ is the standard deviation. If is less than or equal to 0.1, it is determined that the infrared spectrum data is non-uniformly sampled;

[0012] (2) If the infrared spectrum data is non-uniformly sampled, then resample the infrared spectrum data at equal intervals. When it is determined that the infrared spectrum data is non-uniformly sampled, the cubic Lagrange interpolation method is used to resample the infrared spectrum data at equal intervals, and the sampling interval is the average value of the sampling intervals in step (1);

[0013] (3) Perform mean filtering on the equally sampled infrared spectrum data. The mean filtering window radius is set to 1, and mean filtering is performed on the equally sampled infrared spectrum data

[0014]

[0015] Among them, S m (λ k ) is the spectral data obtained by the m-th mean filtering, and S 0 (λ k ) is the original infrared spectrum data; the value rule of i is that if k - 1 < 0, then i = 1, and if k + 1 < N k , then i = N k ;

[0016] (4) Calculate the spectral similarity before and after mean filtering. The calculation method is

[0017]

[0018] (5) Determine whether the spectral similarity is within the threshold range. If δ m and the threshold T satisfy δ m > T, within the threshold range, then continue to perform mean filtering, repeating steps (3) and (4); if δ m and the threshold T satisfy δ m ≤ T, within the threshold range, then output the smoothed infrared spectrum data S m (λ k ).

[0019] The beneficial effects of the present invention are as follows: By adopting multiple consecutive smoothings and judging the spectral similarity of the two consecutive smoothed spectra to select the optimal spectral data smoothing result, replacing the setting of the window size with the setting of the error limit simplifies the parameter setting. Description of the Drawings

[0020] Figure 1 Example of original infrared spectral data

[0021] Figure 2 Example of spectral smoothing result of the method of the present invention

[0022] Figure 3 Technical flow chart of a spectral smoothing method applicable to infrared spectral data Detailed implementation manners

[0023] The following further describes the embodiments of the present invention in detail, but the embodiments are not limited to the present invention. Any similar methods and their similar variations using the present invention shall fall within the protection scope of the present invention.

[0024] (1) Judge whether the infrared spectral data is equally spaced sampled in the spectral dimension. Calculate the sampling interval of the infrared spectral data

[0025] Δλ k =λ k+1 -λ k

[0026] where λ k is the k-th sampling wavelength, k = 1, 2, 3,..., N k and N k is the number of sampling wavelengths; Δλ k is the sampling interval. Calculate the average value and standard deviation of the sampling interval of the infrared spectral data

[0027]

[0028]

[0029] where is the average value of the sampling interval, and σ Δλ is the standard deviation. If is less than or equal to 0.1, it is determined that the infrared spectral data is non-equally spaced sampled;

[0030] (2) If the infrared spectral data is non-equally spaced sampled, then resample the infrared spectral data at equal intervals. When it is determined that the infrared spectral data is non-equally spaced sampled, the cubic Lagrange interpolation method is used to resample the infrared spectral data at equal intervals, and the sampling interval is the average value of the sampling interval in step (1);

[0031] (3) Perform mean filtering on the equally spaced sampled infrared spectral data. The mean filtering window radius is set to 1, and mean filtering is performed on the equally spaced infrared spectral data

[0032]

[0033] Among them, S m (λ k ) is the spectral data obtained by the m-th mean filtering, and S 0 (λ k ) is the original infrared spectral data; the value rule of i is that if k - 1 < 0, then i = 1, and if k + 1 < N k , then i = N k ;

[0034] (4) Calculate the spectral similarity before and after mean filtering. The calculation method is

[0035]

[0036] (5) Determine whether the spectral similarity is within the threshold range. If δ m and the threshold T satisfy δ m > T, within the threshold range, then continue to perform mean filtering, repeating steps (3) and (4); if δ m and the threshold T satisfy δ m ≤ T, within the threshold range, then output the smoothed infrared spectral data S m (λ k ).

[0037] Embodiment

[0038] To verify the effect of the present invention, the infrared spectral data measured by ATHIS is selected as the processing object. The implementation method of this embodiment is as described above, and spectral smoothing is achieved according to the following steps:

[0039] (1) Determine whether the ATHIS infrared spectral data is equally spaced sampled in the spectral dimension. Calculate the sampling interval of the infrared spectral data

[0040] Δλ k = λ k+1 - λ k

[0041] Among them, λ k is the k-th sampling wavelength, k = 1, 2, 3,..., N k , N k is the number of sampling wavelengths, which is 110; Δλ k is the sampling interval. Calculate the average value and standard deviation of the sampling interval of the infrared spectral data

[0042]

[0043]

[0044] Among them, is the average value of the sampling interval, and σΔλ is the standard deviation. The value of ATHIS is 0.03, which is less than 0.1. It is determined that the infrared spectrum data is equally spaced sampled;

[0045] (2) Perform mean filtering on the equally spaced sampled infrared spectrum data. The mean filtering window radius is set to 1, and mean filtering is performed on the equally spaced infrared spectrum data

[0046]

[0047] where S m (λ k ) is the spectral data obtained by the m-th mean filtering, and S 0 (λ k ) is the original infrared spectrum data; the value rule of i is that if k - 1 < 0, then i = 1, if k + 1 < N k , then i = N k ;

[0048] (4) Calculate the spectral similarity before and after mean filtering. The calculation method is

[0049]

[0050] (5) Determine whether the spectral similarity is within the threshold range. If δ m and the threshold T (0.001) satisfy δ m > T, within the threshold range, then continue to perform mean filtering, repeating steps (3) and (4); if δ m and the threshold T satisfy δ m ≤ T, within the threshold range, then output the smoothed infrared spectrum data S m (λ k ).

Claims

1. A spectral smoothing method applicable to infrared spectral data, characterized in that it includes the following steps: (1) Judge whether the infrared spectral data is sampled at equal intervals in the spectral dimension, and calculate the sampling interval of the infrared spectral data: Δλ k = λ k+1 - λ k where, λ k is the k-th sampling wavelength, k = 1, 2, 3, ..., N k , N k is the number of sampling wavelengths; Δλ k is the sampling interval, and the average value and standard deviation of the sampling intervals of the infrared spectral data are calculated: Among them, is the average value of the sampling interval, and σ Δλ is the standard deviation. If is less than or equal to 0.1, it is determined that the infrared spectrum data is non-equidistant sampling; (2) If the infrared spectral data is non-uniformly sampled, then perform uniform resampling on the infrared spectral data. When it is determined that the infrared spectral data is non-uniformly sampled, use the cubic Lagrange interpolation method to perform uniform resampling on the infrared spectral data, and the sampling interval is the average value of the sampling intervals in step (1); (3) Perform mean filtering on the uniformly sampled infrared spectral data, and set the mean filtering window radius to 1 to perform mean filtering on the uniformly sampled infrared spectral data: Among them, S m (λ k ) is the spectral data obtained by the m-th mean filtering, and S 0 (λ k ) is the original infrared spectral data; the value rule of i is that if k - 1 < 0, then i = 1, and if k + 1 < N k , then i = N k ; (4) Calculate the spectral similarity before and after mean filtering, and the calculation method is: (5) Determine whether the spectral similarity is within the threshold range. If δ m and the threshold T satisfy δ m > T, within the threshold range, then continue with mean filtering and repeat steps (3) and (4); if δ m and the threshold T satisfy δ m ≤ T, within the threshold range, then output the smoothed infrared spectral data S m (λ k ).

Images