A multi-slit hyperspectral data reconstruction processing method

Abstract

The present application relates to a kind of multi-slit hyperspectral data processing method, in particular to a kind of multi-slit hyperspectral data reconstruction processing method, solve the problem that real-time acquisition high-precision multi-slit hyperspectral reconstructed data is difficult in prior art.This method includes the following steps: step 1: carry out dark current correction to multi-slit area array data;Step 2: relative radiation correction;Step 3: bad pixel correction;Step 4: define one of the slit as reference slit, define the bad pixel corrected data of reference slit as reference slit hyperspectral data;Then, with reference slit hyperspectral data as correction reference, the bad pixel corrected data of the remaining each slit is respectively corrected to horizontal image shift error;Step 5: spectral position error correction;Step 6: interframe error correction, obtain the corrected data of the remaining each slit;Step 7: the corrected data of the remaining each slit and reference slit hyperspectral data are merged and reconstructed, and multi-slit hyperspectral reconstructed data is obtained.

Description

TECHNICAL FIELD

[0001] The present application relates to a multi-slit hyperspectral data processing method, in particular to a multi-slit hyperspectral data reconstruction processing method. BACKGROUND

[0002] The multi-slit dispersive hyperspectral imaging system obtains the overall data of the entire ground element by push-broom. Referring to Figure 1 The existing multi-slit dispersive hyperspectral imaging system generally includes a front lens group 1, a multi-slit template 2, a collimating lens group 3, a dispersive element 4, an imaging lens group 5, and a detector 6 arranged in sequence along the optical path. Due to the use of multi-slit dispersion, the hyperspectral data of different column elements can be obtained at a single time, and then through push-broom, multiple sets of hyperspectral data of the same ground element after dispersion through different slits at different times can be obtained. Theoretically, the multiple sets of hyperspectral data of the same ground element obtained have consistency, and by averaging the multiple sets of hyperspectral data of the same ground element, hyperspectral data with higher signal-to-noise ratio can be obtained. However, in actual engineering, the multiple sets of hyperspectral data of the same ground element are obtained at different positions of the detector, different positions of the optical system, and different times, so there is a difference between the multiple sets of multi-slit hyperspectral data of the same ground element, which directly affects the accuracy of the final data. In order to ensure the consistency between the multiple sets of multi-slit hyperspectral data of the same ground element, error correction and targeted processing of other multiple sets of hyperspectral data other than the reference data are required to obtain high-precision final data, and this process is the reconstruction processing process of multi-slit hyperspectral data. In the reconstruction processing process, multiple errors and data differences need to be corrected and operated, and the overall process is relatively complicated. Under limited resources, how to simply and efficiently correct multiple errors and data differences to complete the reconstruction processing of multi-slit hyperspectral data and obtain high-precision multi-slit hyperspectral reconstructed data in real time is currently lacking.

[0003] Currently, the processing of single-slit hyperspectral data is relatively mature, and there is no instant and efficient processing method for the reconstruction processing of the inconsistency between multi-slit hyperspectral data. SUMMARY

[0004] The purpose of the present application is to provide a multi-slit hyperspectral data reconstruction processing method to solve the technical problem that the prior art lacks an efficient reconstruction processing method for the inconsistency between multi-slit hyperspectral data, and it is difficult to obtain high-precision multi-slit hyperspectral reconstructed data in real time.

[0005] The inventive concept of the present application is that, in the multi-slit hyperspectral data reconstruction processing, the transverse image shift error, the spectral position error and the inter-frame error among the multi-slit hyperspectral data are corrected through line interpolation, inter-line interpolation and inter-frame interpolation, so as to improve the consistency among the multi-slit hyperspectral data; in the multi-slit hyperspectral data merging and reconstruction link, the reference data and the difference threshold are set, the influence of the moving target on the consistency among the multi-slit hyperspectral data is filtered and avoided, and the accuracy of the reconstructed hyperspectral data is further improved. The main steps include seven links of dark current correction, relative radiation correction, bad pixel correction, transverse image shift error correction (line interpolation), spectral position error correction (inter-line interpolation), inter-frame error correction (inter-frame interpolation correction) and multi-slit hyperspectral data merging and reconstruction, and finally the efficient reconstruction processing of the multi-slit hyperspectral data is realized. The above line refers to the spatial dimension direction of the data, which is perpendicular to the push scan direction of the imaging system; the above inter-line refers to the slit dispersion direction, which is consistent with the push scan direction of the imaging system.

[0006] To solve the above technical problems and achieve the above inventive concept, the technical solution adopted by the present application is:

[0007] A multi-slit hyperspectral data reconstruction processing method, which is characterized by comprising the following steps:

[0008] Step 1: subtracting the dark current correction coefficient from the multi-slit area array data obtained in the multi-slit dispersion type hyperspectral imaging system to realize the dark current correction of the data and obtain the dark current corrected data; the dark current correction coefficient is measured;

[0009] Step 2: multiplying the dark current corrected data obtained in step 1 by the relative radiation correction coefficient to realize the relative radiation correction of the data and obtain the relative radiation corrected data; the relative radiation correction coefficient is measured;

[0010] Step 3: using one-dimensional interpolation method to correct the bad pixels of the relative radiation corrected data obtained in step 2 and obtain the bad pixel corrected data;

[0011] Step 4: defining one of the slits in the multi-slit dispersion type hyperspectral imaging system as a reference slit, and defining the bad pixel corrected data of the reference slit in the bad pixel corrected data obtained in step 3 as reference slit hyperspectral data; then taking the reference slit hyperspectral data as the correction reference, using the line interpolation method to correct the bad pixel corrected data of the remaining slits except the reference slit in the bad pixel corrected data obtained in step 3 respectively to obtain the transverse image shift error corrected data of the remaining slits;

[0012] Step 5: taking the reference slit hyperspectral data defined in step 4 as a correction reference, the transverse image motion error corrected data of each slit obtained in step 4 is respectively corrected for spectral position error by using the interline interpolation method to obtain the spectral position error corrected data of each slit;

[0013] Step 6: taking the reference slit hyperspectral data defined in step 4 as a correction reference, the spectral position error corrected data of each slit obtained in step 5 is respectively corrected for interframe error by using the interframe interpolation correction method to obtain the completed correction data of each slit;

[0014] Step 7: the completed correction data of each slit obtained in step 6 and the reference slit hyperspectral data defined in step 4 are merged and reconstructed to obtain multi-slit hyperspectral reconstructed data, and the reconstruction process is completed.

[0015] Further, in step 3, a one-dimensional interpolation method is used, only considering the interpolation in the line direction, and setting the maximum number of continuous bad pixels as four, the relative radiation corrected data obtained in step 2 is corrected for bad pixels to obtain bad pixel corrected data; specifically:

[0016] 1) when it is a single bad pixel, the bad pixel correction result is 0.5A+0.5B;

[0017] 2) when it is double continuous bad pixels, the bad pixel correction results from left to right are 0.67A+0.33B and 0.33A+0.67B, or 0.75A+0.25B and 0.25A+0.75B, respectively;

[0018] 3) when it is three continuous bad pixels, the bad pixel correction results from left to right are 0.67A+0.33B, 0.5A+0.5B, 0.33A+0.67B, respectively;

[0019] 4) when it is four continuous bad pixels, the bad pixel correction results from left to right are 0.2A+0.8B, 0.4A+0.6B, 0.6A+0.4B, 0.8A+0.2B, respectively;

[0020] The A and B respectively represent the normal pixels adjacent to the left and right of the continuous bad pixels.

[0021] Further, in step 4, when the bad pixel corrected data of each slit except the reference slit in the bad pixel corrected data obtained in step 3 is respectively corrected for transverse image motion error by using the interline interpolation method, for the bad pixel corrected data of each slit, the interline interpolation method is used to correct the pixel data S i,j The formula [1] is used for calculation, and the formula [1] is:

[0022] S i,j = αC i,t + (1 - α)C i,t+1 [1] ;

[0023] In formula [1], S i,j represents the pixel data of the output coordinate point (i, j) after correction of the lateral image shift error; α represents the correction coefficient of the lateral image shift error; C i,t and C i,t+1 respectively represent the pixel data of the line coordinate i and the column coordinate t and the pixel data of the line coordinate i and the column coordinate t+1 in the corrected data of the bad pixel of the slit obtained in step 3; the values of α and t are determined by the size of the lateral image shift error of the slit relative to the reference slit.

[0024] Further, the error sources causing the lateral image shift error in step 4 include optical system distortion error and deflection angle residual error correction error.

[0025] Further, in step 5, the remaining lateral image shift error corrected data of each slit obtained in step 4 is corrected for spectral position error by using the interline interpolation method. For the lateral image shift error corrected data of each slit, the interline interpolation method in column units is used for correction, and the pixel data S i ″ ,j′ is calculated by using formula [2] as follows:

[0026] S i ″ ,j′ = α'C k ′ ,j′ + (1 - α')C k ′ +1,j′ [2] ;

[0027] In formula [2], S i ″ ,j′ represents the pixel data of the output coordinate point (i', j') after correction of the spectral position error; α' represents the correction coefficient of the spectral position error; C k ′ ,j′ and C k ′ +1,j′ respectively represent the pixel data of the line coordinate k and the column coordinate j' and the pixel data of the line coordinate k+1 and the column coordinate j' in the lateral image shift error corrected data of the slit obtained in step 4; the values of α' and k are determined by the size of the spectral position error of the slit relative to the reference slit.

[0028] Further, in step 5, the spectral position error refers to the deviation of the spectral center wavelength between the data of the transverse image shift error correction of the remaining slits obtained in step 4 and the high-spectral data of the reference slit defined in step 4.

[0029] Further, in step 6, the calculation method of the frame interpolation correction method is to calculate the linear interpolation between the adjacent two frames by using formula [3], and the linear interpolation is taken as the correction completed data of the corresponding slit; the formula [3] is:

[0030] D f ′=α″D f,m +(1-α″)D f,m+1 [3];

[0031] In formula [3], D f ′ represents the correction completed data of the slit f, the f represents the slit serial number of the remaining slits, the slit serial number defined as the reference slit in step 4 is set as 1, and the f takes 2, 3, …, R, and the R is the total number of the slit in the multi-slit dispersion type high-spectral imaging system; α″ represents the frame error correction coefficient of the slit f; D f,m , D f,m+1 respectively represent the mth frame and the m+1th frame data of the spectral position error corrected data of the slit f in the spectral position error corrected data of the remaining slits obtained in step 5, and the m is the frame serial number; the values of α″ and m are determined by the frame error size of the slit f relative to the reference slit.

[0032] Further, the error sources causing the frame error in step 6 include the slit distance deviation caused by the assembly accuracy.

[0033] Further, in order to make the accuracy of the multi-slit high-spectral reconstruction data obtained after the merging and reconstruction higher, the step 7 is specifically:

[0034] Step 7.1: set a difference threshold μ;

[0035] Step 7.2: according to the correction completed data of the remaining slits obtained in step 6, the high-spectral data of the reference slit defined in step 4 and the difference threshold μ set in step 7.1, the merging switch matrix P f of the remaining slits except the reference slit is set, and the f takes 2, 3, …, R;

[0036] Step 7.3: according to the merging switch matrix P f, f takes 2, 3, …, R, the remaining each slit of the correction completed data obtained in step 6 is combined with the reference slit hyperspectral data defined in step 4 to reconstruct multi-slit hyperspectral data after reconstruction by formula [4], and the reconstruction processing is completed; the formula [4] is:

[0037] D''=(RD1-(D1-D'2)⊙P2-……(D1-D'2)⊙PR) / R [4]; R R

[0038] In formula [4], D'' represents multi-slit hyperspectral data after reconstruction; D1 represents reference slit hyperspectral data defined in step 4; D'2, …, D'2 represent correction completed data of slit 2, …, slit R obtained in step 6; P2, …, P2 represent the combination switch matrix of slit 2, …, slit R set in step 7.2; and ⊙ represents Hadamard product of two matrices. R R

[0039] Further, the step 7.2 is specifically:

[0040] Step 7.2.1: the matrix size of the combination switch matrix P2, f takes 2, 3, …, R, of each slit except the reference slit is set to be the same as the matrix size of the reference slit hyperspectral data defined in step 4; f

[0041] Step 7.2.2: the correction completed data of each slit in the correction completed data of each slit obtained in step 6 is subtracted from the corresponding data in the reference slit hyperspectral data defined in step 4 one by one, and then the absolute value of each difference value is compared with the difference threshold μ set in step 7.1; if the absolute value of the difference value is greater than the difference threshold μ set in step 7.1, the corresponding position element in the combination switch matrix P2, f takes 2, 3, …, R, of the corresponding slit is set to 0; if the absolute value of the difference value is less than or equal to the difference threshold μ set in step 7.1, the corresponding position element in the combination switch matrix P2, f takes 2, 3, …, R, of the corresponding slit is set to 1; according to the principle, the setting of all elements in the combination switch matrix P2, f takes 2, 3, …, R, of each slit except the reference slit is completed. f f f

[0042] The beneficial effects of the present application are:

[0043] ​​​​​​​​(1) The multi-slit hyperspectral data reconstruction processing method of the present application, through dark current correction, relative radiation correction, bad pixel correction, horizontal image shift error correction (row interpolation), spectral position error correction (inter-row interpolation), inter-frame error correction (inter-frame interpolation correction), multi-slit hyperspectral data merging and reconstruction of the seven steps, the correction of the inconsistency error between multiple multi-slit hyperspectral data is completed, and the influence of the moving target is filtered out in the merging and reconstruction link, the accuracy and precision of the multi-slit hyperspectral reconstructed data are improved, and the precision of the multi-slit hyperspectral reconstructed data can reach within 3%; and the multi-slit hyperspectral data reconstruction processing method of the present application is simple in operation, and can realize efficient and real-time reconstruction processing of multi-slit hyperspectral data; therefore, the present application solves the technical problems that the prior art lacks efficient reconstruction processing method for data inconsistency between multi-slit hyperspectral data, and it is difficult to obtain high-precision multi-slit hyperspectral reconstructed data in real time.

[0044] (2) The multi-slit hyperspectral data reconstruction processing method of the present application realizes efficient error correction and efficient reconstruction processing of multi-slit hyperspectral data, which is simple in operation, relies on less storage resources and computing resources, and can quickly and efficiently complete the reconstruction processing of multi-slit hyperspectral data, realizing efficient, fast and real-time output of multi-slit hyperspectral reconstructed data. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 is a structural schematic diagram of the existing multi-slit dispersive hyperspectral imaging system mentioned in the background technology;

[0046] Figure 1 The explanations of the numbers in are as follows:

[0047] 1-Pre-mirror group, 2-Multi-slit template, 3-Collimating mirror group, 4-Dispersion element, 5-Imaging mirror group, 6-Detector.

[0048] Figure 2 is a flow chart of the multi-slit hyperspectral data reconstruction processing method embodiment of the present application. DETAILED DESCRIPTION

[0049] The present application will be described in detail below in combination with the drawings and specific embodiments.

[0050] Referring to Figure 2 , the multi-slit hyperspectral data reconstruction processing method of the present application comprises the following steps:

[0051] Step 1: Subtract the dark current correction coefficient from the multi-slit area array data obtained in the multi-slit dispersive hyperspectral imaging system to realize data dark current correction, and obtain the dark current corrected data; the above-mentioned dark current correction coefficient is measured;

[0052] Step 2: the dark current corrected data obtained in step 1 is multiplied by a relative radiation correction coefficient to achieve relative radiation correction of the data, and relative radiation corrected data is obtained; the relative radiation correction coefficient is measured;

[0053] Step 3: one-dimensional interpolation is used to correct the relative radiation corrected data obtained in step 2 to obtain bad pixel corrected data;

[0054] Step 4: one of the slits in the multi-slit dispersion type hyperspectral imaging system is defined as a reference slit, and the bad pixel corrected data of the reference slit in the bad pixel corrected data obtained in step 3 is defined as reference slit hyperspectral data; then, the reference slit hyperspectral data is used as a correction reference, and the bad pixel corrected data of the remaining slits except the reference slit in the bad pixel corrected data obtained in step 3 is corrected by using a horizontal interpolation method to obtain horizontal image shift error corrected data of the remaining slits;

[0055] Step 5: the reference slit hyperspectral data defined in step 4 is used as a correction reference, and the horizontal image shift error corrected data of the remaining slits obtained in step 4 is corrected by using an interline interpolation method to obtain spectral position error corrected data of the remaining slits;

[0056] Step 6: the reference slit hyperspectral data defined in step 4 is used as a correction reference, and the spectral position error corrected data of the remaining slits obtained in step 5 is corrected by using an interframe interpolation correction method to obtain correction completed data of the remaining slits;

[0057] Step 7: the correction completed data of the remaining slits obtained in step 6 is combined with the reference slit hyperspectral data defined in step 4 to obtain multi-slit hyperspectral reconstructed data, and the reconstruction process is completed.

[0058] Referring to Figure 2 In the embodiment, a double-slit dispersion type hyperspectral imaging system is selected, the size of the detector is 1024*256, the number of spectral bands is 80, and the number of bits of the detector is 12, so that the dynamic range of the detector is 0-4096, and two groups of hyperspectral data with a size of 1024*80 can be obtained. When the multi-slit hyperspectral data reconstruction processing method of the application is used to reconstruct the two groups of hyperspectral data obtained by the double-slit dispersion type hyperspectral imaging system selected in the embodiment, the steps are as follows:

[0059] Step 1: Subtract the dark current correction coefficient d from the double-slit area array data F obtained by the double-slit dispersion hyperspectral imaging system, i.e. the above-mentioned two groups of 1024x80 size hyperspectral data, to realize the dark current correction of the data, and obtain the dark current corrected data F';

[0060] Step 2: Multiply the dark current corrected data F' obtained in step 1 by the relative radiation correction coefficient x to realize the relative radiation correction of the data, and obtain the relative radiation corrected data F"; the above-mentioned relative radiation correction coefficient x is measured, and in the present embodiment, the measured relative radiation correction coefficient x is two groups of 1024x80 matrix data, and the value is between 0.95 and 1.05;

[0061] Step 3: Use one-dimensional interpolation method to correct the bad pixels of the relative radiation corrected data F" obtained in step 2, and obtain the bad pixel corrected data F''';In step 3 of the present embodiment, when the one-dimensional interpolation method is used to correct the bad pixels of the relative radiation corrected data F" obtained in step 2, only the interpolation in the row direction is considered, and the maximum number of consecutive bad pixels is set to four;For different numbers of consecutive bad pixels, the recovery algorithms for bad pixel correction are respectively:

[0062] 1) When it is a single bad pixel, the bad pixel correction result is 0.5A+0.5B;

[0063] 2) When it is double consecutive bad pixels, the bad pixel correction results from left to right are 0.67A+0.33B and 0.33A+0.67B, or 0.75A+0.25B and 0.25A+0.75B, respectively;

[0064] 3) When it is three consecutive bad pixels, the bad pixel correction results from left to right are 0.67A+0.33B, 0.5A+0.5B and 0.33A+0.67B, respectively;

[0065] 4) When it is four consecutive bad pixels, the bad pixel correction results from left to right are 0.2A+0.8B, 0.4A+0.6B, 0.6A+0.4B and 0.8A+0.2B, respectively;

[0066] A and B above represent the normal pixels adjacent to the left and right sides of the corresponding consecutive bad pixels, respectively;

[0067] Step 4: Define the slit 1 in the double-slit dispersion type hyperspectral imaging system as a reference slit, and define the bad pixel corrected data of the slit 1 in the bad pixel corrected data F''' obtained in step 3 as the reference slit hyperspectral data, denoted as D1; define the bad pixel corrected data of the slit 2 in the bad pixel corrected data obtained in step 3 as D2, that is, divide the bad pixel corrected data F''' obtained in step 3 into two parts D1 and D2; then take the reference slit hyperspectral data D1 as the correction reference, and use the row interpolation method to correct the transverse image shift error of D2, to obtain the transverse image shift error corrected data of the slit 2; the error sources causing the above transverse image shift error in step 4 mainly include optical system distortion error and deflection angle residual correction error and other error sources; in step 4 of the embodiment, when the row interpolation method is used to correct the transverse image shift error of the bad pixel corrected data of each slit except the reference slit in the bad pixel corrected data obtained in step 3, for the bad pixel corrected data of each slit, the row interpolation method is used to correct the data in units of rows, and the pixel data S i,j The formula [1] is used for calculation, and the formula [1] is:

[0068] S i,j = αC i,t + (1-α)C i,t+1 [1];

[0069] In the formula [1], S i,j represents the pixel data of the output coordinate point (i, j) after transverse image shift error correction; α represents a transverse image shift error correction coefficient; C i,t and C i,t+1 respectively represent the pixel data of the row coordinate i and the column coordinate t in the bad pixel corrected data of the slit obtained in step 3, and the pixel data of the row coordinate i and the column coordinate t+1; the values of α and t are determined by the size of the transverse image shift error of the slit relative to the reference slit; in the embodiment, the above row coordinate refers to the coordinate in the 1024 direction, and in the embodiment, the corresponding α and t of the slit 2 are determined according to the size of the transverse image shift error of the slit 2 relative to the slit 1, which are 1024×80 matrix data;

[0070] Step 5: using the line interpolation method to correct the spectral position error of the data of slit 2 after the correction of the transverse image motion error in step 4, and taking the reference slit hyperspectral data D1 defined in step 4 as the correction reference, to obtain the data of slit 2 after the correction of the spectral position error; the spectral position error in step 5 refers to the deviation of the spectral center wavelength between the data of each slit after the correction of the transverse image motion error in step 4 and the reference slit hyperspectral data defined in step 4; in this embodiment, the spectral position error in step 5 refers to the deviation of the spectral center wavelength between the data of slit 2 after the correction of the transverse image motion error in step 4 and the reference slit hyperspectral data D1 defined in step 4; in step 5 of this embodiment, the line interpolation method is used to correct the spectral position error of the data of each slit after the correction of the transverse image motion error in step 4, and for the data of each slit after the correction of the transverse image motion error, the line interpolation method is used to correct the data of the output coordinate point (i', j') in column units, and the pixel data S i ,j′ The formula [2] is used for calculation, and the formula [2] is as follows:

[0071] S i ,j′ = α' C k ' ,j′ + (1-α') C k ' +1,j′ [2];

[0072] In the formula [2], S i ,j′ represents the pixel data of the output coordinate point (i', j') after the correction of the spectral position error; α' represents the correction coefficient of the spectral position error; C k ' ,j′ and C k ' +1,j′ respectively represent the pixel data of the coordinate point with the row coordinate k and the column coordinate j' and the pixel data of the coordinate point with the row coordinate k+1 and the column coordinate j' in the data of the slit after the correction of the transverse image motion error in step 4; the values of α' and k are determined by the size of the spectral position error of the slit relative to the reference slit; in this embodiment, the column coordinate refers to the coordinate in the 80 direction, and in this embodiment, the corresponding α' and k of slit 2 are determined according to the size of the spectral position error of slit 2 relative to slit 1, which are all 1024*80 matrix data;

[0073] ​​​Step 6: using the frame interpolation correction method to correct the spectral position error corrected data of slit 2 obtained in step 5 with the slit 1 high spectral data D1 defined in step 4 as the correction reference, to obtain the correction completed data of slit 2; the error sources causing the above-mentioned frame error in step 6 mainly include the error sources such as the slit distance deviation caused by the assembly accuracy; the assembly accuracy and other reasons can cause the problem of non-integer time interval or horizontal spectral deviation of multi-slit hyperspectral data, and further cause the frame error; in this embodiment, when the frame interpolation correction method is used to correct the spectral position error corrected data of slit 2 obtained in step 5, in step 6, the calculation method of the above-mentioned frame interpolation correction method is to calculate the linear interpolation between the adjacent two frames by using formula [3], and the linear interpolation is taken as the correction completed data of the corresponding slit; the above-mentioned formula [3] is:

[0074] D f ′=α″D f,m +(1-α″)D f,m+1 [3];

[0075] In formula [3], D f ′ represents the correction completed data of slit f, f represents the slit serial number of the remaining slits, and the slit serial number defined as the reference slit in step 4 is set as 1, so f takes 2, 3, …, R, and R is the total number of slits in the multi-slit dispersive type hyperspectral imaging system; α″ represents the frame error correction coefficient of slit f; D f,m , D f,m+1 respectively represent the mth frame and the m+1th frame data of the spectral position error corrected data of slit f in the spectral position error corrected data of the remaining slits obtained in step 5, and m is the frame serial number; the values of α″ and m are determined by the frame error size of slit f relative to the reference slit; in this embodiment, according to the frame error size of slit 2 relative to slit 1, the corresponding α″ and m of slit 2 are determined as 0.978 and 1 respectively, and the correction completed data of slit 2 is D2′;

[0076] Step 7: merging and reconstructing the correction completed data D2′ of slit 2 obtained in step 6 and the reference slit hyperspectral data D1 defined in step 4 to obtain the double-slit hyperspectral reconstructed data, and completing the reconstruction process; the above-mentioned step 7 is specifically:

[0077] Step 7.1: setting a difference threshold μ; the difference threshold μ set in this embodiment is 300;

[0078] Step 7.2: according to the correction completed data of the remaining slits obtained in step 6, the reference slit hyperspectral data defined in step 4 and the difference threshold μ set in step 7.1, the merging switch matrix Pf f takes values ​​of 2, 3, ..., R, and is set accordingly; specifically, in this embodiment, the merging switch matrix P2 of slit 2 is set based on the correction completion data D2′ of slit 2 obtained in step 6, the reference slit hyperspectral data D1 defined in step 4, and the difference threshold μ set in step 7.1; step 7.2 is specifically as follows:

[0079] Step 7.2.1: Combine the switching matrix P of all slits except the reference slit. f The matrix size of f is set to be the same as the matrix size of the reference slit hyperspectral data defined in step 4, where f takes the values ​​2, 3, ..., R. In this embodiment, the matrix size of the merging switch matrix P2 of slit 2 is set to 1024×80.

[0080] Step 7.2.2: For each slit in the calibration data obtained in Step 6, subtract the corresponding data from the hyperspectral data of the reference slit defined in Step 4. Then, compare the absolute value of each difference with the difference threshold μ set in Step 7.1. If the absolute value of the difference is greater than the difference threshold μ set in Step 7.1, then the merging switching matrix P of the corresponding slit is... f The corresponding element is set to 0; if the absolute value of the above difference is less than or equal to the difference threshold μ set in step 7.1, then the merging switch matrix P of the corresponding slit is... f The corresponding element in the matrix is ​​set to 1; based on this principle, the merging switching matrix P for all slits except the reference slit is completed. f f takes the setting of all elements in 2, 3, ..., R; in this embodiment, the merging switch matrix P2 of slit 2 is a matrix in which the elements are composed of 0 and 1;

[0081] Step 7.3: Based on the merging switching matrix P of all slits except the reference slit set in Step 7.2. f f takes values ​​of 2, 3, ..., R. Using equation [4], the corrected data of the remaining slits obtained in step 6 are merged and reconstructed with the reference slit hyperspectral data defined in step 4 to obtain the reconstructed multi-slit hyperspectral data, thus completing the reconstruction process. Equation [4] is:

[0082] D″=(RD1-(D1-D′2)⊙P2-…(D1-D′ R )⊙P R ) / R [4];

[0083] In Equation [4]: ​​D″ represents the data after multi-slit hyperspectral reconstruction; D1 represents the reference slit hyperspectral data defined in step 4; D′2, ..., D′ Rrespectively represent the correction completion data of the slits 2, …, slit R obtained in step 6; P2, …, P R respectively represent the merging switch matrix of the slits 2, …, slit R set in step 7.2; ⊙ represents the Hadamard product of two matrices; the specific calculation formula of the double-slit hyperspectral reconstruction data D" in this embodiment is: D" = (2D1 - (D1 - D'2) ⊙ P2) / 2.

[0084] The multi-slit hyperspectral data reconstruction processing method of the application is simple in operation, can realize efficient and real-time reconstruction processing of multi-slit hyperspectral data, and also improves the accuracy and precision of the multi-slit hyperspectral reconstruction data when the data is reconstructed by the multi-slit hyperspectral data reconstruction processing method of the application, and the precision of the multi-slit hyperspectral reconstruction data can reach within 3%.

Claims

1. A multi-slit hyperspectral data reconstruction processing method, characterized in that, The method comprises the following steps: Step 1: dark current correction of multi-slit area array data obtained by a multi-slit dispersive hyperspectral imaging system is performed by subtracting a dark current correction coefficient, to obtain dark current corrected data; the dark current correction coefficient is measured; Step 2: relative radiation correction of the dark current corrected data obtained in step 1 is performed by multiplying the dark current corrected data by a relative radiation correction coefficient, to obtain relative radiation corrected data; the relative radiation correction coefficient is measured; Step 3: one-dimensional interpolation is used to correct bad pixels in the relative radiation corrected data obtained in step 2, to obtain bad pixel corrected data; Step 4: one of the slits in the multi-slit dispersive hyperspectral imaging system is defined as a reference slit, and the bad pixel corrected data of the reference slit in the bad pixel corrected data obtained in step 3 is defined as reference slit hyperspectral data; then, the reference slit hyperspectral data is used as a correction reference, and horizontal interpolation is used to correct the bad pixel corrected data of the remaining slits except the reference slit in step 3, to obtain horizontal image shift error corrected data of the remaining slits; Step 5: the reference slit hyperspectral data defined in step 4 is used as a correction reference, and interline interpolation is used to correct the horizontal image shift error corrected data of the remaining slits obtained in step 4, to obtain spectral position error corrected data of the remaining slits; Step 6: the reference slit hyperspectral data defined in step 4 is used as a correction reference, and interframe interpolation correction is used to correct the spectral position error corrected data of the remaining slits obtained in step 5, to obtain corrected data of the remaining slits; Step 7: the corrected data of the remaining slits obtained in step 6 and the reference slit hyperspectral data defined in step 4 are merged and reconstructed, to obtain multi-slit hyperspectral reconstructed data, and the reconstruction process is completed.

2. The multi-slit hyperspectral data reconstruction processing method according to claim 1, wherein in step 3, one-dimensional interpolation is used, only considering interpolation in the row direction, and the maximum number of consecutive bad pixels is set to four, and the relative radiation corrected data obtained in step 2 is corrected for bad pixels to obtain bad pixel corrected data; specifically: 1) when there is a single bad pixel, the bad pixel correction result is 0.5A+0.5B; 2) when there are two consecutive bad pixels, the bad pixel correction results from left to right are 0.67A+0.33B and 0.33A+0.67B, or 0.75A+0.25B and 0.25A+0.75B; 3) when there are three consecutive bad pixels, the bad pixel correction results from left to right are 0.67A+0.33B, 0.5A+0.5B, and 0.33A+0.67B; 4) when there are four consecutive bad pixels, the bad pixel correction results from left to right are 0.2A+0.8B, 0.4A+0.6B, 0.6A+0.4B, and 0.8A+0.2B. ​ A and B represent the left and right adjacent normal pixels of the continuous bad pixel, respectively.

3. The multi-slit hyperspectral data reconstruction processing method according to claim 2, characterized in that: In step 4, the row interpolation method is used to correct the horizontal image shift error of the bad pixel corrected data of each slit except the reference slit in the bad pixel corrected data obtained in step 3. For the bad pixel corrected data of each slit, the row interpolation method is used to correct the data in row units, and the pixel data S of the output coordinate point (i, j) after correction is obtained. i,j The calculation is performed using Equation [1] as follows: S i,j = aC i,t + (1 - a)C i,t+1 [1] In formula [1]: S i,j represents the pixel data of the output coordinate point (i, j) after correction of the lateral image shift error; a represents the correction coefficient of the lateral image shift error; C i,t , C i,t+1 respectively represent the pixel data of the row coordinate i and the column coordinate t and the pixel data of the row coordinate i and the column coordinate t+1 in the bad pixel corrected data of the slit obtained in step 3; the values of a and t are determined by the measured lateral image shift error of the slit relative to the reference slit.

4. The multislab hyper spectral data reconstruction method of claim 1, wherein: The error sources causing the lateral image motion error in step 4 include optical system distortion error and bias flow angle residual error correction error.

5. The multi-slit hyperspectral data reconstruction processing method according to claim 3, characterized in that: In step 5, the spectral position error correction is performed on the data of the remaining slits after the lateral image shift error correction in step 4 by using the interline interpolation method. For the data of each slit after the lateral image shift error correction, the interline interpolation method is used to correct the data in column units, and the output coordinate point (i', j') of the pixel data S is obtained i ,j′ The formula [2] is used for calculation, and the formula [2] is as follows:​ S i ,j′ = a' C k ,j′ + (1 - a') C k +1,j′ [2]​​​ In formula [2], S i ,j′ represents the pixel data of the output coordinate point (i', j') after correction of the spectral position error; a' represents the correction coefficient of the spectral position error; C k ,j′ , C k +1,j′ respectively represent the pixel data of the line coordinate k and the column coordinate j' and the pixel data of the line coordinate k+1 and the column coordinate j' in the transverse image shift error corrected data of the slit obtained in step 4; the values of a' and k are determined by the size of the spectral position error of the slit relative to the reference slit.​​​ 6. The multi-slit hyperspectral data reconstruction processing method according to claim 1, characterized in that: In step 5, the spectral position error refers to the spectral center wavelength deviation between the data after correction of the lateral image motion error of each slit in step 4 and the reference slit hyperspectral data defined in step 4.

7. The multi-slit hyperspectral data reconstruction processing method according to claim 5, characterized in that: In step 6, the calculation method of the frame interpolation correction method is to calculate the linear interpolation between two adjacent frames by using formula [3], and the linear interpolation is taken as the corrected data of the corresponding slit; the formula [3] is: D f ′= a" D f,m +(1 - a") D f,m+1 [3] In formula [3], D f f represents the slit sequence number of the remaining slits, the slit sequence number defined as the reference slit in step 4 is set as 1, and f takes 2, 3, …, R, where R is the total number of slits in the multi-slit dispersion type hyperspectral imaging system; and a" represents the inter-frame error correction coefficient of the slit f. D f,m 、D f,m+1 respectively represent the mth frame and the m+1th frame of the spectral position error corrected data of the slit f in the spectral position error corrected data of the remaining respective slits obtained in step 5, and the value of m and a" are determined by the measured interframe error of the slit f relative to the reference slit.

8. The multislab hyper spectral data reconstruction method of claim 1, wherein: The error sources causing the frame error in step 6 include the slit distance deviation caused by assembly accuracy.

9. The multi-slit hyperspectral data reconstruction processing method according to claim 7, characterized in that: The step 7 is specifically: Step 7.1: set a difference threshold μ; Step 7.2: Based on the corrected data of the remaining slits obtained in Step 6, the reference slit hyperspectral data defined in Step 4, and the difference threshold μ set in Step 7.1, the merging switch matrix P of the remaining slits except the reference slit is set as follows: f f = 2, 3,..., R, is set. Step 7.3: The combined switch matrix P of the rest of the slits except the reference slit is set according to step 7.2 f The rest of the slit correction completed data obtained in step 6 is combined and reconstructed with the reference slit hyperspectral data defined in step 4 by using formula [4] with f taking 2, 3, …, R, to obtain multi-slit hyperspectral reconstructed data, and the reconstruction process is completed; the formula [4] is: D" = (RD1- (D1-D'2) 0 P2-... (D1-D'N) 0 PN) / R [4]; where D'N is the desired depth of the Nth layer, and P is the power of the Nth layer. R ) 0 P R ) / R [4]. In formula [4]: D" represents the multi-slit hyperspectral reconstructed data; D1 represents the reference slit hyperspectral data defined in step 4; D'2,..., D' R represent the corrected data of slits 2,..., slits R obtained in step 6 respectively; P2,..., P R represent the combined switch matrix of slits 2,..., slits R set in step 7.2 respectively; and represents the Hadamard product of two matrices. R In formula [4]: D" represents the multi-slit hyperspectral reconstructed data; D1 represents the reference slit hyperspectral data defined in step 4; D'2,..., D' R represent the corrected data of slits 2,..., slits R obtained in step 6 respectively; P2,..., P R represent the combined switch matrix of slits 2,..., slits R set in step 7.2 respectively; and represents the Hadamard product of two matrices. R In formula [4]: D" represents the multi-slit hyperspectral reconstructed data; D1 represents the reference slit hyperspectral data defined in step 4; D'2,..., D' R represent the corrected data of slits 2,..., slits R obtained in step 6 respectively; P2,..., P R represent the combined switch matrix of sl 10. The multi-slit hyperspectral data reconstruction processing method according to claim 9, characterized in that: The step 7.2 is specifically: Step 7.2.1: The combined switch matrix P of the rest of the slits except the reference slit f The matrix size of f = 2, 3, …, R is set to be the same as the matrix size of the reference slit hyperspectral data defined in step 4. Step 7.2.2: In the correction completion data of each of the remaining slits obtained in step 6, each of the data is respectively subtracted from the corresponding data in the reference slit hyperspectral data defined in step 4, and then the absolute value of each difference is respectively compared with the difference threshold μ set in step 7.1; if the absolute value of the difference is greater than the difference threshold μ set in step 7.1, the corresponding position element in the merging switch matrix P f of the corresponding slit is set to 0; if the absolute value of the difference is less than or equal to the difference threshold μ set in step 7.1, the corresponding position element in the merging switch matrix P f of the corresponding slit is set to 1; according to this principle, the setting of all elements in the merging switch matrix P f , f takes 2, 3, …, R is completed.

Images