A shadow detection and correction method for infrared imaging spectrometer

Abstract

The present invention relates to a shadow detection and correction method for an infrared imaging spectrometer, the method comprising: S1, principal component analysis and band combination are respectively performed on VIS / NIR image data, and two normalizations are respectively performed on the first principal component result PC1* and the band combination result BC10*; S2, SId image and SIs image representing two shadow indexes are constructed using the normalized results of the first principal component and the band combination; S3, SId image and SIs image are respectively threshold segmented using the Otsu algorithm to obtain two rough shadow detection result images; S4, the shadow detection results of SId and SIs are further refined using the absolute value of z-score to obtain precise shadow detection results; S5, shadow effects of VIS / NIR and SWIR band spectra are corrected using the precise shadow detection results and spectral data detected by the infrared imaging spectrometer VNIS. Compared with the prior art, the present invention has higher detection accuracy for shadows existing in the hyperspectral data of the infrared imaging spectrometer, and can correct the shadow effects of the VIS / NIR and SWIR band spectra.

Description

Technical Field

[0001] The invention relates to the field of spectral remote sensing data processing, and in particular to a shadow detection and correction method for an infrared imaging spectrometer. Background Art

[0002] Chang'e-4 is a lunar probe launched during the second phase of China's lunar exploration program. It is also the first probe to land on the far side of the moon. It has achieved the first soft landing and patrol survey on the far side of the moon, which is of great significance and far-reaching impact. Among the multiple scientific payloads carried by the Chang'e-4 probe, the infrared imaging spectrometer is the only scientific instrument that serves the detection and research of lunar mineral composition, and can obtain lunar soil hyperspectral images and infrared spectral data with millimeter-level spatial resolution.

[0003] The infrared imaging spectrometer (Visible and Near-Infrared Imaging Spectrometer, VNIS) has a spectral range of 450-2400nm and a spectral resolution of 2-12nm. It has on-orbit calibration and dust-proof functions, can adapt to the temperature environment of -20~+55℃ operation and -50~+70℃ storage, and weighs less than 6kg. It is a high-performance, lightweight, and highly integrated instrument. The infrared imaging spectrometer can image in the visible and near-infrared band (VIS / NIR, 450-950nm) and detect spectral data in the shortwave infrared band (SWIR, 900-2400nm).

[0004] Due to the undulating terrain on the lunar surface, shadows are common in the hyperspectral images detected by the infrared imaging spectrometer VNIS. Since the shadow area loses radiation information, it cannot reflect the true reflectivity of the lunar surface, and thus cannot accurately analyze the composition and properties of the lunar surface materials. Therefore, it is necessary to detect and correct the shadows in the image.

[0005] The existing shadow detection and correction of remote sensing images mainly uses edge detection, region growing and other algorithms for detection of Earth remote sensing images. However, due to the weak texture of the lunar surface, the shadows on lunar remote sensing images are mainly caused by topography and micro-scale roughness, which leads to irregular shadow shapes and usually small areas. Traditional detection algorithms are difficult to apply to the detection of lunar remote sensing images. In addition, in lunar geological research, the spectrum around 1000nm and 2000nm is crucial for inverting mineral composition, while the wavelength of VIS / NIR image data detected by infrared imaging spectrometers is between 450nm and 950nm, so it is impossible to correct the shadow effect in space. Summary of the invention

[0006] The purpose of the present invention is to overcome the defect that the shadow detection and correction algorithm in the above-mentioned prior art is difficult to be applied to the hyperspectral image of the lunar surface detected by the infrared imaging spectrometer, and to provide a shadow detection and correction method for the infrared imaging spectrometer.

[0007] The purpose of the present invention can be achieved by the following technical solutions:

[0008] The present invention provides a shadow detection and correction method for an infrared imaging spectrometer, comprising the following steps:

[0009] Step S1: perform principal component analysis and band combination on the VIS / NIR image data, and perform two types of normalization on the first principal component result PC1* and the band combination result BC10*;

[0010] Step S2: constructing the SId image and SIs image representing two shadow indices using the normalized result of the first principal component and the band combination;

[0011] Step S3: Use the Otsu algorithm to perform threshold segmentation on the SId image and the SIs image respectively to obtain two rough shadow detection result images;

[0012] Step S4: for the shadow detection results of SId and SIs, the absolute value of z-score is used to further refine the detection to obtain the shadow precision detection results;

[0013] Step S5: Correct the shadow effects of the VIS / NIR and SWIR band spectra using the shadow precision detection results and the spectral data detected by the infrared imaging spectrometer VNIS.

[0014] Preferably, step S1 also includes: after performing principal component analysis on the VIS / NIR image data, if the shadow is located at the negative small value end of the first principal component, it is necessary to calculate the inverse of the first principal component; if the shadow is located at the positive large value end, no change is made and the result is recorded as PC1*.

[0015] Preferably, the band combination in step S1 is specifically:

[0016] The VIS / NIR image data are standardized band by band, and the pixel values ​​of each band after standardization are arranged from large to small, and the first 10% and the last 10% of pixels are selected as "non-shadow" and "shadow", respectively;

[0017] Calculate the difference between the mean values ​​of “non-shadow” and “shadow” pixels in each band, select the top 10 standardized bands with the largest difference and add them up, and calculate the opposite of the accumulated result as the band combination result, recorded as BC10*.

[0018] Preferably, the band standardization expression is:

[0019]

[0020] Where, Band i and Respectively represent the pixel values ​​of the i-th band before and after normalization, μ i and σ i Represents the pixel mean and standard deviation of the i-th band.

[0021] Preferably, the two normalizations in step S1 are:

[0022] The first method is to directly normalize PC1* and BC10* to the interval [0,1], and the results are recorded as PC1d and BC10d;

[0023] The second method is to assign 0 to the negative pixels of PC1* and BC10* and then normalize them to the interval [0,1]. The results are recorded as PC1s and BC10s.

[0024] Preferably, the step S2 further comprises: multiplying the normalized results PC1d and BC10d to obtain the SId image, and multiplying the normalized results PC1s and BC10s to obtain the SIs image.

[0025] Preferably, in step S4, the process of refining the detection includes:

[0026] Step S41: taking the non-shadow area detected by the SId image as the initial non-shadow area;

[0027] Step S42: taking the shadow area detected by the SIs image as the initial shadow area;

[0028] Step S43: For the shadow pixels detected by the SId image but not detected by the SIs image, calculate the absolute value of the z-score of the pixel with the average value of the non-shadow area detected by the SId image, denoted as z1, and the absolute value of the z-score of the pixel with the average value of the shadow area detected by the SIs image, denoted as z2; if z1 is greater than z2, the pixel is recorded as a shadow pixel, otherwise it is recorded as a non-shadow pixel, and the determination is terminated to obtain the precise shadow detection result.

[0029] Preferably, the calculation formulas of z1 and z2 in step S43 are expressed as follows:

[0030]

[0031]

[0032] Where PC1 * (i,j) is the value of pixel (i,j) in PC1*, μ f and σf is the mean and standard deviation of the non-shadow area detected by SId in PC1*, μ s and σ s are the mean and standard deviation of the shaded area detected by SIs in PC1*.

[0033] Preferably, in step S5, the process of correcting the shadow effect includes:

[0034] Step S51: using 10 overlapping bands of SWIR and VIS / NIR at 900-950nm, calculating the average of the reflectivity difference, adding the reflectivity of each band of VIS / NIR to the average, and obtaining the adjusted VIS / NIR spectrum;

[0035] Step S52: Calculate the proportion of the shadow area and the non-shadow area in the SWIR field of view, and for each band of the VIS / NIR image, calculate the average brightness value of the shadow area and the non-shadow area in the field of view;

[0036] Step S53: using the proportion and average brightness value of the shadow area and the non-shadow area in the SWIR field of view, calculate the coefficient of the average brightness reduction in the field of view due to the presence of shadows in each VIS / NIR band;

[0037] Step S54: dividing the reflectivity of each band of VIS / NIR by the coefficient to obtain a VIS / NIR spectrum corrected for the shadow effect;

[0038] Step S55: using the 10 overlapping bands of SWIR and VIS / NIR, calculate the average of the difference between the reflectance of VIS / NIR and SWIR for correcting the shadow effect, add the reflectance of each band of SWIR to this average, and obtain the SWIR spectrum for correcting the shadow effect.

[0039] Preferably, the calculation formula of the coefficient in step S53 is expressed as follows:

[0040]

[0041] Where Coeff i is the coefficient of the ith band, DNs i and DNf i are the average brightness values ​​of the shadow area and non-shadow area in the SWIR field of view of the i-th band, respectively. RATIOs and RATIOf are the proportions of shadow and non-shadow in the SWIR field of view, respectively.

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

[0043] The shadow detection and correction method adopted in the present invention overcomes the defect that the existing shadow detection and correction methods are difficult to apply to the hyperspectral images of the lunar surface detected by the infrared imaging spectrometer. Through coarse shadow detection and fine shadow detection, the shadow detection accuracy of the VNIS hyperspectral data of the infrared imaging spectrometer is improved, and by correcting the shadow effect of the VIS / NIR and SWIR band spectra, the accuracy of the spectral data is improved, which is convenient for subsequent further research. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flow chart of the shadow detection and correction method for infrared imaging spectrometer of the present invention;

[0045] Figure 2 The image corresponding to the parameters generated during the algorithm implementation;

[0046] Figure 3 The images of SId index and SIs index constructed in the rough detection stage of the algorithm and their corresponding rough shadow detection result images;

[0047] Figure 4 It is a comparison between the shadow result image in the algorithm's fine detection stage and the original image of the embodiment;

[0048] Figure 5 The original spectra of VIS / NIR and SWIR bands and the spectra after correction of shadow effects. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0050] Example

[0051] like Figure 1 As shown, the shadow detection and correction method for the infrared imaging spectrometer of the present invention comprises the following steps:

[0052] Step S1: perform principal component analysis and band combination on the VIS / NIR image data, and perform two types of normalization on the first principal component and the band combination results;

[0053] In step S1, first perform principal component analysis on the VIS / NIR image. When the shadow is at the negative small value end of the first principal component, it is necessary to calculate the opposite of the first principal component. If the shadow is at the positive large value end, no transformation is performed. The processed first principal component is recorded as PC1*. Then the VIS / NIR image data is standardized band by band, and the pixel values ​​of each band after standardization are arranged from large to small. The first 10% and the last 10% of the pixels are selected as "non-shadow" and "shadow" respectively, and the difference between the mean values ​​of the "non-shadow" and "shadow" pixels of each band is calculated. The first 10 standardized bands with the largest difference are selected and accumulated, and the opposite of the accumulated result is calculated as the band combination result, recorded as BC10*. The formula for standardization is as follows:

[0054]

[0055] Where, Band i and Respectively represent the pixel values ​​of the i-th band before and after normalization, μ i and σ i Represents the pixel mean and standard deviation of the i-th band;

[0056] Next, two normalizations are performed on PC1* and BC10*. The first is to directly normalize PC1* and BC10* to the [0,1] interval, and the results are recorded as PC1d and BC10d. The second is to assign 0 to the negative pixels of PC1* and BC10* and then normalize them to the [0,1] interval, and the results are recorded as PC1s and BC10s.

[0057] Step S2: construct two shadow indices, SId and SIs, using the normalized results of the first principal component and the band combination;

[0058] Using PC1d, PC1s, BC10d and BC10s calculated in step S1, PC1d and BC10d are multiplied to obtain the shadow index SId, and PC1s and BC10s are multiplied to obtain the shadow index SIs.

[0059] Step S3: Perform threshold segmentation on the SId image and the SIs image using the Otsu algorithm to obtain two rough shadow detection result images;

[0060] For the SId index image and the SIs index image, a threshold is calculated respectively using the Otsu algorithm, and the two index images are segmented according to their respective thresholds to obtain two rough shadow detection result images.

[0061] Step S4: for the shadow detection results of SId and SIs, the absolute value of z-score is used to further refine the detection to obtain the shadow precision detection results;

[0062] The rough shadow detection results obtained by SId and SIs and the absolute value of z-score are used to perform precise shadow detection.

[0063] Step S41: taking the non-shadow area detected by SId as the initial non-shadow area;

[0064] Step S42: taking the shadow area detected by SIs as the initial shadow area;

[0065] Step S43: for the shadow pixels detected by SId but not detected by SIs, calculate the absolute value of the z-score between the pixel and the average value of the non-shadow area detected by SId, recorded as z1, and the absolute value of the z-score between the pixel and the average value of the shadow area detected by SIs, recorded as z2. If z1 is greater than z2, the pixel is recorded as a shadow pixel, otherwise it is recorded as a non-shadow pixel, and the determination is terminated to obtain the shadow precision detection result.

[0066] The calculation formulas of z1 and z2 are as follows:

[0067]

[0068]

[0069] Where PC1 * (i,j) is the value of pixel (i,j) in PC1*, μ f and σ f is the mean and standard deviation of the non-shadow area detected by SId in PC1*, μ s and σ s is the mean and standard deviation of the shaded area detected by SIs in PC1*;

[0070] Step S5: Correct the shadow effect of the VIS / NIR and SWIR band spectra using the shadow precision detection results and the spectrum data detected by VNIS;

[0071] The shadow effects in the VIS / NIR and SWIR bands are corrected by using the shadow precision detection result obtained in step S4 and the spectral data of the VIS / NIR and SWIR bands detected by the VNIS.

[0072] Step S51: using 10 overlapping bands of SWIR and VIS / NIR at 900-950nm, calculating the average of the reflectivity difference, adding the reflectivity of each band of VIS / NIR to the average, and obtaining the adjusted VIS / NIR spectrum;

[0073] Step S52: Calculate the proportion of the shadow area and the non-shadow area in the SWIR field of view, and for each band of the VIS / NIR image, calculate the average brightness value of the shadow area and the non-shadow area in the field of view;

[0074] Step S53: using the proportion and average brightness value of the shadow area and the non-shadow area in the SWIR field of view, calculate the coefficient of the average brightness reduction in the field of view due to the presence of shadows in each VIS / NIR band;

[0075] Step S54: dividing the reflectivity of each band of VIS / NIR by the coefficient to obtain a VIS / NIR spectrum corrected for the shadow effect;

[0076] Step S55: using the 10 overlapping bands of SWIR and VIS / NIR, calculating the average of the difference between the reflectance of VIS / NIR and SWIR for correcting the shadow effect, adding the reflectance of each band of SWIR to the average, and obtaining the SWIR spectrum for correcting the shadow effect;

[0077] The calculation formula of the coefficient is as follows:

[0078]

[0079] Where Coeff i refers to the coefficient of the ith band, DNs i and DNf i It refers to the average brightness value of the shadow area and non-shadow area in the SWIR field of view of the i-th band. RATIOs and RATIOf refer to the proportion of shadow and non-shadow in the SWIR field of view.

[0080] This example uses the 0003 image detected by Chang'e 4 and Yutu 2 during the first lunar day to carry out the experiment. The 760nm band image of this image can be found in Figure 2 (a), the main technical parameters of the infrared imaging spectrometer VNIS are shown in Table 1.

[0081] Table 1

[0082]

[0083] The flowchart of the shadow detection and correction algorithm for infrared imaging spectrometer hyperspectral data is shown in Figure 1 The specific implementation process is as follows:

[0084] First, perform principal component analysis on the VIS / NIR image. The first principal component image is shown in Figure 2 (b) In this embodiment, the shadow is located at the negative end of the first principal component, so its inverse is calculated to obtain PC1*. The PC1* image can be found in Figure 2(c). Then the VIS / NIR image data is standardized band by band, and the pixel values ​​of each band after standardization are arranged from large to small. The first 10% and the last 10% of the pixels are selected as "non-shadow" and "shadow", respectively, and the difference between the mean values ​​of "non-shadow" and "shadow" pixels of each band is calculated. The difference calculation results are (3.3148, 3.3047, 3.2656, 3.2924, 3.2946, 3.2464, 3.271, 3.2704, 3.2735, 3.2997, 3.2872, 3.2441, 3.2593, 3.28 35,3.251,3.2356,3.26,3.2606,3.2533,3.2732,3.3001,3.2799,3.265,3.284,3.321,3.3045,3.2733,3.2698,3.2984,3.3044,3.2824,3.2687,3.292,3.3042,3.2831,3.2701,3.2831,3.3258,3.3645,3.3794,3.3577,3.3267,3.309 7,3.3102,3.3198,3.3119,3.2841,3.2638,3.2572,3.2674,3.2829,3.2872,3.2729,3.2583,3.2533,3.262,3.2827,3.2947,3.2991,3.2919,3.2775,3.2775,3.2892,3.3102,3.334,3.3422,3.3388,3.3262,3.3169,3.3193,3.325,3. 3.3427,3.3575,3.3668,3.3683,3.3592,3.35,3.3366,3.3262,3.3286,3.3367,3.3447,3.3363,3.3414,3.3332,3.3254,3.3132,3.306,3.3018,3.3095,3.318,3.3274,3.3355,3.3432,3.3503,3.3532,3.3509,3.3397,3.3376,3.3322). Select the first 10 standardized bands with the largest difference. The band numbers of these ten bands are (40, 75, 74, 39, 76, 41, 73, 96, 97, 95). Add up the standardized pixel values ​​of these ten bands and calculate their opposite numbers to get the band combination result BC10*. BC10* see Figure 2 (d).

[0085] Two normalizations are performed on PC1* and BC10* to obtain PC1d, BC10d, PC1s and BC10s. For images of PC1d, BC10d, PC1s and BC10s, see Figure 2 (e), (f), (g) and (h).

[0086] Multiply PC1d and BC10d to get SId. See SId image for details. Figure 3 (a). SIs are obtained by multiplying PC1s and BC10s. See SIs for the image. Figure 3 (b).

[0087] The Otsu algorithm is used to calculate the threshold segmentation SId and SIs, and the SId shadow rough detection results and SIs shadow rough detection results are obtained. The calculated threshold for segmenting SId is 0.4782, and the SId shadow rough detection results are shown in Figure 3 (c). The calculated threshold for segmenting SIs is 0.2471. The SIs shadow rough detection results are shown in Figure 3 (d).

[0088] Use the rough shadow detection results and the absolute value of the z-score obtained by SId and SIs to perform precise shadow detection. First, take the non-shadow area detected by SId as the initial non-shadow area, and the shadow area detected by SIs as the initial shadow area. Then, for the shadow pixels detected by SId but not detected by SIs, calculate the absolute value of the z-score of the average of the non-shadow area detected by SId, and get z1, and the absolute value of the z-score of the average of the shadow area detected by SIs, and get z2. Here, taking the pixel in the 97th row and the 122nd column as an example, z1 is calculated to be 4.2362, and z2 is 1.8817. z1 is greater than z2, so the pixel is taken as a shadow pixel. Calculate the shadow pixels detected by SId but not detected by SIs one by one until all calculations are completed to obtain the precise shadow detection results, see. Figure 4 (b).

[0089] The shadow effects in the VIS / NIR and SWIR bands are corrected using the shadow precision detection results and the spectral data of the VIS / NIR and SWIR bands detected by VNIS.

[0090] 1) Using 10 overlapping bands of SWIR and VIS / NIR at 900-950nm, calculate the average of the reflectivity difference. The calculation result of this embodiment is 0.0089. Add 0.0089 to the reflectivity of each band of VIS / NIR to obtain the adjusted VIS / NIR spectrum, see Figure 5 .

[0091] 2) Calculate the proportion of shadow area in SWIR field of view RATIOs = 0.3647 and the proportion of non-shadow area RATIOf = 0.6353. For each band of VIS / NIR image, calculate the average brightness value of shadow area and non-shadow area in field of view. The average brightness value of shadow area is (0.0047092, 0.0035426, 0.0030663, 0.0032207, 0.0029567, 0.0025, 0.0028319, 0.0026762, 0.0031372, 0.00325, 0.0033633, 0.002748, 0.0029564, 0.0032778, 0.002683 2,0.0029362,0.0029117,0.0028074,0.003115,0.0032975,0.0036921,0.0034682,0.0033981,0.0038449,0.0043261,0.0041382,0.0039386,0.0 0375,0.0041765,0.0041216,0.0037365,0.0038622,0.0040505,0.0036067,0.0036002,0.0034102,0.0036592,0.0050345,0.0055046,0.005886, 0.0054815,0.0047827,0.0043548,0.0046047,0.0050858,0.0047898,0.0042265,0.0040624,0.0041351,0.004316,0.0041576,0.0042052,0.004 2668,0.0038299,0.0037354,0.0040293,0.0041844,0.004072,0.0042274,0.0041461,0.0038977,0.0039563,0.0042572,0.0044424,0.0045454, 0.004634,0.004421,0.0044644,0.0043423,0.0042027,0.0041891,0.0041422,0.004277,0.0046661,0.0045861,0.0045641,0.0044338,0.00427 44,0.0040699,0.0038731,0.0039178,0.0042113,0.0041016,0.0040707,0.0038221,0.0037342,0.0036633,0.0035468,0.0034761,0.0036623,0.0038972,0.004101,0.0043936,0.0044514,0.004336,0.0042143,0.0039633,0.0039322,0.0037471,0.0037213), the average brightness value of the non-shadow area is (0.011986,0.010436,0.010483,0.010314,0.0094695,0.009605,0.010026,0.0093468,0.010143,0.010194,0.010289,0.010534,0.010925,0.010616,0.010022,0. 011071,0.01101,0.010786,0.011397,0.011774,0.011882,0.011841,0.011974,0.012544,0.012629,0.012624,0.012916,0.013184,0.013225,0 .013228,0.012839,0.01333,0.013209,0.012271,0.0124,0.012527,0.012728,0.013621,0.013409,0.013227,0.01286,0.012592,0.012647,0.0 13098,0.01356,0.013192,0.012869,0.012965,0.013307,0.013355,0.012815,0.012651,0.01268,0.012387,0.012324,0.012715,0.012571,0.0 12182,0.011968,0.011916,0.01175,0.011986,0.012422,0.012207,0.011856,0.011604,0.011239,0.011285,0.011216,0.011024,0.010972,0. 010709,0.010626,0.010781,0.010484,0.010491,0.010316,0.010203,0.010055,0.00987,0.009793,0.0098761,0.0098429,0.0096482,0.00935 33,0.0093446,0.0093487,0.0092946,0.009303,0.0094556,0.0095803,0.0096074,0.0097176,0.0096351,0.0093986,0.0090716,0.0087611,0.0087411,0.008593,0.0085697). .

[0092] 3) Calculate the coefficient Coeff of the average brightness reduction in the field of view due to the presence of shadows in each VIS / NIR band i is(0.77856,0.75908,0.74195,0.74916,0.74915,0.7302,0.73829,0.73969,0.74807,0.75155,0.75449,0.73041,0.73397,0.74788,0.73292,0.73199,0.73172,0.7302,0.73495,0.73741,0.7486,0.7421,0.73877,0.74706,0.7602 ,0.75483,0.74648,0.73901,0.75045,0.74891,0.74141,0.74094,0.74711,0.74247,0.74116,0.73455,0.74012,0.77007,0.785,0.79757,0.79074,0.7738,0.76086,0.76349,0.77206,0.7677,0.75505,0.74954,0.7486,0.75313, 0.7536,0.7565,0.758,0.74804,0.74581,0.75084,0.75667,0.75718,0.76409,0.76217,0.75625,0.75565,0.76027,0.76799,0.7751,0.78091,0.77874,0.77955,0.77647,0.77431,0.77452,0.77634,0.78207,0.79313,0.79481,0 .79394,0.79202,0.78807,0.7829,0.77839,0.78118,0.79079,0.78725,0.78915,0.78431,0.78102,0.77819,0.77445,0.77155,0.77653,0.78364,0.79095,0.80017,0.80377,0.80353,0.8047,0.80026,0.79934,0.79431,0.79365)

[0093] 4) Divide the reflectivity of each VIS / NIR band by the corresponding coefficient Coeff i , and obtain the VIS / NIR spectrum corrected for the shadow effect, see Figure 5 .

[0094] 5) Using the 10 overlapping bands of SWIR and VIS / NIR, the average of the difference between the reflectance of VIS / NIR and SWIR for correcting the shadow effect is calculated. The calculation result of this embodiment is 0.0143. The reflectance of each band of SWIR is added with 0.0143 to obtain the SWIR spectrum for correcting the shadow effect. Figure 5 .

[0095] Figure 4 (a) shows the original image of the 760nm band of the present embodiment. Figure 4 (b) is the shadow detection result. By comparing the two images, we can see that the algorithm has a high accuracy in shadow detection. Figure 5 It can be seen that the algorithm can correct the shadow effect of VIS / NIR and SWIR band spectra.

[0096] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A shadow detection and correction method for an infrared imaging spectrometer, characterized in that: The following steps are involved: Step S1: perform principal component analysis and band combination on the VIS / NIR image data, and perform two types of normalization on the first principal component result PC1* and the band combination result BC10*; Step S2: constructing the SId image and SIs image representing two shadow indices using the normalized result of the first principal component and the band combination; Step S3: Use the Otsu algorithm to perform threshold segmentation on the SId image and the SIs image respectively to obtain two rough shadow detection result images; Step S4: for the shadow detection results of SId and SIs, further refine the detection using the absolute value of the z-score to obtain the shadow precision detection results, specifically including: Step S41: taking the non-shadow area detected by the SId image as the initial non-shadow area; Step S42: taking the shadow area detected by the SIs image as the initial shadow area; Step S43: for the shadow pixels detected by the SId image but not detected by the SIs image, calculate the absolute value of the z-score of the shadow pixel detected by the SId image and the average value of the z-score of the non-shadow area detected by the SId image, recorded as z1, and the absolute value of the z-score of the shadow pixel detected by the SIs image and the average value of the z-score of the shadow area detected by the SIs image, recorded as z2; if z1 is greater than z2, the pixel is recorded as a shadow pixel, otherwise it is recorded as a non-shadow pixel, and the determination is terminated to obtain the shadow precision detection result; Step S5: Correcting the shadow effect of the VIS / NIR and SWIR band spectra using the shadow precision detection results and the spectral data detected by the infrared imaging spectrometer VNIS, specifically including: Step S51: using 10 overlapping bands of SWIR and VIS / NIR at 900-950nm, calculating the average of the reflectivity difference, adding the reflectivity of each band of VIS / NIR to the average, and obtaining the adjusted VIS / NIR spectrum; Step S52: Calculate the proportion of the shadow area and the non-shadow area in the SWIR field of view, and for each band of the VIS / NIR image, calculate the average brightness value of the shadow area and the non-shadow area in the field of view; Step S53: using the proportion and average brightness value of the shadow area and the non-shadow area in the SWIR field of view, calculate the coefficient of the average brightness reduction in the field of view due to the presence of shadows in each VIS / NIR band; Step S54: dividing the reflectivity of each band of VIS / NIR by the coefficient to obtain a VIS / NIR spectrum corrected for the shadow effect; Step S55: using the 10 overlapping bands of SWIR and VIS / NIR, calculate the average of the difference between the reflectance of VIS / NIR and SWIR for correcting the shadow effect, add the reflectance of each band of SWIR to this average, and obtain the SWIR spectrum for correcting the shadow effect.

2. The shadow detection and correction method for infrared imaging spectrometer according to claim 1, characterized in that: The step S1 also includes: after performing principal component analysis on the VIS / NIR image data, if the shadow is located at the negative small value end of the first principal component, it is necessary to calculate the opposite of the first principal component; if the shadow is located at the positive large value end, no change is made and the result is recorded as PC1*.

3. The shadow detection and correction method for infrared imaging spectrometer according to claim 1, characterized in that: The band combination in step S1 is specifically: The VIS / NIR image data is standardized band by band, and the pixel values ​​of each band after standardization are arranged from large to small, and the first 10% and the last 10% of pixels are selected as "non-shadow" and "shadow" respectively; Calculate the difference between the mean values ​​of "non-shadow" and "shadow" pixels in each band, select the top 10 standardized bands with the largest difference and add them up, and calculate the opposite of the accumulated result as the band combination result, recorded as BC10*.

4. The shadow detection and correction method for infrared imaging spectrometer according to claim 3, characterized in that: The band normalization expression is: ， In the formula, and Respectively represent The pixel values ​​of the bands before and after normalization, and Indicates The pixel mean and standard deviation for each band.

5. The shadow detection and correction method for infrared imaging spectrometer according to claim 1, characterized in that: The two normalizations in step S1 are: The first method is to directly normalize PC1* and BC10* to the interval [0,1], and the results are recorded as PC1d and BC10d; The second method is to assign 0 to the negative pixels of PC1* and BC10* and then normalize them to the interval [0,1]. The results are recorded as PC1s and BC10s.

6. The shadow detection and correction method for infrared imaging spectrometer according to claim 5, characterized in that: The step S2 further includes: multiplying the normalized results PC1d and BC10d to obtain the SId image, and multiplying the normalized results PC1s and BC10s to obtain the SIs image.

7. The shadow detection and correction method for infrared imaging spectrometer according to claim 1, characterized in that: The calculation formulas of z1 and z2 in step S43 are as follows: ， ， In the formula, It's pixels The value in PC1*, and is the mean and standard deviation of the non-shadow area detected by SId in PC1*, and are the mean and standard deviation of the shaded area detected by SIs in PC1*.

8. The shadow detection and correction method for infrared imaging spectrometer according to claim 1, characterized in that: The calculation formula of the coefficient in step S53 is as follows: ， In the formula, For the The coefficients of the bands, and Respectively The average brightness value of the shadow area and non-shadow area in the SWIR field of view of each band, and are the proportions of shadow and non-shadow in the SWIR field of view, respectively.

Images