A noise removal method for satellite hyperspectral data

Abstract

This invention belongs to the field of hyperspectral remote sensing image processing technology, specifically relating to a noise removal method suitable for satellite hyperspectral data, comprising: Step 1, acquiring domestic hyperspectral satellite image data, performing atmospheric correction on the images, and obtaining ground reflectance image data; Step 2, resampling the image data to extract band images within the 1956-2462nm band range of the Gaofen-5 satellite or the Ziyuan-02D satellite; Step 3, resampling the image data to obtain pixel values ​​of three adjacent bands N1, N2, and N3 within the 1956-2462nm band range of the Gaofen-5 satellite or the Ziyuan-02D satellite; Step 4, obtaining image pixel values ​​H1 and H2 based on the pixel values ​​of bands N1, N2, and N3; Step 5, obtaining image pixel value H based on image pixel value H1 or H2. This invention overcomes the shortcomings of existing techniques, reduces the impact of noise, and improves the accuracy and efficiency of data extraction.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of hyperspectral remote sensing image processing, and particularly relates to a noise removal method suitable for satellite hyperspectral data. BACKGROUND

[0002] The domestic satellite hyperspectral data has many bands, the Gaofen 5 satellite has 330 bands, and the resource satellite 02D has 166 bands. In the process of extracting information by using the hyperspectral satellite data, the influence of noise is easy to be received, and the extraction result may be misjudged, so it is necessary to remove the interference of the noise on data processing.

[0003] The original commonly used noise removal method is mainly single-band gray interpolation, and the mean values on both sides are used to remove the filled noise, and this method has low efficiency and poor practicability when applied to the hyperspectral satellite image,

[0004] Therefore, it is necessary to develop a new method to quickly and efficiently remove the influence of noise. SUMMARY

[0005] The purpose of the application is to provide a noise removal method suitable for satellite hyperspectral data, which can make up for the deficiencies of the original technical method, reduce the influence of noise, and improve the accuracy and efficiency of data extraction.

[0006] The technical scheme for realizing the purpose of the application is as follows:

[0007] A noise removal method suitable for satellite hyperspectral data, the method comprises:

[0008] Step 1, acquiring domestic hyperspectral satellite image data, performing atmospheric correction on the image to obtain image data of ground reflectivity;

[0009] Step 2, resampling the image data to extract band image in the range of 1956-2462nm bands of the Gaofen 5 satellite or the resource 02D satellite;

[0010] Step 3, resampling the image data to obtain the pixel values of three adjacent bands N1, N2 and N3 of the Gaofen 5 satellite or the resource 02D satellite in the range of 1956-2462nm bands;

[0011] Step 4, obtaining image pixel values H1 and H2 according to the pixel values of bands N1, N2 and N3;

[0012] Step 5, obtaining image pixel value H according to image pixel value H1 or H2.

[0013] The bands N1, N2 and N3 in step 3 are any three consecutive bands in the range of 1956-2462nm bands.

[0014] The step 4 is specifically: dividing the pixel value of the wave band N1 by the pixel value of the wave band N3 to obtain the image pixel value H1; and dividing the pixel value of the wave band N1 by the pixel value of the wave band N2 to obtain the image pixel value H2.

[0015] The step 5 is specifically:

[0016] When the image pixel value H1 is greater than 1.3, it is judged that the pixel satisfies the condition one;

[0017] When the image pixel value H1 is less than 0.7, it is judged that the pixel satisfies the condition two;

[0018] When the image pixel value H2 is greater than 1.2, it is judged that the pixel satisfies the condition three;

[0019] When the image pixel value H2 is less than 0.8, it is judged that the pixel satisfies the condition four;

[0020] The result of one of the condition one, the condition two, the condition three and the condition four is selected to obtain the image pixel value H.

[0021] The method for obtaining the image pixel value H in the step 5 is:

[0022] If one of the condition one, the condition two, the condition three and the condition four is satisfied, the pixel is a noise pixel or the pixel value is 1; if none of the four conditions is satisfied, the pixel value is 0 or a normal pixel.

[0023] The noise pixel obtained in the step 5 is generated into a mask and superimposed to the mineral information extraction result graph, so that the noise area is not displayed on the final result graph.

[0024] In the step 5, the value of the normal pixel is changed to 1 and multiplied by the final result to remove the influence of the noise.

[0025] The beneficial technical effects of the present application are:

[0026] 1. The noise removal method suitable for satellite hyperspectral data provided by the present application can quickly remove the noise by using the wave band ratio discrimination for calculation according to the characteristics of the noise, and has the characteristics of high speed and high precision.

[0027] 2. The noise removal method suitable for satellite hyperspectral data provided by the present application has obvious effect on the strip noise and can effectively remove the influence of the strip noise.

[0028] 3、The noise removal method suitable for satellite hyperspectral data provided by the application utilizes three continuous wave bands, can avoid the disadvantages of two wave bands, and reduces errors and omissions. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 The noise pixel and the spectral curve thereof provided by the embodiment of the application;

[0030] Figure 2 The normal pixel and the spectral curve thereof provided by the embodiment of the application. DETAILED DESCRIPTION

[0031] The application will be further described in detail below in combination with the drawings and embodiments.

[0032] The noise removal method suitable for satellite hyperspectral data provided by the application specifically comprises the following steps:

[0033] Step 1, acquiring domestic hyperspectral satellite image data, performing atmospheric correction on the image to obtain image data of ground reflectivity;

[0034] Step 2, resampling the image data to extract wave band image in the wave band range of 1956-2462nm of the Gaofen 5 satellite or the Resource 02D satellite;

[0035] The wave band range of 1956-2462nm is generally used for information extraction of geological alteration minerals, because the characteristic spectrum of most minerals is located in this wave band range. In addition, selecting this wave band range can reduce the calculation range and data amount, and improve the calculation efficiency of the program.

[0036] Step 3, resampling the image data to obtain the pixel value of three adjacent wave bands N1, N2 and N3 of the Gaofen 5 satellite or the Resource 02D satellite in the wave band range of 1956-2462nm;

[0037] There are about 61 wave bands of the Gaofen 5 satellite in the wave band range of 1956-2462nm, and the three adjacent wave bands N1, N2 and N3 are sequentially selected from the 61 wave bands, and according to the calculation results of the three wave bands, it is judged whether the pixel is noise.

[0038] The present application selects about 61 bands in the range of 1956-2462 nm from the 330 bands of the Gao Fen 5 hyperspectral satellite image, wherein N1, N2 and N3 represent any three consecutive bands in the 61 bands, such as the first, second and third bands, the second, third and fourth bands, the third, fourth and fifth bands, and the like in turn until all the 61 bands are calculated.

[0039] Step 4, obtaining image pixel values H1 and H2 according to the pixel values of the bands N1, N2 and N3, dividing the pixel value of the band N1 by the pixel value of the band N3 to obtain the image pixel value H1;

[0040] Dividing the pixel value of the band N1 by the pixel value of the band N2 to obtain the image pixel value H2.

[0041] Step 5, obtaining the image pixel value H when the image pixel value H1 or H2 meets one of the following conditions:

[0042] When the image pixel value H1 is greater than 1.3, it is judged that the pixel meets condition one;

[0043] When the image pixel value H1 is less than 0.7, it is judged that the pixel meets condition two;

[0044] When the image pixel value H2 is greater than 1.2, it is judged that the pixel meets condition three;

[0045] When the image pixel value H2 is less than 0.8, it is judged that the pixel meets condition four;

[0046] Selecting the result meeting one of the conditions one, condition two, condition three and condition four to obtain the image pixel value H.

[0047] If one of the conditions one, condition two, condition three and condition four is met, the pixel is noise or the pixel value is 1, and if none of the four conditions is met, the pixel value is 0 or a normal pixel.

[0048] The pixel value H is the binary result of the Gao Fen 5 satellite image pixel after the four conditions are judged, the pixel with H equal to 1 is noise, and the pixel with H equal to 0 is a normal pixel. The obtained noise pixel can be generated into a mask and superimposed on the mineral information extraction result map, so that the noise area is not displayed on the final result map. Or the value of the normal pixel is changed to 1, and multiplied by the final result, which can also remove the influence of noise.

[0049] The pixel value H equal to 0 is a normal pixel, and the pixel value H equal to 1 is noise of the subsequent processing of the domestic hyperspectral satellite image data, that is, the pixel value H equal to 0 is used for subsequent processing, so that the amount of data processing can be reduced and the influence of interference factors can be reduced.

[0050] Embodiment

[0051] The application has been described in detail above in connection with the drawings and embodiments, but the application is not limited to the above-described embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the spirit of the application. The contents not described in detail in the application can adopt the prior art.

[0052] Figure 1 is the spectral curve of a noise pixel, and the absorption peak characteristic waveband near 2201 nm is obvious, Figure 2 is the spectral curve of a normal pixel.

[0053] If three continuous wavebands such as 2201 nm (N1), 2209 nm (N2), and 2218 nm (N3) are selected, the H2 of the noise pixel is 0.616 (less than 0.8), and the H1 is 0.56 (less than 0.7); while the H2 of the normal pixel is 1.03, and the H1 is 0.94, the noise pixel satisfies the second and fourth conditions, and is determined as a noise pixel H=1; the normal pixel does not satisfy any one of the four conditions, and is determined as a normal pixel H=0.

[0054] If three continuous wavebands such as 2184 nm (N1), 2193 nm (N2), and 2201 nm (N3) are selected, the H2 of the noise pixel is 1.06, and the H1 is 1.93 (greater than 1.3); while the H2 of the normal pixel is 1.07, and the H1 is 1.11; the noise pixel satisfies the first condition, and is determined as a noise pixel H=1; the normal pixel does not satisfy any one of the four conditions, and is determined as a normal pixel H=0.

[0055] If three continuous wavebands such as 2193 nm (N1), 2201 nm (N2), and 2209 nm (N3) are selected, the H2 of the noise pixel is 1.83, and the H1 is 1.12; while the H2 of the normal pixel is 1.03, and the H1 is 1.06, the noise pixel satisfies the third condition (greater than 1.2), and is determined as a noise pixel H=1; the normal pixel does not satisfy any one of the four conditions, and is determined as a normal pixel H=0.

[0056] In summary, through the calculation of three continuous wavebands, whether the pixel is a noise pixel can be determined. Although whether the pixel is a noise pixel cannot be determined in some other wavebands, through the calculation of all wavebands in the waveband range by using the determination procedure in the application, whether the pixel is a noise pixel or a normal pixel can be finally determined.

Claims

1. A noise removal method suitable for satellite hyperspectral data, characterized in that, The method comprises: Step 1, obtaining domestic hyperspectral satellite image data, performing atmospheric correction on the image, and obtaining image data of ground reflectivity; Step 2, resampling the image data, and extracting band image in the 1956-2462nm band range of the GaoFen 5 satellite or the Resource 02D satellite; Step 3, resampling the image data to obtain the pixel values of three adjacent bands N1, N2 and N3 in the 1956-2462nm band range of the GaoFen 5 satellite or the Resource 02D satellite; Step 4, obtaining image pixel values H1 and H2 according to the pixel values of bands N1, N2 and N3; Dividing the pixel value of band N1 by the pixel value of band N3 to obtain the image pixel value H1; dividing the pixel value of band N1 by the pixel value of band N2 to obtain the image pixel value H2; Step 5, obtaining the image pixel value H according to the image pixel value H1 or H2; When the image pixel value H1 is greater than 1.3, it is judged that the pixel satisfies condition one; When the image pixel value H1 is less than 0.7, it is judged that the pixel satisfies condition two; When the image pixel value H2 is greater than 1.2, it is judged that the pixel satisfies condition three; When the image pixel value H2 is less than 0.8, it is judged that the pixel satisfies condition four; If one of the four conditions of condition one, condition two, condition three and condition four is satisfied, the pixel is a noise pixel or the pixel value is 1, and if none of the four conditions is satisfied, the pixel value is 0 or a normal pixel.

2. The method for noise removal suitable for satellite hyperspectral data according to claim 1, wherein, In step 3, bands N1, N2 and N3 are any three consecutive bands in the 1956-2462nm band range.

3. The method for noise removal for satellite hyperspectral data according to claim 1, wherein, In step 5, the obtained noise pixels are generated into a mask and superimposed on the mineral information extraction result map, so that the noise area is not displayed on the final result map.

4. The method for noise removal for satellite hyperspectral data according to claim 1, wherein, In step 5, the value of the normal pixel is changed to 1, and the final result is multiplied to remove the influence of noise.

Images