JP2007172590A - Pixel value correction program, and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a physically highly reliable pixel value correction program, etc., capable of eliminating the effects of path radiance that includes the case of atmospheric condition that varies in the horizontal direction, losing neither the local information nor worsening apparent state, from a satellite image data in a visible band, including near-infrared band, of a region having large undulations obtained through an optical sensor. <P>SOLUTION: Each pixel in the near-infrared satellite image data is sorted into a sort class, based on a predetermined criterion. The pixel value of the satellite image data in the visible band, corresponding to each pixel classified into identical sort classes, is extracted. From the coordinate data of a numeric altitude model, the altitude of each pixel is extracted on a basis of each classified class, and by an analysis step, the degree of atmospheric influence on each pixel is obtained on the basis of the classified class. With respect to the degree of atmospheric influence on each pixel, the degree of regional atmospheric influence is obtained by applying a median filter, and the pixel value of each pixel is corrected, based on each pixel value, altitude, reference altitude and degree of the regional atmospheric influence. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a pixel value correction program and a recording medium for correcting pixel values of satellite image data in a visible band relating to a target area.

  Earth observation satellites are equipped with various sensors, and the collected satellite image data is used for various purposes such as meteorological phenomena, terrestrial vegetation cover, and water resources surveys. For example, Landsat 4 which is one of the earth observation satellites is equipped with a 7-band TM (Thematic Mapper) sensor. Bands 1 to 3 are used to obtain visible information, and band 6 is used to obtain temperature information. Thermal bands, bands 4, 5 and 7 are near-infrared bands.

  The luminance value (pixel value) of the satellite image data is influenced by path radiance (path radiance), which is scattered light scattered by the atmosphere, in addition to the light reflected by the ground surface. The characteristics of this path radiance depend on atmospheric conditions including haze or fog haze or fog in the atmosphere. Satellite image data acquired by an optical sensor including a near-infrared band, such as a visible band to a mid-infrared band, includes fluctuations based on the effects of both the atmosphere and topography. In particular, it is important to remove this influence because the path radiance caused by the spatial variation of the atmospheric condition has a large influence on the visible band satellite image data in a region with large undulations.

  In Non-Patent Document 1, it is assumed that each pixel in the near-infrared band corresponds to a certain pixel in the visible band. When scattering occurs in the visible band, each pixel in the near-infrared band has a plurality of pixels in the visible band. I think that it corresponds to a pixel. Therefore, we proposed a method to replace the pixel value of each pixel in the visible band classified into the same classification class by the land cover classification etc. in the near infrared band as the atmospheric correction method. Yes. Although this atmospheric correction method is apparently good for satellite image data having local haze, there is a problem that local information is lost. For example, in a near-infrared band, a paddy field in a haze region and a paddy field in a clear region belong to the same classification class, but in the visible band, both belong to different classification classes. Here, if the paddy field is unevenly distributed in the area having the haze, the average value of the pixel values in the visible band is a value biased toward the area having the haze. For this reason, if the pixel value of the paddy field in a clear area is replaced with the average value, the paddy field in the clear area will also be soiled. In other words, the atmospheric correction method of Non-Patent Document 1 has a problem that local information that the paddy field is unevenly distributed in an area with haze is lost. Further, since Non-Patent Document 1 does not deal with undulating areas, there is a problem that local information such as shading information due to undulations is lost and the appearance is not good.

  Non-Patent Document 2 proposes a simple method for correcting atmospheric and terrain effects, focusing on the point that the difference in solar incident illuminance due to the inclination of the ground surface becomes an important problem in mountainous regions where undulations are severe. Specifically, the fact that the path radiance can be expressed as a linear function of altitude is shown using a simulation of radiative transfer and actual satellite image data. However, there is a problem that the case where the atmospheric state changes in the horizontal direction is not handled.

  In Non-Patent Document 3, the fourth index (simply the difference between band 1 and band 3) among the indices obtained by linearly transforming the pixel values of each of the above-mentioned bands of Landsat TM called Tasseled Cap. A method using an empirical formula that is highly correlated with the amount of haze is described. However, Non-Patent Document 3 ignores that the spectral reflection characteristics are different for each pixel. Statistically, it has been pointed out that the correlation between converted bands is high, and the method of Non-Patent Document 3 has a problem that it is physically unreliable.

  In Non-Patent Document 4, first, infrared bands are classified, the optical thickness of aerosol (haze) is determined for each classification class, and then a value obtained by spatially averaging the optical thicknesses is determined. . Specifically, spatial smoothing of the optical thickness is performed using a normal average value in a relatively narrow range (about 150 m). Using this average value, the reflectance of the ground surface is corrected from the pixel value. However, Non-Patent Document 4 has a problem that it does not handle undulating areas. Furthermore, since physical quantities such as optical thickness and reflectance are used, it is necessary to calculate in advance atmospheric scattering in a clear area. Therefore, there is a problem that it is necessary to precisely define a clear area, which is complicated and impractical. In addition, since the value used for the correction is a normal average value in a relatively narrow range, there is a problem that a necessary change cannot be captured when smoothing is performed in a large range.

M.J.Carlooto: “Reducing the effects of space-varying, wavelength-dependent scattering in multispectralimagery”, Int.J.Remote Sensing, 1999, Vol.20, No.17, pp.3333-3344. Yoshikazu Iikura, Ryuzo Yokoyama, “Correction of Atmospheric and Topographic Effects of Landsat TM Image”, Journal of the Remote Sensing Society of Japan, 1999, Vol. 19, No. 1, p. 2-16. (Y.Iikura and R. Yokoyama: “Correction of Atmospheric and Topographic Effects on Landsat TM Imagery”, Journal of Remote Sensing Society of Japan, 1999, Vol. 19, No. 1, pp. 2-16.) J. Lavreau: “De-hazingLandsat Thematic Mapper images”, Photogrammetric Engineering & RemoteSensing, 1991, Vol.57, No.10, pp.1297-1302. S. Liang, H. Fang, and M. Chen: “Atmospheric Correction of Landsat ETM + Land Surface Imagery- Part I: Methods”, IEEE Trans. On Geoscience and Remote Sensing, 2001, Vol.39, No.11, pp. 2490-2498.

  As mentioned above, satellite image data acquired by optical sensors including near-infrared bands, especially visible band satellite image data in areas with large undulations, is greatly affected by the path radiance caused by spatial fluctuations in atmospheric conditions. Therefore, it is important to remove this effect. However, the atmospheric correction method described in Non-Patent Document 1 has a problem that the satellite image data having local haze looks better, but local information is lost. Further, since Non-Patent Document 1 does not deal with undulating areas, there is a problem that local information such as shading information due to undulations is lost and the appearance is not good. The method for correcting the atmospheric / terrain effect described in Non-Patent Document 2 takes into account undulating areas, but there is a problem that it is not handled when the atmospheric state changes in the horizontal direction. The method of Non-Patent Document 3 has a problem that the reliability is physically low. The method described in Non-Patent Document 4 has a problem that it does not handle undulating areas. Furthermore, since physical quantities such as optical thickness and reflectance are used, it is necessary to calculate atmospheric scattering in a clear area in advance. Therefore, there is a problem that it is necessary to precisely define a clear area, which is complicated and impractical. In addition, since the value used for the correction is a normal average value in a relatively narrow range, there is a problem that a necessary change cannot be captured when smoothing is performed in a large range.

  Therefore, an object of the present invention has been made to solve the above problem, and from satellite image data acquired by an optical sensor including a near-infrared band, particularly visible band satellite image data in a region with a large undulation, A physically reliable pixel value correction program that can eliminate the effects of path radiance, including the case where the atmospheric conditions change in the horizontal direction without losing local information and without deteriorating the appearance. It is to provide. Furthermore, an object of the present invention is to provide a pixel value correction program or the like that is simple and practical and that can capture a required change even when smoothing is performed in a large range.

  The pixel value correction program of the present invention is a pixel value correction program for correcting the pixel value of satellite image data in the visible band relating to the target area, and the satellite image data in the near infrared band relating to the target area is stored in the computer. A classification step of classifying each pixel of the satellite image data of the near-infrared band into a classification class based on a predetermined standard; inputting satellite image data of the visible band relating to the target area; and the same classification class in the classification step A visible band data extraction step for extracting a pixel value of satellite image data of a visible band corresponding to each pixel classified into the above, input coordinate data of a numerical elevation model relating to the target area, and an elevation of each pixel for each classification class A digital elevation data extraction step for extracting the visual band data for each of the classification classes. An analysis step for obtaining an atmospheric influence degree for each pixel based on the obtained pixel value, an elevation extracted in the numerical elevation data extraction step and a predetermined reference elevation, and a central to the atmospheric influence degree of each pixel obtained in the analysis step Applying a value filter to obtain a regional atmospheric influence degree for each pixel, a pixel value of each pixel extracted in the visible band data extraction step, an elevation extracted in the numerical elevation data extraction step, and This is a pixel value correction program for executing a correction step for correcting the pixel value of each pixel based on a predetermined reference altitude and the regional atmospheric influence degree of the pixel obtained in the filter step.

  Here, in the pixel value correction program of the present invention, the predetermined standard is a standard for classifying pixels in which the pixel values of the satellite image data in the near-infrared band match or are within a predetermined range into the same classification class. It can be.

  Here, in the pixel value correction program of the present invention, the predetermined reference elevation can be set to a value within a range of about 1000 m to 1500 m, which is higher than the elevation of the pixel of the classification class.

Here, in the pixel value correction program of the present invention, the analysis step corresponds to the elevation Xi and the pixel value Yi of each pixel using a graph in which the horizontal axis is the altitude and the vertical axis is the pixel value for each classification class. using a graphing step of plotting the pixel coordinates (Xi, Yi), a tilt setting step of setting a gradient B _{0 (pixel} value / altitude) in the clear region, the gradient B ₀ set by the slope setting step for all of said pixels are plotted in the graph creation step (Xi, Yi), a temporary pixel value Ypi at the reference altitude _{X 0, Ypi = Yi-B} 0 × (X 0 -Xi), i = 1~ the A provisional pixel value calculation step using the number of pixels of the classification class, and a median value of the provisional pixel value Ypi calculated in the provisional pixel value calculation step is obtained and the median value is used as a reference elevation And pixel value determining step of a pixel value Y ₀ in X _0, using the pixel value Y ₀ determined by the pixel value determining step for all the pixels plotted in the graph creation step (Xi, Yi), the reference The slope Bi with respect to the pixel (X ₀ , Y ₀ ) at the altitude is calculated using Bi = (Yi−Y ₀ ) / (X ₀ −Xi), i = 1 to the number of pixels of the classification class, and the slope Bi. To the atmospheric influence degree of the pixel (Xi, Yi).

  Here, in the pixel value correction program according to the present invention, the filter step extracts the atmospheric influence degree of pixels within about 600 m around each pixel, and smoothes the result of smoothing using the median value of the atmospheric influence degree. It can be obtained as the regional atmospheric influence level of the pixel.

Here, the pixel value correction program according to the present invention, the correction step, for each pixel altitude Xi, the pixel value Yi, reference altitude X ₀ and regional air impact Bi 'corrected pixel value using a Yi ′ can be calculated by Yi ′ = Yi−Bi ′ × (X ₀ −Xi), i = 1 to the total number of pixels.

  The recording medium of the present invention is a computer-readable recording medium that records the pixel value correction program of the present invention.

  According to the pixel value correction program of the present invention, since the atmospheric influence degree is calculated by an expression that takes elevation into consideration for each classification class such as land cover classification, the pixel value that is physically more reliable than the tasseled cap. A correction program or the like can be provided. Since the median filter is used to determine the regional atmospheric influence level from the atmospheric influence level for each pixel, it is possible to naturally grasp changes in discontinuous atmospheric conditions such as haze, and within a large range. Even when smoothing is performed, necessary changes can be captured. Furthermore, it is possible to obtain the regional atmospheric influence level including the case where the atmospheric state changes in the horizontal direction without losing the local information without deteriorating the appearance. By considering the atmospheric influence level or the regional atmospheric influence degree at the pixel as the inclination with respect to the pixel at the reference altitude, the correction amount can be easily obtained as a linear expression of the altitude.

  As a result, from the satellite image data acquired by the optical sensor including the near-infrared band, especially the visible band satellite image data in the region with a large undulation, the atmospheric state can be maintained without losing the local information without losing the appearance. There is an effect that it is possible to provide a physically reliable pixel value correction program or the like that can remove the influence of path radiance including the case of changing in the horizontal direction. Furthermore, since a physical quantity such as optical thickness or reflectance is not used, a simple and practical pixel value correction program can be provided.

  First, the theory underlying the pixel value correction program for correcting pixel values of satellite image data in the visible band relating to the target area of the present invention will be described, and then each embodiment will be described in detail with reference to the drawings. To do.

Explanation of theory.
1. The point that the path radiance can be expressed as a linear function of altitude described in Non-Patent Document 2 of the background art is expanded, and the following formulas 1 and 2 that the slope of the linear function varies depending on the location are expressed in the present invention. A basic model of the pixel value correction program was used. As a result, the amendment of the atmosphere was improved. Elevation (m) is X-axis, pixel value (Digital Number: DN) is Y-axis, pixel value before correction at elevation Xi is Yi, correction amount is Di, slope is Bi, parameter is A, corrected pixel value Is Yi ',

          Yi '= Yi-Di (1)

It can be expressed as. The correction amount Di is

          Di = A−Bi × Xi (2)

It is. Here, if the correction amount at the altitude where the influence of the atmosphere can be ignored (reference altitude X ₀ ) is 0, A can be set as follows.

0 = A−Bi × X ₀
∴ A = Bi x X ₀

  Substituting the parameter A into the equation 1, the equation 1 becomes the following equation 3, and the parameter need only be Bi.

_{Di = Bi × X 0 -Bi ×} Xi = Bi × (X 0 -Xi) (3)

Slope Bi of the target pixel (Xi, Yi) is, knowing the reference altitude X pixel value at ₀ Y _{0 (pixel} value when there is no influence of the atmosphere), it can be obtained from Equation 4 below.

Bi = (Yi−Y ₀ ) / (X ₀ −Xi) (4)

2. In order to obtain Bi of the basic model for each pixel (Xi, Yi), calculation of Equation 4 is performed. Here, the pixel value Y ₀ in Equation 4 is necessary for each classification class classified by land cover classification or the like. The pixel value correction program of the present invention, in order to obtain the pixel values Y _0, the pixel value of each pixel of the visible bands that are classified into the same classification class in the near-infrared band, as described in Non-Patent Document 1 of the related art Is used. In Non-Patent Document 1, each pixel value is replaced with the average value of each pixel in the visible band, but the difference from the average value can also be considered as the influence level of the atmosphere. If more than half of the image is a clear region, it is better to replace the pixel values with their median values rather than the average values. Utilizing gradient B ₀ in a clear area in order to obtain the pixel value Y _0. B ₀ uses a value (7-8 DN / km in band 1) set by 6S (an abbreviation for Second Simulation of the Satellite Signal in the Solar Spectrum, a program developed at Lille University of Science and Technology, France). First, for all the pixels included in the same classification class (Xi, Yi), a temporary pixel value Ypi at the reference altitude X ₀ is calculated using Equation 5 below.

Ypi = Yi−B ₀ × (X ₀ −Xi) (5)

Since Ypi of pixels included in the clear region is constant regardless of the elevation, which can be regarded as a pixel value Y ₀ when there is no influence of the atmosphere. When pixels included in clear areas in the classification class is large, median Ypi becomes Y _0. Since Ypi of the pixels included in the non-clear area is large, when there are many such pixels, the center calculated for the clear area is larger than the median value of Ypi calculated for all the pixels (Xi, Yi). Use the value. Specifically, assuming that the percentage of the image occupied by the clear area is W (%), the Ypi of W / 2 (%) from the Ypi frequency distribution calculated for all the pixels (Xi, Yi). Use the median calculated using.

3. The inclination Bi can be obtained for each pixel by 1 and 2 described above. The inclination Bi can be adopted as the influence degree of the atmosphere. However, the inclination Bi includes various error factors. From the satellite image data, it can be seen that there is a local feature in the degree of influence of the atmosphere, which is clear in some places but thin fog or haze in other places. The influence of the atmosphere is thought to change gradually with a certain extent. Therefore, it is considered that the error can be reduced by using not only the pixel unit but also the data in the vicinity thereof as the atmospheric influence degree. Therefore, the spatial atmospheric influence degree Bi ′ is obtained by applying a spatial median filter that extracts the gradients Bi of the pixels around the target pixel and uses the median of these. The reason why the median value is used instead of the average value is that when the clear area and the non-clear area are clearly separated, the boundary between both areas is blurred when the average value is used compared to the median value. The correction amount Di is obtained by the following equation 6 using the regional atmospheric influence degree Bi ′.

Di = Bi ′ × (X ₀ −Xi) (6)

  The corrected pixel value Yi ′ is expressed by the following Expression 7 using Expressions 1 and 6.

Yi ′ = Yi−Bi ′ × (X ₀ −Xi) (7)

  FIG. 1 is a flowchart showing the flow of a pixel value correction program according to the first embodiment of the present invention. In the following, description of each processing block in FIG. 1 will be described using the example in FIGS. 2 to 4, the detailed flowchart shown in FIG. 5, the graph shown in FIG. 6, and the examples in FIGS.

  As shown in FIG. 1, first, satellite image data of a near-infrared band related to a desired target area (for example, Okawa area, Iwate Prefecture) is input to a computer, and each pixel of the satellite image data of the near-infrared band is input A classification step (step S10) for classifying into classification classes based on predetermined criteria is executed. Since the input satellite image data in the near-infrared band is general, a description thereof will be omitted. As the predetermined reference, for example, a reference for classifying pixels in which pixel values (DN) of satellite image data in the near-infrared band are identical or within a predetermined range into the same classification class can be used. Of course, other criteria may be used.

  FIG. 2 shows a classification data image obtained as a result of executing the classification step (step S10). The classification data image shown in FIG. 2 is a color composite image.

  Next, as shown in FIG. 1, the satellite image data of the visible band relating to the same target area as the target area is input to the computer, and the pixels corresponding to the pixels classified into the same classification class in the classification step (step S10) are supported. The visible band data extracting step (step S20) for extracting the pixel value of the visible band satellite image data is executed.

  FIG. 3 shows an example of satellite image data of the visible band (band 1) input in the visible band classification step (step S20). As shown by the upper right portion Haze in FIG. 3, the portion is brightened due to the influence of haze.

  Subsequently, as shown in FIG. 1, the coordinate data of the numerical elevation model related to the same target area as the target area is input to the computer, and the altitude of each pixel is classified for each classification class classified in the classification step (step S10). The digital elevation data extracting step (step S30) for extracting the.

  FIG. 4 shows an example of the coordinate data image of the digital elevation model input in the pixel value acquisition step (step S30). The coordinate data image is created by converting the Geographical Survey Institute's numerical data 50 m mesh (elevation) into the UTM coordinate system (Universal Transverse Mercator system).

  Next, as shown in FIG. 1, the pixel values extracted in the visible band data extraction step (step S20) and the digital elevation data extraction step (step S30) are extracted by the computer for each classification class. Based on the altitude and a predetermined reference altitude, an analysis step (step S40) for determining the atmospheric influence degree for each pixel is executed.

  FIG. 5 is a flowchart showing the detailed flow of the analysis step (step S40), and FIG. 6 is a graph for explaining the details of the analysis step (step S40). The following processing is executed for each classification class. Using the graph with the horizontal axis shown in FIG. 6 as the elevation (m) and the vertical axis as the pixel value (DN), the pixels are plotted at the coordinates (Xi, Yi) corresponding to the elevation Xi and the pixel value Yi of each pixel. (Graph creation step. Step S41).

Next, the inclination B ₀ (pixel value / elevation) in the clear area is set (inclination setting step, step S42). As described above, B ₀ may be a value set by radiative transfer simulation such as 6S (7 to 8 DN / km in band 1).

Using the inclination B ₀ set in the inclination setting step (step S42), the provisional pixel value Ypi at the reference altitude X ₀ is calculated for all the pixels (Xi, Yi) plotted in the graph creation step (step S41). Equation 5 above

Ypi = Yi−B ₀ × (X ₀ −Xi), i = 1 to the number of pixels of the classification class (5)

(Temporary pixel value calculation step, step S43). The predetermined reference elevation X _0, they are preferable that a value within the range of about 1500m from higher 1000m from elevation of the pixel of the classification classes. It is used to 1200m as a predetermined reference elevation _{X 0} in FIG. 6, a predetermined reference elevation _{X 0} is not limited to 1200m.

Temporary pixel value calculation step (step S43) obtains the median of the calculated temporary pixel value Ypi in, the median value as the pixel value Y ₀ at the reference altitude X ₀ (the pixel value determination step. Step S44). In Figure 6, the pixel value _{Y 0} is approximately 75 (DN).

Using the pixel value _{Y 0} determined by the pixel value determining step (step S44), all the pixels (Xi, Yi) plotted graphically creation step (step S41) for the pixels in the reference altitude _{X 0 (X} 0, The slope Bi with respect to (Y ₀ )

Bi = (Yi−Y ₀ ) / (X ₀ −Xi), i = 1 to the number of pixels of the classification class (4)

The slope Bi is used as the atmospheric influence degree of the pixel (Xi, Yi) (step S45). In FIG. 6, the slope Bi of the pixel (450, 90) with respect to the pixel (1200, 75) at the reference altitude of 1200 m is indicated by a straight line i. This slope Bi, that is, a value that changes by 15 DN (= 90 DN-75 DN) per 750 m (= 1200 m−450 m) is the atmospheric influence degree in the pixels (450, 90). FIG. 7 is an image showing the atmospheric influence degree Bi for each pixel obtained as described above.

  Returning to FIG. 1, the median filter is applied to the atmospheric influence degree Bi of each pixel (Xi, Yi) obtained in the analysis step (step S40), and the region for each pixel (Xi, Yi) is applied to the computer. The filter step (step S50) for obtaining the atmospheric influence degree Bi ′ is executed. FIG. 8 is an image showing the execution result of the filter step (step S50). In FIG. 8, in the filter step (step S50), the atmospheric influence degree Bi of pixels within about 600 m around each pixel (Xi, Yi) is extracted, and smoothed using the median value of the atmospheric influence degree Bi. Is obtained as the regional atmospheric influence degree Bi ′ of each pixel (Xi, Yi).

Finally, the computer, the pixel value Yi of each pixel extracted in the visible band data extracting step (step 20), and elevation Xi extracted with digital elevation data extraction step (step S30), a predetermined reference elevation X _0, Based on the regional atmospheric influence Bi ′ of the pixel obtained in the filter step (step 50),

Di = Bi ′ × (X ₀ −Xi) (6)

The correction amount Di is obtained by the above equation 7,

Yi ′ = Yi−Bi ′ × (X ₀ −Xi) (7)

Thus, the correction step (step 60) for correcting the pixel value Yi of each pixel to obtain Yi 'is executed. Here, i = 1 to the total number of pixels. FIG. 9 is an image showing a result of executing the above-described correction step (step S60).

As described above, according to the first embodiment of the present invention, first, near infrared band satellite image data relating to a desired target area is input to a computer, and each pixel of the near infrared band satellite image data is set to a predetermined value. A classification step for classifying into classification classes based on the criteria is executed. Next, the satellite image data of the visible band relating to the same target area as the target area is input, and the pixel value of the satellite image data of the visible band corresponding to each pixel classified into the same classification class in the classification step (step S10). A visible band data extraction step (step S20) to be extracted is executed. Subsequently, coordinate data of a digital elevation model relating to the same target area as the target area is input, and a digital elevation data extraction step (step S30) is performed for extracting the elevation Xi of each pixel for each classification class. For each of the classification classes, the atmosphere for each pixel based on the pixel value extracted in the visible band data extraction step (step S30), the elevation extracted in the numerical elevation data extraction step (step S30), and a predetermined reference elevation. An analysis step (step S40) for obtaining the influence degree is executed. A filter that applies a median filter to the atmospheric influence degree Bi of each pixel (Xi, Yi) obtained in the analysis step (step S40) and obtains a regional atmospheric influence degree Bi ′ for each pixel (Xi, Yi). Step (step S50) is executed. A pixel value Yi of each pixel extracted in the visible band data extracting step (step S20), and elevation extracted with digital elevation data extraction step (step S30), a predetermined reference elevation X _0, the filter step (Step 50) Based on the obtained regional atmospheric influence degree Bi ′ of the pixel, the correction amount Di is obtained by the above-described equation 6, and each pixel value Yi is corrected by the above-described equation 7 to obtain Yi ′. Step S60) is executed.

In the pixel value correction program according to the present invention, the atmospheric influence degree Bi is calculated by the equation 4 in which the altitude Xi is added for each classification class such as land cover classification, so that the pixel is physically more reliable than the tasseled cap. A value correction program or the like can be provided. In the pixel value correction program of the present invention, the median filter is used when the regional atmospheric influence degree Bi ′ is obtained from the atmospheric influence degree Bi for each pixel. Can be grasped naturally, and even if smoothing is performed over a large range, necessary changes can be captured. Furthermore, it is possible to obtain the regional atmospheric influence degree Bi ′ including the case where the atmospheric state changes in the horizontal direction without losing the local information without deteriorating the appearance. In the pixel value correction program of the present invention, the atmospheric influence degree Bi or the regional atmospheric influence degree Bi ′ at the pixel (Xi, Yi) is considered as the inclination Bi with respect to the pixel (X ₀ , Y ₀ ) at the reference altitude X ₀ . Thus, the correction amount Di can be easily obtained as a linear expression of the altitude Xi. From the above, from the satellite image data acquired by the optical sensor including the near-infrared band, especially the visible band satellite image data in the region with large undulations, the atmospheric state is maintained without losing the local information without losing the appearance. It is possible to provide a physically reliable pixel value correction program or the like that can eliminate the influence of path radiance including the case of changing in the horizontal direction. Furthermore, since a physical quantity such as optical thickness or reflectance is not used, a simple and practical pixel value correction program or the like can be provided.

  FIG. 10 is a block diagram showing an internal circuit 10 of a computer that executes the computer program (pixel value correction program) of the present invention for realizing the first embodiment. As shown in FIG. 10, the CPU 11, ROM 12, RAM 13, image control unit 14, controller 18, input control unit 20, and external interface (I / F) unit 21 are connected to a bus 22. In FIG. 10, the above-described pixel value correction program of the present invention is recorded on a recording medium (including a removable recording medium) such as ROM 12, HDD 17a or DVD 17n (or CD-ROM). The HDD 17a or the DVD 17n or the like stores the input near-infrared band satellite image data, visible band satellite image data, numerical elevation model coordinate data, and the like. The pixel value correction program of the present invention is loaded into the RAM 13 from the ROM 12 via the bus 22 or from a recording medium such as the HDD 17a or DVD 17n via the controller 18 via the bus 22. The input control unit 20 is connected to an input operation unit 19 such as a mouse or a numeric keypad and performs input control and the like. The VRAM 15 serving as an image memory includes a classification data image, an image indicating the atmospheric influence Bi for each pixel, an image indicating the execution result of the filter step (step S50), an image indicating the result of executing the correction step (step S60), and the like. It has a capacity corresponding to the data capacity of at least one screen of the display unit 16 such as a display for display, and the image control unit 16 has a function of converting data in the VRAM 15 into image data and sending it to the display unit 16. is doing. The external I / F unit 21 has an interface function when inputting near-infrared band satellite image data, visible band satellite image data, coordinate data of a digital elevation model, and the like.

  As described above, the CPU 11 executes the pixel value correction program of the present invention described above, thereby achieving the object of the present invention. The pixel value correction program can be supplied to the computer CPU 11 in the form of a recording medium such as the HDD 17a or the DVD 17n as described above, and the recording medium such as the HDD 17a or the DVD 17n on which the pixel value correction program is recorded similarly applies the present invention. Will be composed. As a recording medium on which the pixel value correction program is recorded, for example, a memory card, a memory stick, or the like can be used in addition to the recording medium described above.

  As an application example of the present invention, the present invention can be applied to correction of pixel values of satellite image data acquired by a Landsat TM sensor.

It is a flowchart which shows the flow of the correction program in Example 1 of this invention. It is a figure which shows the classification | category data image of the result of having performed a classification | category step (step S10). It is a figure which shows an example of the satellite image data of the visible band (band 1) input at a visible band classification step (step S20). It is a figure which shows an example of the coordinate data image of the numerical elevation model input at an analysis step (step S40). It is a flowchart which shows the detailed flow of an analysis step (step S40). It is a graph for demonstrating the detail of an analysis step (step S40). It is an image which shows the atmospheric influence degree Bi for every pixel calculated | required at the analysis step (step S40). It is an image which shows the execution result of a filter step (step S50). It is an image which shows the result of having performed the correction step (Step S60). It is a block diagram which shows the internal circuit 10 of the computer which performs the pixel value correction program of this invention for implement | achieving the Example of this invention.

Explanation of symbols

Bi, B ₀ inclination, area affected by Haze haze, pixel value at X ₀ reference elevation, Y ₀ reference elevation, 10 internal circuit, 11 CPU, 12 ROM, 13 RAM, 14 image control unit, 15 VRAM, 16 display Section, 17a HDD, 17n DVD, 18 controller, 19 input operation section, 20 input control section, 21 external I / F section, 22 bus.

Claims (7)

A pixel value correction program for correcting pixel values of satellite image data in a visible band relating to a target area,
A classification step of inputting satellite image data of the near-infrared band related to the target area, and classifying each pixel of the satellite image data of the near-infrared band into a classification class based on a predetermined criterion;
A visible band data extracting step of inputting satellite image data of a visible band related to the target area, and extracting a pixel value of the satellite image data of the visible band corresponding to each pixel classified into the same classification class in the classification step;
Numerical elevation data extraction step of inputting coordinate data of a digital elevation model related to the target area and extracting the elevation of each pixel for each classification class,
For each classification class, the pixel value extracted in the visible band data extraction step, the analysis step for determining the atmospheric influence level for each pixel based on the elevation extracted in the numerical elevation data extraction step and a predetermined reference elevation,
Applying a median filter to the atmospheric influence degree of each pixel obtained in the analysis step, and a filter step for obtaining a regional atmospheric influence degree for each pixel;
The pixel value of each pixel extracted in the visible band data extraction step, the elevation extracted in the numerical elevation data extraction step, the predetermined reference elevation, and the regional atmospheric influence degree of the pixel obtained in the filter step And a pixel value correction program for executing a correction step for correcting the pixel value of each pixel.   2. The pixel value correction program according to claim 1, wherein the predetermined criterion is a criterion for classifying pixels in which pixel values of near-infrared band satellite image data match or are within a predetermined range into the same classification class. A pixel value correction program characterized by being.   3. The pixel value correction program according to claim 1, wherein the predetermined reference altitude is higher than an altitude of the pixels of the classification class and is in a range of about 1000 m to 1500 m. The pixel value correction program according to any one of claims 1 to 3, wherein the analysis step is performed for each classification class.
A graph creation step of plotting pixels at coordinates (Xi, Yi) corresponding to the elevation Xi and pixel value Yi of each pixel using a graph in which the horizontal axis is the elevation and the vertical axis is the pixel value;
An inclination setting step for setting an inclination B ₀ (pixel value / elevation) in a clear area;
Using the slope B ₀ set in the slope setting step, for all the pixels (Xi, Yi) plotted in the graph creating step, the temporary pixel value Ypi at the reference altitude X ₀ is set as Ypi = Yi−B _0. A provisional pixel value calculation step of calculating using x (X ₀ −Xi), i = 1 to the number of pixels of the classification class;
And pixel value determining step of a pixel value Y ₀ at the reference altitude X ₀ the median calculated the median value of pixel values Ypi provisional calculated by the pixel value calculation step of the temporary,
Using the pixel value Y ₀ determined by the pixel value determining step, all the pixels (Xi, Yi) plotted in the graph creation step for the slope Bi with respect to the pixel (X _{0, Y 0)} in the reference altitude, Calculating Bi = (Yi−Y ₀ ) / (X ₀ −Xi), i = 1 to the number of pixels of the classification class, and setting the slope Bi as the atmospheric influence level of the pixel (Xi, Yi); A pixel value correction program comprising:   5. The pixel value correction program according to claim 1, wherein the filter step extracts an atmospheric influence degree of pixels within about 600 m around each pixel, and smoothes using the median value of the atmospheric influence degree. The pixel value correction program characterized by calculating | requiring the result obtained as a regional atmospheric influence degree of each pixel. The pixel value correction program according to any one of claims 1 to 5, wherein the correction step, the corrected using elevation Xi, the pixel value Yi, reference altitude X ₀ and regional atmospheric influence Bi 'of each pixel Pixel value Yi ′ is calculated by Yi ′ = Yi−Bi ′ × (X ₀ −Xi), i = 1 to the total number of pixels.   A computer-readable recording medium on which the pixel value correction program according to claim 1 is recorded.