JP5327419B2 - Hyperspectral image analysis system, method and program thereof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To extract a target object automatically by spectrum collation, while suppressing an influence of absorption by atmospheric air, in a system for hyper-spectrum image analysis. <P>SOLUTION: This system is equipped with a reference reflecting plate. Spectrum data of the reference reflecting plate from which the influence of atmospheric air absorption is removed are held. Spectrum data of the reference reflecting plate are acquired. An atmospheric air correction coefficient is calculated based on the acquired spectrum data and the held spectrum data. Spectrum data of a target image are acquired. The atmospheric air correction is performed to the spectrum data of the target image by using the atmospheric air correction coefficient. The reference spectrum data of the target image are held. The first collation is performed between the reference spectrum data and the spectrum data subjected to the atmospheric air correction relative to the whole wavelength domain. The second collation is performed relative to only a specific wavelength domain. The target object is selected based on the first collation result and the second collation result. <P>COPYRIGHT: (C)2009,JPO&amp;INPIT

Description

  The present invention relates to automatic atmospheric absorption correction and automatic target extraction in a hyperspectral image analysis system.

  In order to improve the target extraction capability of remote sensing technology, efforts have been made to increase the resolution, but there is a limit to target extraction only by shape recognition.

For example, it is difficult to detect what is camouflaged. Therefore, it is expected that the target extraction capability will be remarkably improved by the hyperspectral sensing technology using the optical characteristics of the substance by the hyperspectral imaging device (see, for example, Patent Document 1).
JP 2006-012036 A

  However, when using the hyperspectral sensing technology, there were the following two problems.

  The first problem is that the spectrum obtained by the imaging condition differs even for the same object. This is because the visible / near-infrared hyperspectral image is affected by the absorption of sunlight by the atmosphere, and its absorption characteristics depend on the state of the atmosphere and are not constant.

  The second problem is that it is difficult to identify an object having a similar color with a general spectral matching function. This is because similar color objects have similar spectral shapes, so that there is no significant difference in the result of matching by a general matching function.

  Therefore, the present invention provides a hyperspectral image analysis system, a hyperspectral image analysis method, and a hyperspectral image analysis program that can automatically extract a target object by spectral matching while suppressing the influence of sunlight absorption by the atmosphere. Objective.

  According to the present invention, in the analysis system that includes the imaging device including the reference reflector and the image analysis device as the first system and performs the hyperspectral image analysis, the imaging device acquires the spectrum data of the reference reflector. A reflection plate imaging means; and a target image imaging means for acquiring spectral data of the target image, wherein the image analysis apparatus holds the spectral data of the reference reflection plate excluding the influence of atmospheric absorption; and the reference reflection plate An atmospheric correction coefficient calculating means for calculating an atmospheric correction coefficient based on the spectral data acquired in the spectral data imaging means and the spectral data held in the holding means; and a spectrum of the target image using the atmospheric correction coefficient. Atmospheric correction means for performing atmospheric correction on the data, and atmospheric correction by the atmospheric correction means Analysis means for performing a hyperspectral image analysis using spectral data, the analysis system comprising: a provided.

Further, as a second system, in an analysis system that includes an imaging device and an image analysis device including a reference reflector and performs hyperspectral image analysis, the imaging device acquires the spectral data of the reference reflector. And a target image capturing means for acquiring spectral data of the target image, wherein the image analysis apparatus holds the spectral data of the reference reflector excluding the influence of atmospheric absorption, and the reference reflector spectral data imaging An atmospheric correction coefficient calculating means for calculating an atmospheric correction coefficient based on the spectral data acquired in the means and the spectral data held in the holding means; and for the spectral data of the target image using the atmospheric correction coefficient. Atmospheric correction means for correcting the atmosphere and a reference for holding the reference spectrum data of the target object Spectrum data holding means; first reference means for checking the reference spectrum data and spectrum data corrected by the atmosphere correction means for all wavelength regions; and correction of the atmosphere by the reference spectrum data and the atmosphere correction means. A second collation unit that collates the spectrum data with respect to only a specific wavelength region, and a selection unit that selects a target object based on a collation result in the first collation unit and a collation result in the second collation unit. And an analysis system comprising: means.

  Further, as a first method, in the analysis method for performing the hyperspectral image analysis in the system including the imaging device and the image analysis device including the reference reflection plate, the reference reflection plate from which the image analysis device excludes the influence of atmospheric absorption. A holding step for holding the spectrum data of the reference reflector, a reference reflector imaging step for the imaging device to acquire the spectrum data of the reference reflector, the spectral data acquired by the image analyzer in the imaging step of the reference reflector spectral data, and the holding An atmospheric correction coefficient calculating step for calculating an atmospheric correction coefficient based on the spectrum data held in the step, a target image imaging step for the imaging device to acquire spectral data of the target image, and an image analyzer for the atmospheric correction coefficient Is used to perform atmospheric correction on the spectral data of the target image. And atmospheric correction step that the analysis method by the image analyzing apparatus characterized by comprising a an analysis step of performing hyperspectral image analysis using atmospheric correction spectral data in the atmospheric correction step is provided.

Further, as a second method, in an analysis method for performing hyperspectral image analysis in a system including an imaging device and an image analysis device including a reference reflection plate, the reference reflection plate from which the image analysis device excludes the influence of atmospheric absorption A holding step for holding the spectrum data of the reference reflector, a reference reflector imaging step for the imaging device to acquire the spectrum data of the reference reflector, the spectral data acquired by the image analyzer in the imaging step of the reference reflector spectral data, and the holding An atmospheric correction coefficient calculating step for calculating an atmospheric correction coefficient based on the spectrum data held in the step, a target image imaging step for the imaging device to acquire spectral data of the target image, and an image analyzer for the atmospheric correction coefficient Is used to perform atmospheric correction on the spectral data of the target image. An atmospheric correction step, a reference spectrum data holding step in which the image analysis apparatus holds reference spectrum data of the target object, and an image analysis apparatus that converts the reference spectrum data and the spectrum data corrected in the atmosphere in the atmospheric correction step to all wavelengths. A first collation step for collating the region; a second collation step in which the image analysis device collates the reference spectrum data and the spectrum data corrected in the atmosphere in the atmospheric correction step only for a specific wavelength region; and image analysis There is provided an analysis method, characterized in that the apparatus includes a selection step of selecting a target object based on a result of the matching in the first matching step and a result of the matching in the second matching step.

  Further, as a first program, in a hyperspectral image analysis program incorporated in an image analysis apparatus in a system including an imaging apparatus and an image analysis apparatus provided with a reference reflector, the spectrum of the reference reflector excluding the influence of atmospheric absorption Atmosphere for calculating the atmospheric correction coefficient based on the holding function for holding data, the spectrum data acquired in the reference reflector spectral data imaging function transmitted from the imaging device, and the spectrum data held in the holding function Using the correction coefficient calculation function, the atmospheric correction function that performs atmospheric correction on the spectral data of the target image transmitted from the imaging device using the atmospheric correction coefficient, and the spectral data that has been atmospherically corrected in the atmospheric correction function With an analysis function to perform hyperspectral image analysis. Analysis program for causing realized Yuta is provided.

Further, as a second program, in the analysis program for performing the hyperspectral image analysis incorporated in the image analysis apparatus in the system including the imaging apparatus having the reference reflector and the image analysis apparatus, the reference reflection excluding the influence of the atmospheric absorption. An atmospheric correction coefficient is calculated based on the holding function for holding the spectral data of the plate, the spectral data acquired in the reference reflector spectral data imaging function transmitted from the imaging device, and the spectral data held in the holding function. An atmospheric correction coefficient calculation function for calculating, an atmospheric correction function for performing atmospheric correction on the spectral data of the target image transmitted from the imaging device using the atmospheric correction coefficient, and a reference spectrum for holding the reference spectral data of the target object A data holding function, the reference spectrum data and the The first collation function that collates the spectrum data corrected in the atmosphere in the air correction function for all wavelength regions, and the reference spectrum data and the spectrum data corrected in the atmosphere in the air correction function only in a specific wavelength region And a second collation function for selecting a target object based on a collation result in the first collation function and a collation result in the second collation function. An analysis program is provided.

  According to the present invention, it is possible to automatically extract a target object by spectral matching while suppressing the influence of absorption of sunlight by the atmosphere.

  Next, embodiments of the present invention will be described in detail with reference to the drawings.

  Referring to FIG. 1, a configuration diagram of a hyperspectral image analysis system is shown as a first embodiment of the present invention.

  FIG. 2 is an explanatory diagram of the all-wavelength matching process and the specific wavelength matching process. First, a case will be described in which the spectrum shape [1] and the spectrum shape [2] shown in FIG. Hereinafter, a region sandwiched between wavy lines is particularly referred to as a “specific wavelength region”.

  When the shape of the spectrum of [1] and the shape of the spectrum of [2] are collated only for a specific wavelength region, both wavelengths are greatly different, so that the shape of the spectrum of [1] and the shape of the spectrum of [2] are identified. Can do.

  On the other hand, when the shape of the spectrum of [1] and the shape of the spectrum of [2] are collated for all regions, it is difficult to identify. This is because although the specific wavelength region is different, the wavelengths are approximate in regions other than the specific wavelength region, so that it can be said that the two wavelengths do not change greatly as a whole.

  Next, verification of the spectrum shape [1] and the spectrum shape [3] will be described. When the shape of the spectrum of [1] and the shape of the spectrum of [3] are collated only for a specific wavelength region, both wavelengths are approximated so that the shape of the spectrum of [1] and the shape of the spectrum of [3] are identified. It is difficult.

  On the other hand, when the shape of the spectrum of [1] and the shape of the spectrum of [3] are collated for all regions, they can be identified. This is because although the specific wavelength region is approximated, the wavelengths are different in regions other than the specific wavelength region, so that it can be said that the two wavelengths are greatly different as a whole.

  From the above, it can be seen that by combining the all-wavelength matching process and the specific wavelength matching process, it is possible to accurately check the shape of the spectrum.

  Referring to FIG. 1, the hyperspectral image analysis system of the present invention includes a hyperspectral imaging device 1 and an image analysis device 2.

  The hyperspectral imaging device 1 includes a reference reflector 3, a scanning device 4, and an imaging device 5.

  Further, the image analysis device 2 includes a reference data storage unit 6, an atmospheric correction processing unit 7, an all-wavelength matching processing unit 8, and a specific wavelength matching processing unit 9.

  The presence of the reference reflector 3 and the atmospheric correction processing unit 7 according to the present invention enables automatic atmospheric correction. Therefore, the reference spectrum can be verified in the subsequent spectrum verification process.

  In addition, since the all-wavelength matching processing unit 8 and the specific wavelength matching processing unit 9 perform matching processing using the characteristics of the entire spectrum and the characteristics unique to the target object, the target extraction accuracy is improved.

  Next, the processing method of the first embodiment will be described with reference to FIGS.

  First, as shown by the dotted line in FIG. 1, the scanning device is directed in the direction of the reference reflector to obtain spectral data of the reference reflector.

  As shown in FIG. 3, the atmospheric correction processing unit obtains an atmospheric correction coefficient by the following equation.

α (k) = S (k) / S0 (k)
However,
α (k): Atmospheric correction factor
S (k): Reference reflector spectrum
S0 (k): Reference reflector spectrum stored in the reference data storage unit, excluding the effects of atmospheric absorption
k: Wavelength number.

  Next, as shown by the solid line in FIG. 1, the scanning device is directed in the target direction to acquire the target image, and as shown in FIG. 4, the atmospheric correction processing unit 7 uses the correction coefficient α (k) to The atmospheric correction of the target image is performed by the equation.

Zi (k) = Xi (k) / α (k)
However,
Zi (k): Target spectrum after correction
Xi (k): Target spectrum before correction
α (k): Atmospheric correction factor
i: Pixel number
k: Wavelength number.

  Next, the all-wavelength matching processing unit performs primary extraction of the target object based on the characteristics of the entire spectrum. As an example, FIG. 5 shows a reference spectrum, a target 1 spectrum, a target 2 spectrum, and a target 3 spectrum. Here, the reference spectrum refers to target object spectrum data excluding the influence of atmospheric absorption held in the reference data storage unit 6.

  Target 1 is a spectrum equivalent to the reference spectrum, Target 2 is similar to the reference spectrum in the shape of the whole spectrum, but the spectrum shape of a specific region is different from the reference spectrum, and Target 3 is the shape of the whole spectrum as the reference spectrum. However, the spectral shape of the specific region is similar to the reference spectrum.

  The result of collating with the reference spectrum by SAM (Spectral Angle Mapper) which is a typical collation function for the target 1, the target 2 and the target 3 is as follows. Here, SAM is a method of calculating the matching degree by replacing the spectrum shape with a vector and using the angle of the vector as a scale.

Target 1 collation result 0 degrees Target 2 collation result 1 degree Target 3 collation result 20 degrees (The smaller the angle, the higher the collation degree. The maximum collation degree is 0 degree and the minimum is 90 degrees)
As a result, target 1 and target 2 are extracted as the result of all wavelength comparison.

  Finally, the specific wavelength matching processing unit 9 performs matching at a specific wavelength to extract a target object. FIG. 6 shows the specific wavelength matching process.

  First, the wavelength selection unit 91 acquires a processing wavelength range corresponding to the reference spectrum from the reference data storage unit 6, and selects the wavelength range for the target spectrum. Similarly, the wavelength selector 92 selects a wavelength region with respect to the reference spectrum.

  The minimum luminance calculation unit 93 and the minimum luminance calculation unit 94 obtain the minimum value of the selected spectrum, and perform offset removal by subtracting the minimum value. The matching processing unit 95 performs matching processing on the spectrum after the offset removal using a spectrum matching function, and calculates a matching degree.

  For the target 1 and target 2 in FIG. 5, the result of collating with the reference spectrum at a specific wavelength by the SAM (Spectral Angle Mapper) which is a typical collation function is as follows, and the target 1 is extracted as the specific wavelength collation result. The

Target 1 collation result 0 degrees Target 2 collation result 32 degrees (For reference, the target 3 collation result is 7 degrees.)
In this embodiment, since atmospheric correction is performed using the reference reflector, the influence of atmospheric absorption can be automatically removed, and matching is performed using characteristics of the entire spectrum and characteristics of a specific wavelength range according to the object. Goals can be extracted.

[Other Embodiments of the Invention]
In the above-described embodiment, a plurality of spectrum specific wavelength matching processing units may be stacked.

  The advantage of having multiple spectrum specific wavelength matching processing units stacked in this way is that objects with multiple characteristic wavelength ranges can be extracted, and the extraction accuracy for targets for which characteristic wavelength ranges cannot be determined is improved. It can be done.

  Further, by storing the spectrum data after atmospheric correction in the reference data storage unit, it is possible to accumulate the reference data from the imaging spectrum.

[Effect of the invention]
The first effect is that the atmospheric correction can be automatically performed by using the spectrum of the reference reflector, and the spectral matching process can be performed without the influence of the atmosphere.

  The second effect is that an image analysis apparatus having a highly accurate target extraction function can be provided by performing a matching process using the characteristics of the entire spectrum and the characteristics unique to the target object.

  Note that the hyperspectral image analysis system can be realized by a combination of hardware and software.

  Examples of the utilization of the present invention include hyperspectral image analysis systems used for remote sensing such as disaster investigation, vegetation investigation, search, and rescue.

It is a figure which shows the structure of the hyperspectral image analysis system which is embodiment of this invention. It is explanatory drawing regarding the all wavelength collation process in a hyperspectral image analysis, and a specific wavelength collation process. It is explanatory drawing about the air | atmosphere correction coefficient in embodiment of this invention. It is explanatory drawing of the air | atmosphere correction | amendment in embodiment of this invention. It is a figure showing the example of the reference | standard spectrum and target spectrum in embodiment of this invention. It is a block diagram of the specific wavelength collation process part of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Hyperspectral imaging device 2 Image analysis device 3 Reference reflector 4 Scanning device 5 Imaging device 6 Reference data storage unit 7 Atmospheric correction processing unit 8 All-wavelength matching processing unit 9 Specific wavelength matching processing units 91 and 92 Wavelength selection units 93 and 94 Minimum luminance calculation unit 95 Collation processing unit

Claims (10)

In an analysis system that includes an imaging device and an image analysis device equipped with a reference reflector and performs hyperspectral image analysis,
The imaging device
A reference reflector imaging means for obtaining spectral data of the reference reflector;
A target image capturing means for acquiring spectral data of the target image;
Image analysis device
Holding means for holding the spectral data of the reference reflector excluding the influence of atmospheric absorption;
Atmospheric correction coefficient calculating means for calculating an atmospheric correction coefficient based on the spectral data acquired in the reference reflector spectral data imaging means and the spectral data held in the holding means;
Atmospheric correction means for performing atmospheric correction on spectral data of the target image using the atmospheric correction coefficient;
Analyzing means for performing hyperspectral image analysis using the spectrum data corrected in the atmosphere by the atmospheric correction means;
An analysis system comprising: In an analysis system that includes an imaging device and an image analysis device equipped with a reference reflector and performs hyperspectral image analysis,
The imaging device
A reference reflector imaging means for obtaining spectral data of the reference reflector;
A target image capturing means for acquiring spectral data of the target image;
Image analysis device
Holding means for holding the spectral data of the reference reflector excluding the influence of atmospheric absorption;
Atmospheric correction coefficient calculating means for calculating an atmospheric correction coefficient based on the spectral data acquired in the reference reflector spectral data imaging means and the spectral data held in the holding means;
Atmospheric correction means for performing atmospheric correction on spectral data of the target image using the atmospheric correction coefficient;
Reference spectrum data holding means for holding reference spectrum data of the target object;
First collating means for collating the reference spectral data and the spectral data corrected in the atmosphere by the atmospheric correction means for all wavelength regions;
A second collating unit that collates the reference spectral data with the spectral data that has been atmospherically corrected by the atmospheric correction unit only for a specific wavelength region;
Screening means for selecting a target object based on the result of matching in the first matching means and the result of matching in the second matching means;
An analysis system comprising: The analysis system according to claim 2 ,
Image analysis device
The analysis system further comprising recording means for recording the spectrum data corrected by the atmosphere correction means as the reference spectrum data. The analysis system according to claim 2 or 3,
Image analysis device
The analysis system characterized in that the second collating means is operated a plurality of times with different specific wavelength regions. In an analysis method for performing a hyperspectral image analysis in a system including an imaging device and an image analysis device including a reference reflector,
A holding step in which the image analyzer retains the spectral data of the reference reflector excluding the influence of atmospheric absorption;
A reference reflector imaging step in which the imaging device acquires spectral data of the reference reflector;
An atmospheric correction coefficient calculating step for calculating an atmospheric correction coefficient based on the spectral data acquired by the image analysis device in the reference reflector spectral data imaging step and the spectral data held in the holding step;
A target image imaging step in which the imaging device acquires spectral data of the target image; and
An atmospheric correction step in which the image analysis apparatus performs atmospheric correction on the spectral data of the target image using the atmospheric correction coefficient;
An analysis step in which the image analyzer performs a hyperspectral image analysis using the spectrum data corrected in the atmosphere in the atmosphere correction step;
An analysis method comprising: In an analysis method for performing a hyperspectral image analysis in a system including an imaging device and an image analysis device including a reference reflector,
A holding step in which the image analyzer retains the spectral data of the reference reflector excluding the influence of atmospheric absorption;
A reference reflector imaging step in which the imaging device acquires spectral data of the reference reflector;
An atmospheric correction coefficient calculating step for calculating an atmospheric correction coefficient based on the spectral data acquired by the image analysis device in the reference reflector spectral data imaging step and the spectral data held in the holding step;
A target image imaging step in which the imaging device acquires spectral data of the target image; and
An atmospheric correction step in which the image analysis apparatus performs atmospheric correction on the spectral data of the target image using the atmospheric correction coefficient;
A reference spectrum data holding step in which the image analysis device holds the reference spectrum data of the target object;
A first collation step in which the image analysis apparatus collates the reference spectral data with the spectral data that has been atmospherically corrected in the atmospheric correction step for all wavelength regions;
A second collation step in which the image analysis apparatus collates the reference spectral data with the spectral data that has been atmospherically corrected in the atmospheric correction step only for a specific wavelength region;
A selection step in which the image analysis apparatus selects a target object based on a result of the matching in the first matching step and a result of the matching in the second matching step;
An analysis method comprising: The analysis method according to claim 6 , comprising:
Image analysis device
The analysis method further comprising a recording step of recording the spectrum data corrected in the atmosphere in the atmosphere correction step as the reference spectrum data. The analysis method according to claim 6 or 7 , wherein
Image analysis device
The analysis method characterized in that the second matching step is performed a plurality of times with different specific wavelength regions. In the hyperspectral image analysis program incorporated in the image analysis apparatus in the system including the imaging apparatus and the image analysis apparatus including the reference reflector,
A holding function to hold the spectrum data of the reference reflector excluding the influence of atmospheric absorption,
An atmospheric correction coefficient calculation function for calculating an atmospheric correction coefficient based on the spectral data acquired in the reference reflector spectral data imaging function transmitted from the imaging device and the spectral data held in the holding function;
An atmospheric correction function for performing atmospheric correction on the spectral data of the target image transmitted from the imaging device using the atmospheric correction coefficient;
An analysis function for performing a hyperspectral image analysis using the spectrum data corrected in the atmosphere in the atmosphere correction function;
An analysis program characterized in that a computer is realized. In an analysis program for performing a hyperspectral image analysis incorporated in an image analysis apparatus in a system including an imaging apparatus and an image analysis apparatus including a reference reflector,
A holding function to hold the spectrum data of the reference reflector excluding the influence of atmospheric absorption,
An atmospheric correction coefficient calculation function for calculating an atmospheric correction coefficient based on the spectral data acquired in the reference reflector spectral data imaging function transmitted from the imaging device and the spectral data held in the holding function;
An atmospheric correction function for performing atmospheric correction on the spectral data of the target image transmitted from the imaging device using the atmospheric correction coefficient;
A reference spectrum data holding function for holding the reference spectrum data of the target object;
A first collation function for collating the reference spectrum data and the spectrum data corrected in the atmosphere by the atmospheric correction function for all wavelength regions;
A second collation function for collating the reference spectral data with the spectral data that has been atmospherically corrected in the atmospheric correction function only for a specific wavelength region;
A selection function for selecting a target object based on a result of matching in the first matching function and a result of matching in the second matching function;
An analysis program characterized in that a computer is realized.

Images