JP2008152709A - Geographic image processor and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processing system which automatically selects a band suitable for extracting a region to be extracted from a hyper-spectrum image photographed from a satellite. <P>SOLUTION: This geographic image processor for processing geographic images having a plurality of wavelength bands comprises: a data-for-learning storing means for storing spectral data corresponding to the plurality of wavelength bands in each classification class; a band selection processing means for selecting a characteristic wavelength band in one classification class to be processed comparing with the other classification class based on the spectral data in each classification class; and a classification band data storing means for storing information on the wavelength band of the classification class to be processed selected by the band selection processing means as classification band data used in performing the covering classification processing. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention uses a multi-band spectrum image (hereinafter referred to as a hyperspectral image) obtained from satellite images, aerial photographs, etc., to classify photographed areas into classes such as forests, water areas, and upland fields ( The present invention relates to a geographic image processing apparatus that performs a covering classification process.

  In recent years, the use of satellite images and aerial photograph images (hereinafter referred to as geographic images) has become widespread, and is used for applications such as covering classification and exploration of vast areas in map creation / update efficiently. Yes. The use of geographic images can reduce the cost of on-site surveys and efficiently conduct a wide range of surveys.

  A method using a hyperspectral image of a geographic image has been proposed when performing coverage classification of a region (Non-Patent Document 1). The hyperspectral image is an image of the spectrum of the object detected by the hyperspectral sensor for each band. The hyperspectral sensor can collect data in many bands from visible light to near infrared. Since the spectral characteristics of the hyperspectral image vary depending on the object (see FIG. 3), the cover classification can be performed by using this feature. Conventionally, the wavelength resolution has been about several tens of nanometers, but recently, with the progress of technology, sensors whose wavelength resolution has been improved from several nanometers to about 10 nanometers have been put into practical use.

MicroImages, Inc. “Introduction to TNT Hyperspectral Image Classification”, 2001, [Search June 10, 2006], Internet, <URL: http://www.microimages.com>

  In recent years, it has become possible to photograph geographic images under a wide range and high resolution conditions, and thus the image size has become larger than conventional geographic images. Furthermore, since a hyperspectral image generally has information on the number of bands of several hundreds of geographic images, the data is larger by the number of bands than a single-band geographic image. In order to capture such a large-size hyperspectral image and cover classification processing, a large-capacity storage device is required to store the captured geographical image (hyperspectral image), and its cost is a problem. It has become. For example, when a 10 km square is photographed at a ground resolution of 1 m, the image capacity of 100 bands is 10 gigabytes, which is a very large capacity. Further, for example, when land cover classification processing is performed using such a large amount of data, the processing time also becomes enormous. In addition, there is a theory that the amount of essential information included in the hyperspectral image data is not so large, and there is a problem that noise increases as the number of dimensions increases.

  Therefore, realization of a method for efficiently selecting a band necessary for covering classification from a band of a hyperspectral image is desired.

  The present invention has been made in view of such circumstances, and provides a process for efficiently selecting a band of a hyperspectral image necessary for cover classification.

  The present invention also provides a system for efficiently performing a cover classification process on an input map image.

  As a result of diligent research, the present inventor previously obtained a combination of bands necessary and sufficient for recognizing each class, recorded only the necessary bands at the time of photographing a geographic image, and used only those bands to determine the target class. The inventors have conceived that the above-mentioned problems can be solved by performing the covering classification.

  In order to solve the above problems, the present invention is a system for selecting a band that can classify a target classification class from hyperspectral information included in a geographic image, and includes a band selection processing unit. The system characterized by this is provided.

  That is, the geographic image device according to the present invention is a geographic image processing device that processes a geographic image having a plurality of wavelength bands, and stores learning data for storing spectrum data corresponding to the plurality of wavelength bands for each classification class. Selected by the band selection processing means, the band selection processing means for selecting a characteristic wavelength band for one processing target classification class compared to other classification classes, based on the spectrum data for each classification class Classification band data storage means for storing information on the wavelength band of the classification class to be processed as classification band data used when performing the cover classification process.

  More specifically, the band selection processing means calculates the absolute difference between the spectrum data of the processing target classification class and the other classification class for each wavelength region, acquires the minimum difference absolute value in each wavelength region, Among the minimum difference absolute values in the wavelength region, information on the wavelength region of a combination in which the sum exceeds a predetermined threshold value is selected. More specifically, the band selection processing means rearranges the minimum difference absolute value in each wavelength region in order of increasing values, calculates a cumulative addition value from the large minimum difference absolute value, and the cumulative addition value is the predetermined value. When the threshold value is exceeded, information on the wavelength region constituting the cumulative addition value may be selected.

  A geographic image processing apparatus according to the present invention is a geographic image processing apparatus that covers and classifies a geographic image having spectral data corresponding to a plurality of wavelength bands, and uses a wavelength used when covering classification processing is performed for a plurality of classification classes. Classification band data storage means for storing classification band data indicating a region, means for obtaining data of a designated classification class, and obtaining classification band data for a designated classification class, and included in the obtained classification band data Processing image acquisition means for acquiring geographic images consisting of spectral data corresponding to the wavelength region to be covered, and for each designated classification class, covering classification processing is executed using the spectral data of the geographic images obtained by the processing image acquisition means And a cover classification processing means. The information on the wavelength region forming the classification band data is selected by the band selection processing means according to the above-described procedure, and is stored in the classification band data storage means as the classification band data.

  In addition, according to this invention, the program for functioning a computer as the above-mentioned geographic image processing apparatus is also provided.

  Further features of the present invention will become apparent from the best mode for carrying out the present invention and the accompanying drawings.

  According to the present invention, it is possible to efficiently select a band of a hyperspectral image necessary for cover classification, and to efficiently perform a cover classification process on an input map image.

  The best mode for carrying out the geographic image processing system of the present invention will be described below in detail with reference to the accompanying drawings. 1 to 7 are diagrams illustrating embodiments of the present invention. In these drawings, the same reference numerals denote the same components, and the basic configuration and operation are the same. To do. Note that this embodiment is merely an example for realizing the present invention, and does not limit the present invention. A person skilled in the art can make various changes and improvements to the configuration and function of the invention according to the present embodiment without departing from the gist of the present invention.

<Configuration of map image processing system>
FIG. 1 is a block diagram schematically showing the configuration of a geographic image processing system of the present invention. In FIG. 1, this system includes a storage device 20 including a processing device 10 including a personal computer, a workstation and the like, and an auxiliary storage device such as a RAM (Random Access Memory) and a magnetic disk storage device used as a main memory. And an input / output device 30.

  The input / output device 30 includes an input device 31 including a pointing device such as a keyboard and a mouse, a display device 32 such as a CRT display device, and a printer 33. The input device 31 is used for input of parameters by the user, activation of commands, and the like, and generation of optimal spectrum combination data for each classification class by this system. The display device 32 and the printer 33 are used for presenting geographic data and spectrum combination data according to the present system to the user. Note that the display device may include only one of the display device 32 and the printer 33.

  The processing device 10 includes a geographic image processing program 40. The geographic image processing program 40, as a program module, a band selection processing unit 100 for obtaining an optimum combination of bands for covering classification of each classification class, and a cover for performing the cover classification of the classification class designated for the geographic image. A classification processing unit 200.

  The storage device 20 stores teacher data 21, classification band data 22, and geographic image data 23. Among these, the teacher data 21 is spectrum data of each classification class available on the Internet or the like. FIG. 3 is a diagram illustrating teacher data, and shows a function of wavelength and reflectance in classification classes such as water area and soil.

  On the other hand, the classification band data 22 is data generated based on the teacher data 21 in this system. The classification band data 22 stores optimal band combination data for each class to be classified based on the teacher data 21 (see FIG. 5). The classification class is a classification item for covering classification such as “water area”, “forest”, and “road”. The classification class can be freely set according to the user's request.

  The geographic image data 23 is hyperspectral image data. However, only the band data necessary for classifying the cover classification class designated by the user is stored. That is, only the classification band data corresponding to the classification class shown in FIG. 5 to be described later is stored out of all hand data constituting the geographic image data. For example, when a user classifies three forests, water bodies, and soils with respect to a certain aerial image data, not all the spectral band data constituting the aerial image data but the bands 2 and N- 1. Only the spectrum data of the bands 3, N-1 and N corresponding to the water area, and the bands 1 and 3 corresponding to the soil are stored in the storage device 20 as geographic images.

<Details of geographic image processing>
Hereinafter, details of the processing of the band selection processing unit 100 and the cover classification processing unit 200 in the image processing system of the present invention will be described. In this system, when the geographic image processing program 40 is started, the band selection processing unit 100 or the cover classification processing unit 200 is executed. Either process is selected by the user. Details of the processing will be described below.

  First, the band selection processing unit 100 will be described. FIG. 2 is a flowchart showing a processing flow by the classification band selection processing unit 100 of the geographic image processing program 40. In FIG. 2, the classification band selection processing unit 100 first reads the teacher data 21 from the storage device 20 (S101). Next, one unprocessed classification class is selected (S102). Next, the degree of independence in each band of the classification class is calculated. The degree of independence is an index indicating how different the spectrum of a certain classification class in each band is from the spectrum of another classification class. In other words, if the spectrum in a band of the class of interest is the same as the spectrum of the band of another classification class, the degree of independence is small, and if the spectrum is different, the degree of independence is large. . A method of calculating the degree of independence will be described with reference to FIG. FIG. 4A shows the reflectance at each wavelength of the classification classes A to D. FIG. Here, as an example, the degree of independence in classification class A is obtained.

In each band, the absolute value of the difference in reflectance between the class A of interest and each class is taken. Next, the minimum value is obtained from the obtained difference absolute values. This minimum value is set as the degree of independence in the band of the classification class of interest. FIG. 4B is an explanatory diagram of the band 5, and d _AB , d _AC , and d _AD represent the absolute difference values of the reflectances of the classification classes B, C, and D with respect to the classification class A. _{D AD} is independent of d ₅ at the band 5 so _{d AD} in this example takes the minimum value.

Returning to the flowchart, the independence in each band obtained in step S103 is rearranged in descending order (S104). Thereafter, in steps S105, S106, and S107, band selection processing for discriminating the classification class is performed. First, the values of independence are cumulatively added in the order of bands having the highest independence (S105). For example, in FIG. 4 (a), the first adding this because high d ₆ most degree of independence. Next, it is determined whether or not the cumulative addition value is equal to or greater than a preset threshold value (S106).

If it is determined in step S106 that the cumulative addition value is equal to or greater than the threshold value, the loop is exited. If the threshold value is not satisfied, the process returns to step S105, and the next highest degree of independence after the one used in the previous order is cumulatively added. In step S106, the process is repeated until the cumulative independence becomes equal to or greater than the threshold value. In FIG. 4C, the degree of independence is added in the order of d ₆ , d ₇ , and d _1, and the cumulative degree of independence exceeds the threshold when d ₁ is added. The threshold value used here can be arbitrarily set by the user. If the threshold value is high, the accuracy of classification increases, but a larger number of bands are required, and the storage capacity increases. The opposite is true if the threshold is set low.

  The band used for the calculation of the cumulative independence is set as a classification band of the classification class (S107). That is, since bands having high independence with respect to other classification classes are selected, it is sufficient to use those bands when identifying the classification classes. For example, in the example of FIG. 4, if the cover classification process is performed using the data of bands 1, 6, and 7, the classification class A can be distinguished from other classification classes. Furthermore, the number of bands for specifying the classification class can be limited to the number of selected bands. In this way, the band for identifying the classification class selected in step S102 is selected. Then, the processes from step S102 to S107 are repeated until there is no unprocessed classification class. In this way, bands for identification are selected for all classification classes, and the corresponding data of each classification class and its classification band is stored in the storage device 20 as classification band data 22.

  FIG. 5 is an example of the classification band data 22. For example, when the user needs a distribution of “forest” in a certain area, refer to the classification band data and categorize the “forest” class when photographing the area with a satellite or a camera mounted on an aircraft. It is sufficient to store only the band on the storage medium. For this reason, since it is not necessary to hold all the band data of the hyperspectral image data, the storage capacity can be reduced. Further, since it is sufficient to perform the cover classification process using only the necessary bands, the processing time can be shortened.

  It is also possible to obtain a distribution of a plurality of classification classes for one geographic image. In that case, a classification band for each classification class may be stored in the storage device at the time of photographing a geographic image. For example, when the distribution of “forest” and “water area” is obtained, the classification band of “forest” class and the classification band of “water area” class may be stored in the storage medium at the time of photographing a geographic image.

  Then, the process of the covering classification processing unit 200 will be described. The cover classification processing unit 200 obtains the classification band data 22 by the classification band selection processing unit 100, and obtains band image data necessary for the cover classification obtained by photographing from an artificial satellite or the like. FIG. 6 is a flowchart showing the flow of processing by the cover classification processing unit 200 of the geographic image processing program 40. In FIG. 6, the cover classification processing unit 200 acquires information on a classification class that the user wants to cover and classifies, and based on the information, the classification band data 22 (spectrum band information in FIG. 5) and its classification band are stored from the storage device 20. Geographic image data 23 composed of spectral data corresponding to the data is acquired (S201). The geographical image data 23 here is hyperspectral image data, and only the data of the band necessary for classifying the classification class designated by the user is stored in the storage device 20. FIG. 7A is a diagram schematically showing one band data in the geographic image data 23.

  Next, in steps S202 to S205, each classification class specified by the user is covered and classified one class at a time. First, one unprocessed classification class is selected (S202). Next, a classification band of the classification class is acquired from the classification band data (S203). Then, covering classification of the classification class is performed using the classification band (S204) (refer to Non-Patent Document 1 for a specific method of covering classification). As a result, as shown in FIG. 7B, the distribution of the classification class (forest in FIG. 7B) selected in step S202 is known. If there is only one classification class designated by the user, the process ends here, and if there are two or more classification classes, the process returns from step S206 to step S202 to repeat the process. In this way, the cover classification is performed for all the specified classification classes. FIG. 7C is a diagram schematically showing the processing result. In this figure, forests, soils, and water areas are specified as classification classes. Areas that do not correspond to any classification class are grouped together as other areas.

<Summary>
According to the present embodiment, as the teacher data, each classification class has spectrum data corresponding to a plurality of wavelength bands (see FIG. 3), and one processing target classification is based on the spectrum data for each classification class. A characteristic wavelength band is selected for the class compared with other classification classes. Information on the wavelength band of the selected classification class to be processed is stored as classification band data used when performing the covering classification process (see FIG. 5). If this classification band data is used, it is not necessary to process all bands (wavelength regions), so that the covering classification process can be performed efficiently in a short time. Further, since it is not necessary to store spectrum data for all bands of geographic image data such as an aerial photograph image that is the target of the cover classification process, the capacity of the storage device of the apparatus can be greatly reduced.

  More specifically, for the band selection process, for each wavelength region, the difference absolute value of the spectrum data of the processing target classification class and the other classification class is calculated, and the minimum difference absolute value is obtained in each wavelength region. Among the minimum difference absolute values in the wavelength region, information on the wavelength region of a combination in which the sum exceeds a predetermined threshold value is selected. More specifically, when the minimum difference absolute value in each wavelength region is rearranged in order of increasing values, a cumulative addition value is calculated from the large minimum difference absolute value, and the cumulative addition value exceeds the predetermined threshold value. The information of the wavelength region constituting the cumulative addition value may be selected. In this way, a characteristic wavelength region (band) can be selected reliably and efficiently for each classification class (for example, forest, soil, water region, etc.).

  In the present embodiment, a process of covering and classifying a geographic image having spectrum data corresponding to a plurality of wavelength bands is also disclosed. In the present embodiment, classification band data indicating a wavelength region used when covering classification processing is stored in a storage device for a plurality of classification classes (for example, forest, soil, water area, etc.). Then, a classification class is designated by the user, and classification band data is acquired from the storage device for the designated classification class. A geographic image composed of spectral data corresponding to a wavelength region included in the obtained classification band data is acquired, and a cover classification process is executed using the spectral data of the geographic image for each designated classification class. In this way, when photographing the area to be covered and classified, only the necessary bands need be stored, so that the capacity of the recording device for storing geographic image data can be reduced. For example, when a 10 km square is captured at a ground resolution of 1 m, the image capacity of 100 bands is 10 gigabytes, but if the required number of bands is 20 bands, it becomes 2 gigabytes, and the storage capacity can be greatly reduced. In addition, the processing time can be reduced because the number of data to be processed is small. Note that the information on the wavelength region forming the classification band data is selected by the above-described procedure and stored in the storage device as the classification band data.

  The present invention can also be realized by a program code of software that realizes the functions of the present embodiment. In this case, a storage medium in which the program code is recorded is provided to the system or apparatus, and the computer (or CPU or MPU) of the system or apparatus reads the program code stored in the storage medium. In this case, the program code itself read from the storage medium realizes the functions of the above-described embodiments, and the program code itself and the storage medium storing the program code constitute the present invention. As a storage medium for supplying such program code, for example, floppy (registered trademark) disk, CD-ROM, DVD-ROM, hard disk, optical disk, magneto-optical disk, CD-R, magnetic tape, non-volatile A memory card, ROM, or the like is used.

  Also, based on the instruction of the program code, an OS (operating system) running on the computer performs part or all of the actual processing, and the functions of the above-described embodiments are realized by the processing. May be. Further, after the program code read from the storage medium is written in the memory on the computer, the computer CPU or the like performs part or all of the actual processing based on the instruction of the program code. Thus, the functions of the above-described embodiments may be realized.

  In addition, the program code of the software that realizes the functions of the embodiment is distributed via a network, so that it is stored in a storage means such as a hard disk or memory of a system or apparatus or a storage medium such as a CD-RW or CD-R. It may also be achieved by the computer (or CPU or MPU) of the system or apparatus reading and executing the program code stored in the storage means or the storage medium.

It is a block diagram which shows roughly the structure of the geographic image processing system of this invention. It is a flowchart of the classification band selection process of a geographic image processing program. It is a figure which illustrates the outline | summary of the hyperspectral characteristic of a covering classification item. It is explanatory drawing of the classification band selection process of the geographic image processing program shown in FIG. It is a figure which illustrates the classification | category band data of this invention. It is a flowchart of the cover classification process of a geographic image processing program. It is a figure which illustrates the geographic image data and cover classification result of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Processing device 20 Storage device 21 Teacher data 22 Classification band data 23 Geographic image data 30 Input / output device 31 Input device 32 Display device 33 Printer 40 Geographic image processing program 100 Classification band selection processing unit 200 Cover classification processing unit

Claims (8)

A geographic image processing apparatus for processing a geographic image having a plurality of wavelength bands,
Learning data storage means for storing spectrum data corresponding to a plurality of wavelength bands for each classification class;
Band selection processing means for selecting a characteristic wavelength band in comparison with other classification classes for one classification class based on spectrum data for each classification class;
Classification band data storage means for storing wavelength band information of the classification class to be processed selected by the band selection processing means as classification band data used when performing the covering classification processing;
A geographic image processing apparatus comprising:   The band selection processing unit calculates a difference absolute value of spectrum data of the classification class to be processed and another classification class for each wavelength region, acquires a minimum difference absolute value in each wavelength region, 2. The geographic image processing apparatus according to claim 1, wherein information on a wavelength region of a combination in which the total sum exceeds a predetermined threshold value among absolute difference absolute values is selected.   The band selection processing means rearranges the minimum difference absolute value in each wavelength region in order of increasing values, calculates a cumulative addition value from the large minimum difference absolute value, and the cumulative addition value exceeds the predetermined threshold value. The geographic image processing apparatus according to claim 2, wherein information on a wavelength region constituting the cumulative addition value is selected. A geographic image processing apparatus for covering and classifying a geographic image having spectral data corresponding to a plurality of wavelength bands,
For a plurality of classification classes, classification band data storage means for storing classification band data indicating a wavelength region to be used when covering classification processing,
Means for obtaining data of a specified classification class;
Processing image acquisition means for acquiring the classification band data for the specified classification class, and acquiring a geographic image composed of spectrum data corresponding to a wavelength region included in the acquired classification band data;
Cover classification processing means for performing cover classification processing using spectral data of the geographic image obtained by the processed image acquisition means for each designated classification class;
A geographic image processing apparatus comprising:   The information on the wavelength region forming the classification band data is selected by the band selection processing means based on the spectrum data for each classification class, and a characteristic wavelength band is selected for one processing target classification class compared to other classification classes. The geographic image processing apparatus according to claim 4, wherein:   The band selection processing unit calculates a difference absolute value of spectrum data of the classification class to be processed and another classification class for each wavelength region, acquires a minimum difference absolute value in each wavelength region, 6. The geographic image processing apparatus according to claim 5, wherein information on a combination of wavelength regions in which the sum exceeds a predetermined threshold value among the minimum difference absolute values is selected.   The band selection processing means rearranges the minimum difference absolute value in each wavelength region in order of increasing values, calculates a cumulative addition value from the large minimum difference absolute value, and the cumulative addition value exceeds the predetermined threshold value. The geographic image processing apparatus according to claim 6, wherein information on a wavelength region constituting the cumulative addition value is selected.   A program for causing a computer to function as the geographic image processing apparatus according to any one of claims 1 to 7.

Images