JP3228920B2 - Image processing method and storage medium storing image processing program - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a method for determining, using a computer, a change area between two images, new and old, which are taken at an interval in the same area. INDUSTRIAL APPLICABILITY The present invention is suitably applied to, for example, a task of grasping land cover change based on an image obtained by photographing from an aircraft or an artificial satellite. In this specification, land cover change refers to the construction, remodeling, and construction of terrestrial structures (buildings) such as roads, houses, and factories.
Changes in land cover caused by remodeling, demolition, etc. Note that the land change area may be simply referred to as a change area. Note that the present invention is not limited to the above-described case of obtaining land cover change, but can be applied to various applications for determining a change area by comparing two images taken of the same area with a time difference. It is.

[0002]

2. Description of the Related Art To compare land and land changes in an area including buildings such as roads, houses, factories, and the like by comparing two new and old images taken from the sky with time, for example, This is useful for understanding changes in fixed assets. Conventionally, the following manual operation has been required to specify such a change area.

First, new and old images (for example, a 20 cm square color positive film) of the same area photographed from the sky are created. Next, these color positive films are placed on a translucent plate or a transparent plate of a light table, and a map of a corresponding area is prepared on the side of the light table. Looking at both color positive films, search for the changed area, and transfer the searched changed area to a paper map. After writing all the changed areas on the map, the map is digitized using a digitizer such as a scanner to digitally process the changed areas.

[0004]

The above-mentioned conventional "identification of the change area by the human eye" has the following problems. That is, since the criterion for determining whether or not the area is a change area differs depending on the operator, the quality of the products is not uniform, and the quality control of the final result is very difficult. Furthermore, there is a problem that a space for placing the light table and the map is required, so that a large work area is required. Furthermore, it takes a long time to transfer the changed area determined by comparing the two films to the map, and a considerable burden is imposed on the operator.

[0005]

SUMMARY OF THE INVENTION Accordingly, it is an object of the present invention to solve the conventional problem associated with manual work by digitally determining the change area of a new and old image using a computer.

[0006]

According to an image data processing method of the present invention, a new image and an old image obtained by photographing the same area from the sky at a time interval are prepared. First and second converted images are obtained from each of the new and old images by a first image conversion method, and the first and second converted images are obtained.
Are compared to obtain a first binarized image having data estimated to be a change area, and (b) third and fourth converted images are converted from each of the new and old images by a second image conversion method. And the third and fourth converted images are compared to obtain a second binarized image having data estimated to be a change region. (C) The first binarized image is converted to a predetermined number of pixels. Is divided into a plurality of pixel regions, and the number of pixels indicating a region change in each pixel region is obtained, and the number of pixels is compared with a threshold value to obtain a third binarized image. The second binarized image is divided into a plurality of pixel regions each having a predetermined number of pixels, the number of pixels indicating an area change in each of the pixel regions is obtained, and the number of pixels is compared with a threshold value to obtain a fourth number. Calculating a binarized image, and (e) specifying a change area between the new and old images based on the third and fourth binarized images That.

[0007]

Embodiments of the present invention will be described below with reference to the accompanying drawings.

FIG. 1 is a flowchart schematically showing an embodiment according to the present invention. As shown in FIG. 1, new and old images of a multi-band (to be described later) are input to a computer via an image input unit 10. Next, the color tone of the new and old images is adjusted by the color tone correction unit 12.

The image input unit 10 uses a known analog-to-digital converter to digitize the input new and old analog images and inputs them to a computer.

On the other hand, a color tone correction unit 1 which is a pre-processing unit for an image
In 2, the color tone correction is performed on new and old image data having different photographing times (hereinafter, the image data may be simply referred to as an image). In other words, the two images, old and new, having different photographing times have a color tone difference based on a difference in the altitude of the sun and atmospheric conditions at the time of photographing. For example, if the old image has a strong blue hue overall, but the new image has a red hue, the two images need to be unified in order to perform the subsequent processing accurately. There is. This color tone correction data processing is performed using a density histogram. Since this data processing technique is a well-known technique, a detailed description is omitted.

The following data processing units 14, 16, and 18
Will be described later with reference to FIG. 2 to FIG. 6, so that only the outline will be described here.

Data processing unit 14: a first changing region extracting unit, which converts (compresses) each of the multi-band new and old images into a one-band image by a first image conversion method, and performs first and second conversion. An image is obtained, the first and second converted images are compared to obtain (estimate) a change area, and a first binarized image indicating the change area is obtained.

A data processing section 16 is a second change area extracting section, and converts each of the new and old images into one by a second image conversion method.
Converting into band images to obtain third and fourth converted images,
The third and fourth converted images are compared to determine (estimate) a change area, and a second binary image indicating the change area is determined.

A data processing unit 18 for dividing each of the first and second binarized images into a plurality of pixel areas each including a predetermined pixel, and calculating the number of pixels indicating a change area in each of the pixel areas; The number of pixels is compared with a threshold value to obtain third and fourth binarized images. Based on the third and fourth binarized images, image information indicating a change area between the new and old images is obtained. Ask.

The final image data (information) indicating the change area between the new and old images obtained by the data processing section 18 is recorded on an appropriate storage medium in the data output section 20, or is stored in a display device, a printer, or the like. Is output.

Each of the new and old images input to the image input unit 10 of FIG. 1 is a three-band image of RGB (R: red, G: green, B: blue), or a three-band image such as a near infrared band. Is an image of four or more bands. The data processing unit (first change region extracting unit) 14 first converts (compresses) the above-described multiband image into one band.

Referring to FIG. 2, the operation of data processing unit 14 will be described.

The multiband old image that has been subjected to the color tone correction by the color tone correction section 12 in FIG. 1 is converted into a one-band first principal component image by principal component analysis (steps 22 and 2).
4). Converted image obtained by principal component analysis (first principal component image)
Is denoted by A. By performing the principal component analysis, it is possible to process not only a three-band image (such as an aerial photograph image of an RGB band) but also a multi-band image of four or more bands including a near infrared band and the like.

Principal component analysis is a technique of multivariate analysis, and is a known technique that focuses on the correlation between bands of a multiband image and combines them into one band without losing information as much as possible. That is, the principal component analysis is a method of obtaining a feature axis of multidimensional data, which obtains a correlation matrix or a variance / covariance matrix of original data, and obtains an eigenvalue and an eigenvector. The inner products of the eigenvectors corresponding to the eigenvalues and the original data in the descending order of the eigenvalues are referred to as a first principal component, a second principal component, and so on, respectively. For the principal component analysis, see, for example, “Remote Sensing Dictionary” published by Kyoritsu Shuppan Co., Ltd.
Please refer to.

Similarly to the data processing described above, the new multi-band image which has been subjected to the color tone correction by the color tone correction section 12 in FIG. 1 is subjected to principal component analysis in steps 26 and 28 to obtain the first principal component image. B is required.

In steps 30 and 32, the absolute value of the difference between the corresponding pixels of the first principal component images A and B is determined.
Create an absolute difference image. Next steps 34 and 36
Then, the absolute value image of the difference is binarized by using “threshold value (for convenience of explanation, threshold value A)” to obtain a binarized image. A well-known “representative value filter process” is performed on the binarized image to remove a change portion of a detail caused by an alignment error between the new and old images, a difference in a shadow position, and the like.
The first output image shown in FIG.
Is obtained (step 40).

Incidentally, the first principal component images A and B are considered to represent the total reflection luminance in a Landsat image in Japan. In the above description, the first principal component image is obtained by performing the principal component analysis. However, the brightness (luminance) information is represented using a method other than the principal component analysis for each of the input new and old images after the color tone correction. After determining the images and determining the "absolute difference image" of these images, step 34,
After the processing of 36, 38, and 40, the data processing unit 14
The first output image which is the final output image in FIG.
May be obtained.

Referring to FIG. 3, the operation of data processing unit 16 (FIG. 1) will be described.

The multiband old image that has been subjected to the color tone correction by the color tone correction unit 12 in FIG. 1 is converted (compressed) into one band using the HSI conversion, and a saturation image is obtained (steps 42 and 44). The HSI conversion is for converting an RGB image represented by three primary colors of light into information on three attributes (H: hue, S: saturation, I: lightness) of the color of the image.
Among them, the saturation has a feature that it is hardly affected by a shadow of a building or the like. In the next step 46, the shape of a building or a road is extracted by applying a known Sobel filter, and a shape image A is obtained in a step 48.

As in the case described above, the color tone correction unit 1 shown in FIG.
In steps 50, 52, and 54, the same data processing as described above is performed on the new multiband image that has been subjected to the color correction in step 2, and a shape image B is obtained in step 56.

In step 58, the absolute value of the difference between the corresponding pixels of the shape images A and B is determined, and an image of the absolute value of the difference is created (step 60). Next, in steps 62 and 64, the absolute value image of the difference is binarized using "threshold value B" to obtain a binarized image. Further, in step 64,
A representative value filter process is performed on the binarized image to remove a part where details change due to an alignment error between the new and old images, a difference in the position of a shadow, and the like. A second binarized image "indicating the change area is obtained.

The data processing step 18 of FIG. 1 will be described in detail with reference to FIGS.

Referring to FIG. FIG. 4A shows an example of the “binary image showing the changing area” obtained in step 40 of FIG. In this embodiment, it is assumed that the binarized image shown in FIG. 4A includes three change regions. In order to divide the image of FIG. 4A into a plurality of pixel regions including a predetermined number of pixels, a grid is overlaid (the image is divided into a grid) as shown in FIG. 4B.
Assuming that each side of each grid corresponds to 25 m on the ground and 0.5 m square on the ground corresponds to one pixel, each grid has 2
500 pixels (= 50 × 50 pixels) are included.

After the image shown in FIG. 4A is divided into a plurality of pixel areas, the number of pixels indicating a change area in each pixel area is counted. FIG. 4C shows an example of “the number of pixels indicating a region change” in each grid. The "number of pixels indicating the area change" in each grid is compared with a threshold value (for example, 686 pixels (about 27% of 2500 pixels)) to obtain a binarized image (FIG. 4D) in grid units. In other words, grids that are equal to or larger than the threshold value have a region change, while grids that are smaller than the threshold value have no region change. The number of all pixels (2500 pixels) and the threshold value included in each of the above-described grids are examples, and these numbers may be changed according to a captured image in order to accurately obtain a change area. .

Referring to FIG. FIG. 5A shows an example of a “binary image showing a changing area” obtained in step 68 of FIG. In this embodiment, it is assumed that the binarized image shown in FIG. 5A includes one change area. The following image data processing, that is, the image of FIG. 5A is divided into a plurality of pixel regions, the number of pixels indicating a region change in each pixel region (grid) is obtained, and a threshold value (in the case of FIG. The process of obtaining a binarized image in grid units in comparison with 353 pixels (about 14% of 2500 pixels) is the same as in the case of FIG. That is, (A) to (D) of FIG.
(A) to (D).

Referring to FIG. 6, data processing for finally specifying a change area will be described. FIG. 4 (D) and FIG. 5 (D)
The OR operation is performed on the binarized image shown in (1) to obtain a “binary image specifying a change area” 70.

As described above, the binarized image obtained by the data processing of FIGS. 2 and 3 is divided into a plurality of pixel areas, the number of pixels showing the area change in each pixel area is obtained, and the threshold value is obtained. Is used to obtain a binarized image with "pixel area as a unit" (FIGS. 4 and 5), and a logical OR operation is performed on these two binarized images to obtain final image data. (FIG. 6).
By dividing the image into a plurality of pixel areas and performing data processing in this way, it is possible to ignore small parts that are not clear as to whether they are change areas included in the captured image. Therefore, it is possible to remove an error caused by a shift in the photographing position and a shift in the position of the shadow.

The above-described image data processing procedure is processed using a computer. The present invention includes the data processing procedure itself and a storage medium that records a computer program that records the data processing procedure.

[0034]

As described above, according to the present invention,
Since the change area of the old and new images is digitally obtained by using a computer, the conventional problems associated with manual work can be solved. Each of the two binarized images obtained in the first stage is divided into a plurality of pixel regions, the number of pixels indicating a region change in each pixel region is obtained, and “pixel unit” is determined using a threshold. Are obtained, and a logical OR operation is performed on these two binary images to obtain final image data. By dividing data into a plurality of pixel regions and performing data processing in this manner, there is an effect that a small portion that is not clear as to whether it is a change region included in a captured image can be ignored.

[Brief description of the drawings]

FIG. 1 is a flowchart of image data processing for obtaining a change area between new and old images according to the present invention.

FIG. 2 is a flowchart showing data processing in a first change area extraction unit according to the present invention.

FIG. 3 is a flowchart showing data processing in a second change area extraction unit according to the present invention.

FIG. 4 is a diagram showing “previous data processing” for finally specifying a change area between new and old images according to the present invention.

FIG. 5 is a view showing “previous data processing” for finally specifying a change area between new and old images according to the present invention.

FIG. 6 is a view showing data processing for finally specifying a change area between new and old images according to the present invention.

[Explanation of symbols]

 22, 26: principal component analysis 23: first principal component image A 28: first principal component image B 32: absolute value image of difference 36: binary image 40: binary image 42 showing a change area HSI conversion 44, 52: Saturation image 48: Shape image A 50: Shape image B 64: Binary image 68: Binary image showing a change area

Continued on the front page (58) Fields investigated (Int.Cl. ⁷ , DB name) G06T 1/00 285 G06T 7/20 200 H04N 7/18 Patent file (PATOLIS) JICST file (JOIS)

Claims (9)

(57) [Claims] 1. A method for identifying a change region between a new image and a new image photographed from above in the same area with a time interval, comprising: (a) first and second images from each of the new and old images by a first image conversion method; (B) comparing the first and second converted images to obtain a first binarized image having data estimated as a change area, and (b) obtaining a second binarized image from each of the new and old images. The third and fourth converted images are obtained by the image conversion method described above, the third and fourth converted images are compared, and a second binarized image having data estimated as a change area is obtained, (c The first binary image is divided into a plurality of pixel regions each having a predetermined number of pixels, the number of pixels indicating a region change in each of the pixel regions is obtained, and the number of pixels is compared with a threshold to determine the number of pixels. (D) converting the second binarized image into a plurality of images each having a predetermined number of pixels; (E) calculating a fourth binarized image by comparing the number of pixels with a threshold to obtain a fourth binarized image; 4. An image data processing method for specifying a change area between the new and old images based on the binarized image of No. 4. 2. The image data processing method according to claim 1, wherein said first and second transformed images are first principal component images created by using principal component analysis. 3. The image data processing method according to claim 1, wherein said first and second converted images are saturation images created by an HSI conversion method. 4. The image data processing method according to claim 1, wherein the first and second converted images are images created based on brightness information of the image. 5. The image data processing method according to claim 1, wherein the plurality of pixel regions are regions obtained by partitioning the entire image into a grid. 6. The image data processing method according to claim 5, wherein, when comparing the third and fourth binarized images, a logical OR of corresponding binary values given to a pixel area is calculated. . 7. A method for identifying a change area between a new image and a new image taken from above in the sky with a time interval in the same area, comprising: (a) first and second images from each of the new and old images by a first image conversion method; (B) comparing the first and second converted images to obtain a first binarized image having data estimated as a change area, and (b) obtaining a second binarized image from each of the new and old images. The third and fourth converted images are obtained by the image conversion method described above, the third and fourth converted images are compared, and a second binarized image having data estimated as a change area is obtained, (c The first binarized image is divided into a plurality of pixel regions each having a predetermined number of pixels, the number of pixels indicating a region change in each of the pixel regions is obtained, and the number of pixels is compared with a threshold to determine the number of pixels. (D) converting the second binarized image into a plurality of images each having a predetermined number of pixels; (E) calculating the fourth binarized image by calculating the number of pixels indicating the area change in each pixel area and comparing the number of pixels with a threshold value; 4. A recording medium which records a program for specifying a change area between the new and old images based on the binarized image of No. 4. 8. A storage medium storing a program according to claim 7, wherein said plurality of pixel areas are areas obtained by partitioning the entire image data in a grid pattern. 9. A storage medium storing a program according to claim 7, wherein when comparing the third and fourth binarized images, a logical sum of corresponding binary values given to a pixel area is obtained.