JP2003288582A - Method, apparatus, and program for urban area extraction, and recording medium for the program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To stably extract an urban area by paying attention to texture features in the urban area when the urban area is automatically extracted from a satellite photographic image, an aerial photographic image, etc. <P>SOLUTION: A two-dimensional edge vector is obtained as to each pixel of an image of an object of urban area extraction and in a block including the pixel as the center, the sum of outer products of a combination of two of all vectors is calculated to decide whether a representative point in the block is in an urban area. Further, lateral and longitudinal edge differences of each pixel in the block including each pixel as the center are calculated to find an edge tangent parameter and Hough transformation is performed for it to decide whether a representative point in the block is in the urban area according to a polling result. Furthermore, a water area and a forest area are found by using images of the red range and near infrared ranges of visible light and used to decide on the urban area. <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an urban area area extraction for automatically extracting an urban area area from an image taken from the sky such as a satellite photographic image or an aerial photographic image based on its texture information and multispectral information. The present invention relates to a processing method, an urban area extraction processing device, an urban area extraction processing program, and a recording medium of the program.

[0002]

2. Description of the Related Art In the conventional urban area extraction technology, attention is paid to a closed area satisfying a specific color or brightness condition, and when the shape of the closed area is close to that defined in advance, the closed area is constructed. It was assumed that it was a building area, and the buildings extracted by such a method were integrated and recognized as an urban area.

[0003]

However, the conventional method of this type is closed when the image is monochrome or grayscale and has no color information, or when a part of the building is hidden by a tree or another building. Stable urban area extraction can be performed when the area shape is not taken, or when each building does not have a sufficient size due to the low image resolution and the shape of the closed area is collapsed. There was a problem that I could not.

The present invention focuses on the texture feature in the urban area, that is, the feature that the edge power is strong and the edge tangential directions are orthogonal to each other in the urban area. Even in such a case, the objective is to enable stable extraction of urban areas.

Here, the edge power means the length of the two-dimensional edge vector of each point of the image, and the edge tangent line is the straight line whose normal vector is the two-dimensional edge vector of the point at each point of the image. Say.

[0006]

In order to solve the above-mentioned problems, the present invention determines the urban area based on the feature quantity which quantitatively expresses the orthogonality of the edge power and the edge tangent of the urban area. The method is used. 1. A water region and a forest region are extracted using an image capturing an electromagnetic wave having a wavelength in the red region of visible light (referred to as an R image) and an image capturing an electromagnetic wave in the near infrared region (referred to as an NIR image). 2. Appropriate preprocessing using the result of the above 1 is applied to the image of the urban area extraction target. 3. A two-dimensional edge vector is acquired for each pixel of the image of the urban area extraction target, and the absolute value of the vector cross product is calculated and added for all combinations of two vectors in the block centered on each pixel. The absolute value of the cross product of the vectors A and B is calculated as | A × B | = | (| A || B | sin θ) |, where θ is the angle formed by these two vectors. The larger the power and the stronger the orthogonality between the two edge vectors, the larger the value. The vector sum of vector cross products obtained as a result is given to the representative point in the block as a feature amount, and based on this value, it is determined whether or not the point is an urban area. 4. An arbitrary size block centered on each pixel is set for the image of the urban area extraction target, horizontal and vertical edge difference calculations are performed at each pixel, and the angle parameter of the edge tangent is calculated based on the obtained edge difference value. Find the distance parameter and. Hough transform is performed on the calculated edge tangent parameter, and each edge tangent parameter is voted in a two-dimensional voting space having two axes, an angle parameter and a distance parameter. An appropriate feature amount is obtained from the obtained voting space, and based on this, it is determined whether the representative point of the block is the urban area. 5. The final urban area is obtained by adding the urban area extracted in 3 above and the urban area extracted in 4 above and excluding the forest area extracted in 1 above.

The above-mentioned method 3 is effective for extracting an urban area where small houses and the like are gathered, and a good result is obtained when the block size is about the size of a house. In addition, the above 4
Method is effective for extracting large buildings such as factories,
It is preferable to use a block size larger than that used in 3 above.

By the above method, it is possible to stably extract the urban area based on the texture characteristics of the urban area.

[0009]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 1 shows a structural example of an urban area extraction processing apparatus according to this embodiment. In the figure, 1 is an urban area extraction processing device including a CPU and memory, 10 is an R image obtained by capturing the ground surface with electromagnetic waves having a wavelength in the red region of visible light, such as satellite photographic images and aerial photographic images, 11 is an R image 10
NIR image obtained by capturing the same point with electromagnetic waves in the near infrared region, 12 is an image obtained by visualizing the same point as the R image 10 and the NIR image 11 with visible light (hereinafter referred to as P image), 20
Is a water area extraction processing unit that extracts water areas from the R image 10 and the NIR image 11, and 30 is the R image 10 and the NIR image 11
A forest area extraction processing unit for extracting a forest area from the above, 13 an extracted forest area, 14 an extracted water area, 40 a histogram flattening processing unit for the P image 12, 5
0 is an urban area extraction processing unit having a vector outer product sum as a feature amount, 15 is an urban area A extracted when the inner product sign is not checked in the urban area extraction processing unit 50,
Reference numeral 16 is an urban area area B extracted when the sign of the inner product is checked, 60 is an urban area area extraction processing unit using Hough features, 17 is an urban area area C extracted by the processing of the urban area area extraction processing unit 60, and 70 is an urban area. A merge processing unit that integrates the area A15 and the urban area C17, 18 is a merge result obtained as a result of the merge processing unit 70, 80 is a forest area correction processing unit that corrects the forest area for the merge result 18, and 19 is the final result. It is an urban area that can be obtained.

The R image 10 is an image obtained by photographing the ground surface from the sky with an electromagnetic wave having a wavelength in the red region of visible light using an artificial satellite or an aircraft.

The NIR image 11 is an image obtained by photographing the ground surface from the sky with an electromagnetic wave having a wavelength in the near infrared region, using an artificial satellite or an aircraft. Generally, an artificial satellite or an aircraft for surface imaging is equipped with a device for capturing an R image and an NIR image.

The P image 12 is an image obtained by photographing the ground surface with visible light from above using an artificial satellite or an aircraft. Generally, the P image has a higher resolution than the R image and the NIR image. Therefore, in the present embodiment, a method of extracting the urban area from the P image will be described. R image or NI instead of P image
An R image or an image photographed in another wavelength region can be used as an image for extracting the urban area.

The water area extraction processing unit 20 is a processing means for extracting a water area from the R image 10 and the NIR image 11 taken at the same point. First, there may be a case where alignment of two images is required due to affine transformation or the like. After alignment, at each corresponding point, X = (luminance value in NIR image) / (luminance value in R image) is calculated. The reflectance of water differs greatly between the red region of visible light and the near-infrared region, and the reflectance in the near-infrared region is much smaller than that in the red region of visible light. Therefore, X has a small value in the water region. Therefore, when the value of X is equal to or less than a certain value, the point is determined to be the water region. Although the value of this constant value varies greatly depending on the illumination conditions and the camera characteristics, for example, when NIR / R <0.7, the pixel is determined to be the water region.

In addition to the above X, a method in which the difference between the brightness value in the NIR image 11 and the brightness value in the R image 10 or NDVI described below is used as the water area judgment scale is also conceivable.

The forest area extraction processing unit 30 is similar to the water area extraction processing unit 20 in that the R image 10 and the N image taken at the same point are taken.
It is a processing means for extracting the forest area from the IR image 11. In this case as well, there may be a case where the two images need to be aligned by affine transformation or the like. Then, at each corresponding point, NDVI = (luminance value in NIR image−luminance value in R image) / (luminance value in NIR image + luminance value in R image) is calculated. In the forest area, the value of NDVI becomes larger than the spectral reflectance characteristic of chlorophyll (see: "Illustrated Remote Sensing", Japan Surveyors Association). When the value of NDVI is a certain value or more, the point is determined as a forest area.

The forest area 13 is a forest area extraction processing unit 30.
Is the region extracted by. For the forest area 13,
There is also a method of performing image processing such as noise removal processing, blurring processing, maximum value filtering processing, and minimum value filtering processing as needed.

The water area 14 is an area extracted by the water area extraction processing section 20. There is also a method of performing image processing such as noise removal processing, blurring processing, maximum value filter processing, and minimum value filter processing on the water region 14 as necessary.

The histogram flattening processing section 40 is a processing means for performing luminance conversion on the P image as necessary,
In this embodiment, histogram flattening (reference: “Introduction to computer image processing”, Research Institute Publishing, supervised by Hideyuki Tamura) is performed. By this processing, the contrast between the roof part and the shadow part of the building is increased, and the texture features in the urban area are further emphasized.

Further, in the present embodiment, a method of using the water region 14 at the time of histogram flattening will be described. If there are large areas of low brightness in the image, such as the sea or lake, the histogram flattening process may not show the appropriate contrast enhancement effect. Therefore, the luminance histogram necessary for the histogram flattening process is acquired in the part excluding the water region, and the histogram flattening process itself is applied only to the part excluding the water region. By using such a method, the contrast in the urban area is enhanced even when there are large seas or lakes in the image.

The urban area extraction processing unit 50 using the vector cross product sum is a processing means for extracting the urban area using the vector cross product sum as a feature amount. FIG. 2 shows the flow of processing by the urban area extraction processing unit 50 using the vector cross product sum.

The input image (P image) 51 is an image output by the histogram flattening processing unit 40. The edge vector acquisition process 52 is a process of performing a difference calculation on an image to acquire an edge vector. The effect of noise can be reduced by performing blurring processing with a Gaussian filter or the like before performing the difference calculation.

In the present embodiment, a method of calculating the difference between the horizontal direction and the vertical direction will be described. FIG. 3 is a diagram illustrating a method of calculating the edge vector. The difference calculation in the horizontal direction is performed by subtracting the pixel values on the left and right of the pixel of interest.
The difference calculation in the vertical direction is performed by subtracting the upper and lower pixel values.

In the example shown in FIG. 3A, the edge vector C (C _X , C _Y ) is defined by assuming that the horizontal difference value at the pixel C (point C) is C _X and the vertical difference value is C _Y. Is calculated as C _X = (D−B) / 2 C _Y = (E−A) / 2 from the pixel values of the adjacent pixels A, B, D and E.

Further, with respect to the pixels at the edges of the image, there are cases where there are no adjacent upper, lower, left and right pixels. In this case, the difference between the adjacent pixel and the pixel of interest itself is calculated and used as the edge vector component. Such an example is shown in FIG. Edge vector G at pixel G (point G)
(G _X , G _Y ) is calculated as G _X = H−G G _Y = (I−F) / 2 from the pixel values of the adjacent pixels F, H, and I.

The edge vector information 53 is obtained as a result of the edge vector acquisition processing 52, and two-dimensional edge vectors are given to all pixels. FIG. 4 shows an example of the edge vector information 53. FIG. 4A shows the pixel value of the original image,
(B) shows the value of the edge vector component.

The block acquisition process 54 is a process for acquiring an area of a fixed size centered on the pixel of interest. Figure 5
Shows an example of a block when the size of the area is 3 pixels × 3 pixels.

The outer product sum calculation process 55 is a process of calculating a vector outer product between all combinations of two edge vectors in the block acquired by the block acquisition process 54 and adding the absolute values thereof. here,
As in the example shown in FIG. 6, a combination of all two vectors in a block consisting of 9 pixels, that is, ₉
The absolute value of the vector cross product is calculated and added among C ₂ = 36 combinations.

Further, as shown in FIG. 7A, in order to distinguish a black region on a white background and a white region on a black character, an edge vector V _E of either one of two vectors for calculating an outer product, The inner product of the edge vector V _E and the direction vector V _D from the given edge point toward the other edge point is calculated, and the vector outer product is calculated only when the inner product value is a positive value as shown in FIG. 7B. Figure 7
As shown in FIG. 7C, when the inner product value is a negative value, a method of skipping the addition process of the vector outer product in the combination of these two edge vectors can be considered (FIG. 7D).

The outer product sum 56 is obtained as a result of the outer product sum calculation processing 55, and is given to each point of the image as a feature amount.

The city area determination processing 57 is the outer product sum calculation processing 5
This is a process for determining whether or not the representative point of the block is the urban area based on the outer product sum 56 obtained in 5. The representative point can be arbitrarily set, such as the center point of the block, all points forming the block, and points near the center of the block. Here, the representative point is the center point of the block. Thresholding is performed on the value of the sum of outer products to classify it into two areas, an urban area and a non-urban area. It is also conceivable that the value of the sum of outer products is output as it is as a value representing the urban area without performing threshold processing.

Further, in order to suppress erroneous extraction of coastlines,
A method of canceling the urban area determination in the suburbs of the water area using the water area 14 extracted by the water area extraction processing unit 20 may be considered.

The end determination processing 58 is processing for determining whether or not the block acquisition processing 54, the outer product sum calculation processing 55, and the city area determination processing 57 have been performed for all pixels. If there is a pixel that has not been processed yet, the process returns to the block acquisition process 54, and the city area determination is similarly performed for the unprocessed pixel.

Urban area A (no inner product check) 15
This is the urban area extracted when the sign of the inner product is not checked by the urban area extraction processing unit 50 based on the vector outer product sum. Noise reduction processing, blurring processing, maximum value filtering processing, as necessary for the urban area A15,
There is also a method of performing image processing such as minimum value filter processing.

Urban area B (with inner product check) 1
6 is an urban area extraction processing unit 50 based on the sum of vector cross products
It is an urban area extracted when the sign of the inner product is checked in. There is also a method of performing image processing such as noise removal processing, blurring processing, maximum value filter processing, and minimum value filter processing on the city area B16 as necessary.

The urban area extraction processing unit 60 based on the Hough feature is a processing means for extracting the urban area based on the Hough feature amount calculated by the Hough transform of the edge tangent parameter. FIG. 8 shows a processing flow of the urban area extraction processing unit 60 based on the Hough feature.

The input image (P image) 61 is an image output by the histogram flattening processing unit 40. The block acquisition process 62 is a process of acquiring a block of a fixed size centered on the pixel of interest. The shape of the block is arbitrary, such as a square or a circle, but in the present embodiment, a circular block is acquired. This is because it is convenient for the Hough transform of the edge tangent parameter that is performed thereafter. An example of a circular block is shown in FIG.

The edge tangent line parameter calculation process 63 is a process for obtaining an edge tangent line at each point in the block. Here, a method of first performing a difference calculation in the horizontal and vertical directions on an image to acquire an edge vector, and obtaining an edge tangent parameter based on the acquired edge vector will be described. The processing for obtaining the edge vector is the same as the edge vector acquisition processing 52 described with reference to FIG.

Next, edge tangent parameters are calculated from the obtained edge vector. Here, the center of the block is the origin, and the distance r from the origin to the tangent and the angle θ formed by the tangent normal vector with the X direction are the edge tangent parameters to be obtained.

FIG. 10 is a diagram for explaining an example of edge tangent line parameters. As is apparent from FIG. 10, the edge tangent parameters θ _P and r _{P at} the point _P are easily calculated from the edge vector components P _X and P _Y at the point P.

The Hough transform process 64 performs Hough transform on the calculated edge tangent parameters, and 2
This is a process of voting in a three-dimensional voting space. Figure 11
It is a figure explaining the example of the voting process. Edge tangent parameter theta _P and r _P at point P shown in FIG. 10, be voted to the coordinates (theta _P, r _P) in voting space.

The voted value may be +1 in this case, but in the present embodiment, the edge power value itself, that is, (P
_The voting value is the square root of ( _X ² + P _Y ² ). By this method, the parameters of the edge with strong power, which are often present in the urban area, are more emphasized.

Further, for the purpose of suppressing the adverse effects of noise, it is also effective to perform no voting process when the edge power is a fixed value or less, or to give a weight to the edge power to obtain a voting value. Is.

The voting space 65 is the Hough conversion processing 64.
This is the voting space obtained as a result of performing for all pixels in the block. FIG. 12 shows an example thereof. 12A shows an original image, and FIG. 12B shows an example of a voting space. In the example of FIG. 12A, voting is performed on all the pixels in the circular block 69 having a diameter of 50 pixels. In the voting space of FIG. 12B, the size of the vote value is shown by changing the density.
The horizontal axis of the voting space is the axis of the angle parameter θ (range 0 to 3).
60 °), the vertical axis is the axis of the distance parameter r (range 0 to 2)
5). It can be seen that the straight line in the image is converted into a point with a high vote value in the voting space.

The in-block processing end determination 66 is an edge tangent parameter calculation processing 6 for all pixels in the block.
3, a process of determining whether or not the Hough conversion process 64 is completed. When there is an unprocessed pixel, the process returns to the edge tangent line parameter calculation process 63 again to process the unprocessed pixel.

The urban area B (with inner product check) 16 is
Urban area extraction processing unit 50 based on vector sum of cross products in FIG.
It is the urban area B extracted in. The water area 14 is shown in FIG.
It is the water region extracted by the water region extraction processing unit 20 of FIG.

The city area determination processing 67 is processing for determining whether the representative point of the block is the city area based on the voting space 65. As a representative point of the block, a center pixel of the block, a pixel near the center of the block, all pixels forming the block, and pixels having an edge power of a certain value or more in the block are considered. In the present embodiment, a method is used in which a two-dimensional voting space is not used as it is, but a projection is taken in the X-axis direction (angle parameter axis direction) and a one-dimensionally degenerated histogram (hereinafter referred to as an angle histogram) is used. explain.

FIG. 13 shows an angle histogram created from the voting space shown in FIG. There is also a method to obtain the city area from statistics such as the maximum value, minimum value, average value, and variance value of the angle histogram, but here, the judgment processing based on (1) maximum value / minimum value, (2) edge tangent orthogonality explain.

(1) Maximum value / minimum value The maximum value and the minimum value of the angle histogram are obtained, and the maximum value is divided by the minimum value to calculate the ratio. If this value is greater than a certain value, the block's representative point is determined to be an urban area. Since the origin of tangent parameter calculation is set to the center of the block, the angle parameter θ is 1
There will be a difference of 80 °. So add the 180-360 ° part of the angle histogram to the 0-180 ° part,
By setting the range of the angle histogram to 0 to 180 degrees,
A method of correcting straight lines having the same inclination so as to have the same angle parameter may be considered.

In the case of an image in which the edge directions are scattered like random noise, the maximum value / minimum value is a small value, and the large value is that there is a straight line having a strong edge power in the block. And so on.

(2) Edge tangent orthogonality If there are orthogonal straight lines in the block, two peaks having an interval of 90 ° occur in the angle histogram. Whether or not there are two peaks at 90 ° intervals can be determined by the following method. FIG. 14 shows an explanatory diagram of the determination method.

The histogram shown in FIG. 14 corresponds to the above (1).
The 180 to 360 ° portion is 0 to 1 as described above.
The angle range of the angle histogram is 0-overlapped on the 80 ° part.
It is 180 degrees. Introduction this histogram
The maximum angle at θ ₁Ask for. Then θ₁Or
Angle θ with a difference of 90 ° from₂With any width around
Second peak search area S_AIs set, and the maximum
A large value (second peak value) is calculated. In the example of FIG. 14, θ₂
With a width of ± 10 ° around the second peak search area S_Aof
The range was set.

After the second peak value is obtained, this value is divided by the minimum value of the histogram, and if the ratio is equal to or greater than a certain value, the representative point of the block is determined to be the urban area. In addition to the ratio between the second peak value and the minimum value, the value of the second peak value itself, the difference between the second peak value and the minimum value, or the like may be used as the judgment scale, or the value of the maximum value itself may be added to the judgment scale. Can also be considered.

Further, an angle histogram having a range of 0 to 180 ° is further overlapped with a 90 to 180 ° portion on a 0 to 90 ° portion to create a histogram having an angle range of 0 to 90 °. It is also conceivable to use the ratio between the value and the minimum value as the judgment scale.

Although the method of determining the city area based on the angle histogram has been described here, a method of calculating an appropriate feature amount using the two-dimensional voting space as it is may be considered.

In order to prevent erroneous extraction of paddy fields and fields as buildings, there is also a method of using the urban area B (with inner product check) 16 and performing re-determination together with the above determination. Buildings tend to exist as white areas on a black background, whereas paddy fields and fields are black areas on a white background, and therefore are not extracted in the urban area B (with inner product check) 16. Therefore, when the urban area B (with inner product check) 16 does not exist in the block determined to be the urban area by the above method, the block rejects the previous determination result as the non-urban area.

In addition, in order to suppress erroneous extraction of bridges, landfills, etc., a method of using the water area 14 and rejecting the city area judgment in the suburbs of the water area can be considered.

The end determination process 68 is a process of determining whether or not the urban area determination process has been performed for all pixels. If there are still unprocessed pixels, the process returns to the block acquisition processing 62 again, and the same processing is performed on the unprocessed pixels.

The urban area C17 is an urban area extracted as a result of the urban area extraction processing 60 using the Hough feature. There is also a method of performing image processing such as noise removal processing, blurring processing, maximum value filter processing, and minimum value filter processing on the city area C17 as necessary.

The merge processing unit 70 is a processing means for merging the urban area A15 extracted with the vector outer product sum as the feature quantity and the urban area C17 extracted based on the Hough feature. In the present embodiment, the urban area A15
And the urban area C17 are binarized and added by a simple method. Besides this method, urban area A15
It is also conceivable that the urban area likeness of the urban area C17 is represented by a multi-valued value and the value is added with some weight.

The merge result 18 is a result obtained as a result of the merge processing unit 70. For the merge result 18,
There is also a method of performing image processing such as noise removal processing, blurring processing, maximum value filtering processing, and minimum value filtering processing as needed.

The forest area correction processing unit 80 determines the merge result 1
8 is a processing means for excluding the forest area 13 and extracting the final urban area 19. As a method to exclude
In the present embodiment, the forest area 13 and the merge result 18
Will be binarized and subtracted by a simple method.

The urban area 19 includes the forest area correction processing unit 8
It is the final extraction result of the urban area obtained as a result of 0. There is also a method of performing image processing such as noise removal processing, blurring processing, maximum value filter processing, and minimum value filter processing on the urban area 19 as necessary.

The above-described processing performed by the urban area extraction processing device 1 can be realized by a computer and a software program, and the program is an appropriate computer-readable portable medium memory, semiconductor memory, hard disk, or the like. It can be executed by a computer by storing it in a recording medium and reading it from there. Further, the program can be downloaded from another computer via a communication line, and can be installed and executed.

[0065]

As described above, according to the present invention, the texture feature and the multi-spectral feature in the urban area, especially the texture in which the edge power is strong and the edge tangential directions are orthogonal in the urban area are strong. The urban area is extracted by focusing on the features. Urban area extraction using vector sum of cross products, which is effective for extracting urban areas where small buildings such as houses are gathered, and urban area extraction based on Hough features, which is effective for extracting large-scale buildings such as factories By combining with, it is possible to perform stable urban area extraction that does not depend on the size of the building. Moreover, by extracting the forest area and the water area using the multi-spectral information, it becomes possible to extract the city area with higher accuracy.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration example of an urban area extraction processing device according to the present embodiment.

FIG. 2 is a diagram showing a flow of processing by an urban area extraction processing unit using vector sum of cross products.

FIG. 3 is a diagram illustrating a method of calculating an edge vector.

FIG. 4 is a diagram showing an example of edge vector information.

FIG. 5 is a diagram showing an example of blocks.

FIG. 6 is a diagram showing an example of calculation of vector outer product sum.

FIG. 7 is a diagram illustrating an example of an inner product code check method.

FIG. 8 is a diagram showing a flow of urban area extraction processing based on Hough features.

FIG. 9 is a diagram showing an example of a circular block.

FIG. 10 is a diagram illustrating an example of edge tangent parameters.

FIG. 11 is a diagram illustrating an example of voting processing.

FIG. 12 is a diagram showing an example of a voting space.

FIG. 13 is a diagram showing an example of an angle histogram.

FIG. 14 is a diagram illustrating a second peak search process.

[Explanation of symbols]

1 Urban area extraction processing device 10 R image 11 NIR images 12 P image 13 Forest area 14 Water area 15 Urban area A (no inner product check) 16 Urban area B (with inner product check) 17 Urban area C 18 merge results 19 urban area 20 Water region extraction processing unit 30 Forest area extraction processing unit 40 Histogram flattening processor Urban area extraction processing unit by 50 vector sum of cross products Urban area extraction processing unit based on 60 Hough features 70 Merge Processing Unit 80 Forest Area Correction Processing Section

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Naoki Ito             2-3-1, Otemachi, Chiyoda-ku, Tokyo             Inside Telegraph and Telephone Corporation (72) Inventor Akio Shio             2-3-1, Otemachi, Chiyoda-ku, Tokyo             Inside Telegraph and Telephone Corporation F term (reference) 5B057 AA14 CA08 CA12 CA16 CB08                       CB12 CB16 CE03 CH01 DB02                       DB09 DC13 DC16 DC23                 5L096 AA06 BA20 EA05 FA06 FA24                       FA36

Claims (7)

[Claims] 1. A processing method for automatically extracting an urban area from a photographed surface of a ground by a computer, which comprises an R image capturing an electromagnetic wave having a wavelength in the red region of visible light and a near infrared ray. Using the NIR image that captures the electromagnetic waves in the region, the brightness value and R of the NIR image at each corresponding point
A first process of extracting a water region or a forest region based on a value calculated from the brightness value of the image and a brightness for increasing the contrast of the image of the urban area extraction target by using the result of the first process. The second process of conversion and the calculation of a two-dimensional edge vector for each point of the image that has been subjected to luminance conversion, and in the block of the local area centered on each pixel, a combination of all two vectors of all edge vectors The third process of determining whether the representative point in the block is the urban area or not based on the vector cross product sum obtained as a result of adding and calculating the absolute value of the vector outer product, and the image of the urban area extraction target A block centered on each point is set, an edge tangent is obtained at each point in the block, and an angle parameter and a distance parameter of the calculated edge tangent are calculated. A Hough transform is performed for the two parameters to create a two-dimensional voting space having the two parameters as two axes, and it is determined whether the representative point of the block is an urban area based on the obtained voting space. Process, and the first and third
And a step of extracting a final urban area from the result of the fourth step, the urban area extraction processing method. 2. The urban area extraction processing method according to claim 1, wherein when the outer product of the edge vectors is calculated in the third step, one of the two vectors and the edge vector are calculated. An urban product characterized by calculating the inner product of the direction vector from the given edge point to the other edge point and skipping the addition process of the vector outer product in this combination of edge vectors when the inner product value is a negative value. Area extraction processing method. 3. The urban area extraction processing method according to claim 1, wherein the edge power value is used as a voting value when the Hough transform is performed on the edge tangent parameter in the fourth step. An urban area extraction processing method characterized by the above. 4. The city area extraction processing method according to claim 1, claim 2 or claim 3, wherein in the fourth step, a two-dimensional voting space is projected in the angle parameter axis direction to form a one-dimensional histogram. A method for extracting a city area region, characterized in that the representative point of the block is judged to be the city area based on the histogram. 5. A processing device for automatically extracting a city area area from an image of the surface of the earth by a computer, the R image capturing an electromagnetic wave having a wavelength in the red region of visible light, and a near infrared ray. Using the NIR image that captures the electromagnetic waves in the region, the brightness value and R of the NIR image at each corresponding point
A first means for extracting a water area or a forest area based on a value calculated from the brightness value of the image and a brightness for increasing the contrast of the image of the urban area extraction target by using the result of the first means. A second means for performing a conversion and a combination of all two vectors of all edge vectors in a block of a local area centered on each pixel by calculating a two-dimensional edge vector for each point of the image subjected to the brightness conversion The third means for determining whether or not the representative point in the block is the urban area based on the vector cross product sum obtained as a result of adding and calculating the absolute value of the vector outer product, and the image of the urban area extraction target A block centered on each point is set, an edge tangent is obtained at each point in the block, and an angle parameter and a distance parameter of the calculated edge tangent are calculated. A Hough transform is performed for the two parameters to create a two-dimensional voting space having the two parameters as two axes, and it is determined whether the representative point of the block is an urban area based on the obtained voting space. Means and the first and third
And a means for extracting a final urban area from the result of the fourth means. 6. A city area extraction processing program for causing a computer to execute the city area extraction processing method according to any one of claims 1 to 4. 7. A recording of a city area extraction processing program, characterized in that a program for causing a computer to execute the city area extraction processing method according to any one of claims 1 to 4 is recorded. Medium.