JP2007139421A - Morphological classification device and method, and variation region extraction device and method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a morphological classification device for classification by quantitatively analyzing a morphology of an urban area. <P>SOLUTION: The morphological classification device includes: a comparison image storage part 10 for storing the image data formed based on the aerial photograph taken with an arbitrary urban area; a fractal sequence operation part 11 for calculating the fractal sequence of whole of input image data formed on the basis of the aerial photograph of the arbitrary urban area, and calculating the whole of stored image data stored in the comparison image data storage part 10; an image characteristic distance operation part 12 for calculating the image characteristic distance d based on the fractal sequence of whole input image data and whole stored image data calculated by the fractal sequence operation part 11; and a morphological classification part 13 for classifying the morphology of arbitrary urban area based on the image characteristic distance d calculated by the image characteristic operation part 12. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a form classification apparatus and method for classifying a form of an urban area based on an aerial photograph taken of an arbitrary urban area, and a change area extraction apparatus and method for extracting a change area of the urban area.

  When implementing city planning, it is necessary to conduct a prior assessment that takes into account the impact on the surrounding environment. To that end, it is necessary to quantitatively analyze the structure of a complex city. In addition, it is necessary to quickly grasp changes in the environment due to construction of buildings and roads in the area and collapse of buildings due to disasters. In such a case, a human technique that has been widely used in the past is to discriminate an aerial photograph of a city area by human eyes. In this case, enormous time and labor are required, so the efficiency is low and a great amount of cost is required. Therefore, recently, a method applying pattern recognition has been proposed, and attempts have been made to automate by a computer (see, for example, Patent Document 1).

Japanese Patent Laid-Open No. 11-83478

  However, in the case of a method using pattern recognition, the algorithm is generally complicated, and when implemented in an apparatus, the cost is high and a huge amount of processing time is required. Furthermore, there is a problem that the determination result does not necessarily match the human sense.

  Therefore, the present invention has been made in view of the above-described points, can perform analysis with high reliability equivalent to the case where work is performed by a human technique, and can quickly and easily change the form of an urban area. It is an object of the present invention to provide a form classification apparatus and method for classification, and a change area extraction apparatus and method.

  In order to solve the above-described problems, the form classification device according to the present invention includes an image data storage unit that stores image data generated based on an aerial photograph in which an urban area of a predetermined form is captured. Fractal sequence calculation means for calculating a fractal sequence of input image data generated based on an aerial photograph in which an arbitrary urban area is photographed, and calculating a fractal sequence of the image data stored in the image data storage means And image feature distance calculation means for calculating an image feature distance based on the fractal sequence of the input image data and the fractal sequence of the image data calculated by the fractal sequence calculation means, and the image feature distance calculation. Based on the image feature distance calculated by the means, the arbitrary city Of and a form classification means for classifying the form.

  In addition, the form classification method according to the present invention calculates a fractal sequence of input image data generated based on an aerial photograph of an arbitrary urban area in order to solve the above-described problem, and an image data storage unit Is stored in the image data, and a fractal sequence of stored image data generated based on an aerial photograph of an urban area of a predetermined form is calculated, and the fractal sequence of the input image data and the fractal sequence of the stored image data are calculated. An image feature distance is calculated based on the image feature distance, and the form of the arbitrary urban area is classified based on the image feature distance.

  In addition, in order to solve the above-described problem, the change area extraction apparatus according to the present invention stores image data generated based on an aerial photograph taken of a city in a predetermined form. And the input image data generated based on the aerial photograph of the city area taken at a time different from the time when the aerial photograph was taken, and divided into small areas of a predetermined size and stored in the image data storage means A small area dividing means for dividing the image data being divided into small areas having the same size as the predetermined size, and the input image data divided into small areas by the small area dividing means for each small area. And the image data divided into small areas by the small area dividing means is used to calculate a fractal sequence for each small area. And an image feature distance for each corresponding small area based on the fractal sequence for each small area of the input image data and the fractal sequence for each small area of the image data calculated by the fractal sequence calculating means. Image feature distance calculating means for calculating the image area, and change area extracting means for extracting the change area of the urban area based on the image feature distance calculated by the image feature distance calculating means.

  In addition, in order to solve the above-described problem, the change area extraction method according to the present invention is stored in an image data storage unit and is generated based on an aerial photograph in which a city area of a predetermined form is captured. The stored image data is divided into small areas of a predetermined size, and the input image data generated based on the aerial photograph of the urban area taken at a time different from the time when the aerial photograph was taken is Dividing the input image data divided into small areas, calculating a fractal sequence for each small area, calculating the stored image data divided into small areas, calculating a fractal sequence for each small area, Based on the fractal sequence for each small area of the input image data and the fractal sequence for each small area of the stored image data, an image is generated for each corresponding small area. Calculating a symptom distance, based on the calculated the image feature distance, extracting the city of the changing area.

  According to the present invention, it is possible to quickly and easily analyze the form of a reliable city area comparable to the case where the work is performed manually, pre-survey in city planning, creation of map information, and disaster It can be useful for understanding the affected areas at times.

  The present invention is applied to a form classification apparatus and method for classifying urban areas, and a change area extraction apparatus and method for extracting a change area of an urban area.

  The inventors of the present invention obtain an image feature distance d based on a fractal sequence of an aerial photograph of an arbitrary urban area and a fractal sequence of an aerial photograph of a sample urban area based on the fractal theory, and the image feature The present invention was created by elucidating that the form of an arbitrary urban area can be quantified by the distance d.

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

  First, a fractal sequence calculation method and an image feature distance d calculation method will be described. Hereinafter, the light intensity distribution (aerial photograph pixel value) of the image data is represented by f (x, y).

First, an arbitrary small region of r × r [Pixel] is considered, and the light intensity {f (x _i , y _j ), f (x _i + r, y _j ), f (x _i , y _j + at the four corners thereof is considered. r), f (x _i + r, y _j + r)} (where i, j = 0,1,2, ..., N-1) is f _0, and the light intensity at the four corners F ₀ is subtracted from the virtual volume of the small cube with the vertices as V _r (x _i , y _j ), and the average volume V (r) at this time is obtained based on equation (1) ( FIG. 1).

Here, the small region _{{r 0, r 1, ···} , r M-1} average volume to _{{V (r 0), V} (r 1), ···, V (r M-1)} , Fractal sequence is obtained by {logV (r ₀ ), logV (r ₁ ),..., LogV (r _M−1 )} d.

Next, fractal sequences of two image data having the same resolution are respectively expressed as S = {s ₀ , s ₁ ,..., S _M−1 }, T = {t ₀ , t _{1 ,. 1} }, the image feature distance d is calculated based on the equation (2).

  The image feature distance d represents the similarity between two images. For example, if the two images are the same, d = 0. On the other hand, the value of the image feature distance d increases as the two images differ, and the result is closer to the human visual sense.

<Regarding the shape classification device>
As shown in FIG. 2, the form classification apparatus 1 includes a comparison image data storage unit 10 that stores image data generated based on aerial photographs in which urban areas of a predetermined form are captured, and an arbitrary urban area. A fractal sequence calculation unit that calculates the entire fractal sequence of the input image data generated based on the captured aerial photograph and calculates the entire fractal sequence of the stored image data stored in the comparison image data storage unit 10 11 and an image feature distance calculation unit 12 that calculates the image feature distance d based on the entire fractal sequence of the input image data and the entire fractal sequence of the stored image data calculated by the fractal sequence calculation unit 11. Based on the image feature distance d calculated by the image feature distance calculation unit 12, the shape of an arbitrary city area And a form classification unit 13 that classifies.

  The comparison image data storage unit 10 includes, for example, a first shape comparison image data storage unit 20 in which grayscale grid-like image data generated based on an aerial photograph of an urban area having a grid shape is stored. A second-shaped comparative image data storage unit 21 storing gray-scale concentric image data generated based on an aerial photograph of an urban area having a concentric shape, and an aerial image of an urban area having a radial shape A comparison image data storage unit 22 of a third shape in which grayscale radial image data generated based on a photograph is stored. The comparative image data storage unit 10 has another storage unit that stores grayscale comparative image data generated based on aerial photographs of urban areas other than grids, concentric circles, and radial shapes. May be. Note that the input image data input to the fractal sequence calculation unit 11 is fractal image data.

  The fractal sequence calculation unit 11 calculates the entire fractal sequence of input image data having fractal characteristics. Further, the fractal sequence calculation unit 11 reads out the lattice image data stored in the image data storage unit 20 having the first shape, and calculates the entire fractal sequence of the lattice image data. The fractal sequence calculation unit 11 reads the concentric image data stored in the image data storage unit 21 having the second shape, and calculates the entire fractal sequence of the concentric image data. Further, the fractal sequence calculation unit 11 reads the radial image data stored in the image data storage unit 22 having the third shape, and calculates the entire fractal sequence of the radial image data.

Next, the configuration of the image feature distance calculation unit 12 will be described. The image feature distance calculation unit 12 calculates the image feature distance d ₁ based on the entire fractal sequence of the input image data and the entire fractal sequence of the grid image data calculated by the fractal sequence calculation unit 11. . Further, the image feature distance calculation unit 12 calculates the image feature distance d ₂ based on the entire fractal sequence of the input image data and the entire fractal sequence of the concentric image data calculated by the fractal sequence calculation unit 11. calculate. Further, the image feature distance calculation unit 12 calculates the image feature distance d ₃ based on the entire fractal sequence of the input image data and the entire fractal sequence of the radial image data calculated by the fractal sequence calculation unit 11. To do.

  The image feature distance d represents the similarity between two images. For example, if the two images are the same, d = 0. On the other hand, the value of the image feature distance d increases as the two images differ, and the result is closer to the human visual sense.

  Further, the image feature distance d represents the degree of change between the two images as described above, and changes mainly depending on the three-dimensional arrangement of buildings, roads, etc. An extremely small object has an error (noise component) and has a feature that does not contribute to change.

The form classification unit 13 compares the image feature distance d ₁ obtained by the image feature distance calculation unit 12, the image feature distance d _2, and the image feature distance d _3, and based on the comparison result, the input image data The form of an arbitrary urban area that is the basis of is classified into, for example, a grid, a concentric circle, or a radial pattern.

  Moreover, the structure provided with the registration part 31 which registers input image data into the input image data storage part 30 based on the form classified by the form classification | category part 13 may be sufficient. Further, the input image data storage unit 30 is concentric with the input image data storage unit 30A having the first shape for storing the input image data in a lattice form, for example, according to the result of classification by the form classification unit 13. A second shape input image data storage unit 30B for storing the input image data in the form of a shape, and a third shape input image data storage unit 30C for storing the input image data in the form of a radial shape.

<Specific example>
FIG. 3 shows the result of calculating the image feature distance d from the image data based on the actual aerial photograph. In FIG. 3, the size of one side of the aerial photograph is a square of about 400 [m], and the resolution of the grayscale image data generated from the aerial photograph is 512 × 512 [pixel]. In addition, in order to reduce the influence due to the difference in the photographing conditions, aerial photographs taken around noon in the middle of October were used for all the aerial photographs.

  FIG. 3 is a result of obtaining image feature distances d of a plurality of urban areas based on “Nijo Castle Town, Nakagyo-ku, Kyoto Prefecture” as a reference. The image feature distance d of “Nijojo-cho, Nakagyo-ku, Kyoto” is about 0.01 from “Oyama-cho, Kita-ku, Tokyo” (lattice-shaped urban area). The image feature distance d with “Hichiyoshi-ku” (concentric urban area) is about 0.03, and the image feature distance d with “Minami-Otsuka, Toshima-ku, Tokyo” and “Kita-Otsuka, Toshima-ku, Tokyo” (radial urban area) Is about 0.11. Therefore, it can be seen that the form of the urban area in Nijo Castle Town, Nakagyo Ward, Kyoto Prefecture is close to that of Oyama Town, Kita Ward, Tokyo, which is a grid-like urban area. It can be seen that the image feature distance d is larger in the urban area. This is in good agreement with the human visual sense.

  In addition, as described above, the inventors of the present invention have obtained the image feature distance d based on other forms of urban areas, and as a result, when the urban areas are similar to each other, the image characteristic distance d is obtained. Is approximately 0.01, and in the case of a concentric urban area with a grid-shaped urban area, the image characteristic distance d is approximately 0.03, and in the case of a concentric urban area and a radial urban area, the image characteristic distance is It was elucidated that d was about 0.09, and the image feature distance d was about 0.11 in the case of a grid-like urban area and a radial urban area.

  Therefore, since the difference in the form of the urban area is sufficiently reflected in the image feature distance d, the form classification of the urban area can be performed by using the image feature distance d.

  In this way, the form classification device 1 according to the present invention obtains the image feature distance d between a reference city area and a plurality of different city areas, compares the obtained image feature distances d, and compares the result. When the image feature distances d are substantially equal, they can be classified as urban areas of the same form, and when the image feature distances d are different, they can be classified as urban areas of different forms.

<Regarding the change area extraction device>
The image feature distance d represents the degree of change between the two images as described above. Therefore, according to the present invention, two image data based on aerial photographs are divided into small areas of the same size, and the image feature distance d for each corresponding small area is calculated, thereby accurately extracting the change area. It can be performed. The configuration of the change area extraction device will be described below.

  As shown in FIG. 4, the change area extraction device 2 stores image data (hereinafter referred to as stored image data) generated based on aerial photographs of an arbitrary urban area taken at a certain age. The input image data generated based on the storage unit 40 and the aerial photograph of the same urban area taken at a time different from the time when the aerial photograph was taken is divided into small areas of a predetermined size, and the image A small area dividing unit 41 that divides the stored image data stored in the data storage unit 40 into small areas having a size equivalent to a predetermined size, and the input image data divided into small areas by the small area dividing unit 41 Fractal sequence for calculating a fractal sequence for each region, and storing image data divided into small regions by the small region dividing unit 41 to calculate a fractal sequence for each small region. Based on the fractal sequence for each small area of the input image data and the fractal sequence for each small area of the stored image data calculated by the sequence calculation section 42 and the fractal sequence calculation section 42, an image is generated for each corresponding small area. An image feature distance calculation unit 43 that calculates a feature distance and a change region extraction unit 44 that extracts a change region of an urban area based on the image feature distance calculated by the image feature distance calculation unit 43 are provided. Note that the input image data input to the small area dividing unit 41 is fractal image data.

  The image data storage unit 40 stores image data generated based on aerial photographs of urban areas with different shooting ages.

  The small area dividing unit 41 divides input image data having fractal properties into small areas of a predetermined size. The small area dividing unit 41 reads the stored image data stored in the image data storage unit 40 and divides the stored image data into small areas of a predetermined size.

  The fractal sequence calculation unit 42 calculates a fractal sequence for each small area of the input image data divided into small areas. Further, the fractal sequence calculation unit 42 calculates a fractal sequence for each small area from the stored image data divided into the small areas.

  Based on the fractal sequence for each small area of the input image data and the fractal sequence for each small area of the stored image data calculated by the fractal sequence calculation section 42, the image feature distance calculation section 43 corresponds to each small area. Each image feature distance is calculated.

  Here, a series of operations by the fractal sequence verification unit 42 and the image feature distance calculation unit 43 will be described.

As shown in FIG. 5A, the fractal sequence calculation unit 42, as shown in FIG. 5A, has a small area {p _n } of size u × u [pixel] (where n = 1, 2,... L ² ) Divide into ² ) and find each fractal sequence. In addition, the fractal sequence calculation unit 42, as shown in FIG. 5B, outputs other image data to a small region {q _n } having a size u × u [pixel] (where n = 1, 2,... • Divide into L ² ) and find each fractal sequence.

Then, the image feature distance calculator 43, a fractal sequence of the small region {p _1} of one image data, the fractal sequence of small regions subregions other image data corresponding to {p _1} {q _1} The image feature distance d ₁ is calculated on the basis of the image feature distance, and the image feature distance d _n (where n = 1, 2,... L ² ) is sequentially calculated on the basis of the fractal sequence of the small area at the corresponding position. .

  Based on the image feature distance d for each small region calculated by the image feature distance calculation unit 43, the change region extraction unit 44 extracts and extracts only the small regions having the image feature distance d equal to or greater than a certain value. The light intensity distribution of the small area is calculated. The change area extraction unit 44 extracts a small area whose calculated value is a certain value or more as a change area.

  Moreover, the structure provided with the size setting part 45 which sets the size at the time of dividing | segmenting image data into a small area by the small area dividing part 41 may be sufficient. In addition, when the size set by the size setting unit 45 is small, there is a merit that a fractal sequence can be calculated more precisely. However, there is a demerit that it takes time for the calculation. Although less time is required, the value of the fractal sequence becomes rough. Therefore, the set size is appropriately determined depending on how much precision is required.

<Extraction of change area>
Here, as shown in FIG. 6, a building of size [w _p × h _p ] inside the small area {p _L } of the past image data is in the small area {q _L } of the current image data. Is considered to be a building of size [w _q × h _q ].

  In general, since the light intensity distribution on the ground and the upper part of the building is quite different, for example, when the light intensity distribution is obtained along the x direction and the y direction by a predetermined amount from the center of the small area, FIG. 6 (A) and FIG. As in (B), the light intensity distribution changes at the boundary between the ground and the upper part of the building.

Further, in FIG. 6 (C), the small regions {q _L} subregion {p _L} and the current image data of the previous image data, the comparison of the light intensity distribution obtained by scanning along the x-direction Each example is shown.

Therefore, a changing small region can be extracted by examining a range (w _p ≦ x ≦ w _q ) where the light intensity distribution is different.

  In this way, the change area extraction apparatus 2 according to the present invention can extract changes in small areas that are difficult to visually identify in the image data. For example, when some disaster occurs in a certain urban area, Rapidly detecting changes in urban areas by calculating image feature distance d and light intensity distribution for each small area based on image data from aerial photographs before disaster and image data from aerial photographs after disaster Can be used to identify disaster areas.

  Further, if the present invention is applied, for example, based on image data based on past aerial photographs at a certain location and image data based on current aerial photographs at the same location, image feature distance d and light intensity distribution for each small area. Can be used to create a map by extracting a small area that has changed, and updating only the image data corresponding to the extracted small area.

  The present invention is not limited to the above-described embodiments described with reference to the drawings, and various modifications, substitutions or equivalents thereof can be made without departing from the scope and spirit of the appended claims. Of course, it can be done.

It is a figure which shows the virtual volume of a small cube. It is a block diagram which shows the structure of the form classification device to which this invention is applied. It is a figure which shows the result of having calculated the image feature distance d from the image data based on an actual aerial photograph. It is a block diagram which shows the structure of the change area extraction apparatus to which this invention is applied. It is a figure where it uses for description of the division | segmentation of the image performed by the fractal sequence calculating part. It is a figure where it uses for description which extracts a change area | region.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 Form classification apparatus, 2 Change area extraction apparatus, 10 Comparison image data storage part, 11 Fractal sequence calculation part, 12 Image feature distance calculation part, 13 Form classification part, 20 Comparison image data storage part of 1st shape, 21st 2 shape comparison image data storage unit, 22 3rd shape comparison image data storage unit, 30 input image data storage unit, 31 registration unit, 40 image data storage unit, 41 small region division unit, 42 fractal sequence calculation unit 43 Image feature distance calculation unit 44 Change region extraction unit 45 Size setting unit

Claims (5)

Image data storage means for storing image data generated based on aerial photographs in which urban areas of a predetermined form are captured;
Fractal sequence calculation means for calculating a fractal sequence of input image data generated based on an aerial photograph in which an arbitrary urban area is photographed, and calculating a fractal sequence of the image data stored in the image data storage means; ,
Image feature distance calculation means for calculating an image feature distance based on the fractal sequence of the input image data calculated by the fractal sequence calculation means and the fractal sequence of the image data;
A form classification device, comprising: a form classification means for classifying the form of any urban area based on the image feature distance calculated by the image feature distance calculation means. Calculate the fractal sequence of the input image data generated based on the aerial photograph taken of any urban area,
Stored in the image data storage unit, calculates a fractal sequence of the stored image data generated based on the aerial photograph of the urban area of a predetermined form,
Calculating an image feature distance based on the fractal sequence of the input image data and the fractal sequence of the stored image data;
A form classification method characterized in that the form of the arbitrary urban area is classified based on the image feature distance. Image data storage means for storing image data generated based on aerial photographs taken of urban areas of a predetermined form;
The input image data generated based on the aerial photograph of the urban area taken at a time different from the time when the aerial photograph was taken is divided into small areas of a predetermined size and stored in the image data storage means. A small area dividing means for dividing the image data being divided into small areas having a size equivalent to the predetermined size;
The input image data divided into small areas by the small area dividing means calculates a fractal sequence for each small area, and the image data divided into small areas by the small area dividing means is divided into fractals for each small area. A fractal sequence calculation means for calculating a sequence;
Based on the fractal sequence for each small area of the input image data and the fractal sequence for each small area of the image data calculated by the fractal sequence calculation means, the image feature distance is calculated for each corresponding small area. Image feature distance calculation means;
A change area extraction device comprising: a change area extraction means for extracting the change area of the urban area based on the image feature distance calculated by the image feature distance calculation means.   4. The change area extraction apparatus according to claim 3, wherein the change extraction means extracts the change area by scanning a light intensity distribution of the image data. The stored image data that is stored in the image data storage unit and is generated based on the aerial photograph in which the urban area of a predetermined form is captured is divided into small areas of a predetermined size,
The input image data generated based on the aerial photograph of the city area taken at a time different from the time when the aerial photograph was taken is divided into small areas of a predetermined size,
Calculate the input image data divided into small areas, fractal sequences for each small area,
Calculate the fractal sequence for each small area from the stored image data divided into small areas,
Based on the fractal sequence for each small area of the input image data and the fractal sequence for each small area of the stored image data, an image feature distance is calculated for each corresponding small area,
A change area extraction method characterized by extracting a change area of the city based on the calculated image feature distance.

Images