JP2013088999A - Building extraction device, method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To quickly extract the shape of a building from three-dimensional point group data.SOLUTION: The building extraction device: generates an irregular triangular net from the three-dimensional point group data (S12); calculates a normal vector on each triangular net (S13); converts the normal vectors to grids (S14); groups those in the same direction within a predetermined direction range (S15); normalizes the dimension (S15); extracts a contour line of the building from a surface that comprises the grids having the normal vectors parallel to the ground within a predetermined error range (S17 and S18); applies an elevation value to the contour line (S19); and three-dimensionally vectorizes the shape determined by the contour line (S20).

Description

  The present invention relates to an apparatus, method, and program for extracting a shape of a building from three-dimensional point cloud data.

  Aviation laser surveying is a technology (see Patent Document 1) that mounts a laser rangefinder on an airplane and measures the three-dimensional shape of the ground plane including the features by irradiating the ground with laser light. Efficiently collect 3D point data.

  Patent Document 2 describes a technique for generating a three-dimensional shape model of an object from three-dimensional point cloud data obtained by scanning laser light.

  In Patent Document 3, a rectangular shape is generated from three-dimensional point cloud data collected at high density by aviation laser surveying. Between the rectangular shapes, the continuity between adjacent polygons, the normal direction, the distance, etc. A technique for recognizing edges and surfaces and generating three-dimensional shape data is described.

JP 2004-170429 A Japanese Patent Laid-Open No. 2004-272459 JP 2009-014643 A

  In the technique described in Patent Document 1, since the surface and edge are recognized in units of randomly located points, which are three-dimensional point cloud data collected with high density by aviation laser surveying, the burden on computer processing is large. Requires enormous computing power.

  It is an object of the present invention to provide a building extraction apparatus, method, and program capable of extracting a building shape with a necessary accuracy from three-dimensional point cloud data collected at high density by aviation laser surveying by simple computer processing. .

  A building extraction apparatus according to the present invention is an apparatus that extracts the shape of a building from 3D point cloud data obtained by three-dimensionally measuring the building, and generates an irregular triangular network from the 3D point cloud data. Triangular network generating means, normal vector calculating means for calculating a normal vector for each triangular network generated by the triangular network generating means, and grid forming means for gridting the normal vectors into a grid of predetermined distance intervals According to the normal vector gridded by the gridting means, grouping means for grouping the same direction within a predetermined direction range, and the normal vector grouped by the grouping means on the ground On the other hand, contour extracting means for extracting the contour of the building from a plane composed of grids having normal vectors parallel within a predetermined error range, and assigning an elevation value to the contour. And elevation applying means for, characterized by comprising a three-dimensional vectorization means for three-dimensional vector of the shape determined by the contour line.

  The building extraction method according to the present invention is a method of extracting the shape of a building from 3D point cloud data obtained by three-dimensionally measuring the building, and generates an irregular triangular network from the 3D point cloud data. A triangular network generation step, a normal vector calculation step for calculating a normal vector for each triangular network generated in the triangular network generation step, and a grid forming step for gridting the normal vector into a grid of predetermined distance intervals; A grouping step for grouping objects in the same direction within a predetermined direction range according to the normal vector gridded in the gridting step, and a normal vector grouped in the grouping step on the ground An outline extraction process for extracting the outline of the building from a plane composed of grids having normal vectors parallel to each other within a predetermined error range. And-up, characterized by comprising the altitude imparting step of imparting altitude value on the contour line, and a three-dimensional vectorization step of three-dimensional vectors of the shape determined by the contour line.

  A building extraction program according to the present invention is a program for extracting the shape of a building from three-dimensional point cloud data obtained by three-dimensional measurement of the building, and the computer extracts an irregular triangular network from the three-dimensional point cloud data. A triangle network generation function for generating a normal vector, a normal vector calculation function for calculating a normal vector for each triangle network generated by the triangle network generation function, and a grid that grids the normal vectors into a grid at predetermined distance intervals A grouping function, a grouping function for grouping objects in the same direction within a predetermined direction range, and a normal vector grouped by the grouping function. An outline extractor for extracting an outline of a building from a plane composed of a grid having normal vectors parallel to the ground within a predetermined error range When, characterized in that to achieve and elevation imparting function of imparting the altitude value on the contour line, and a three-dimensional vector function of three-dimensional vectors of the shape determined by the contour line.

  According to the present invention, the normal vector is gridded, it is determined whether or not the building outline is configured with a horizontal or nearly horizontal direction component on the ground, and the building outline is extracted. Can be determined. Thereby, the three-dimensional shape data indicating the building shape can be obtained in a short time.

It is a schematic block diagram of one Example of this invention. It is a main flow of the computer program of a present Example. It is a detailed flowchart of a schematic shape extraction program. It is a detailed flowchart of a detailed shape extraction program. It is a detailed flowchart of a display and shaping program.

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

  FIG. 1 shows a schematic block diagram of an embodiment of the present invention, and FIGS. 2 to 5 show flowcharts of characteristic operations of this embodiment.

  The aviation laser surveying device 10 and the digital camera 12 are mounted on an airplane, a laser pulse is irradiated from the aviation laser surveying device 10 to the ground, and the ground shape including the feature and the ground is three-dimensionally measured by the reflected light. The three-dimensional point data obtained by the laser aerial survey equipment includes reflected light from trees in addition to the reflected light from the outer wall including the roof of the ground building and the reflected light from the ground, and further, the laser from the airplane to the ground. Since the light irradiation direction also depends on the attitude of the airplane, the obtained three-dimensional point data is usually random with respect to the ground coordinates. In addition, the survey target area of the aviation laser surveying apparatus is photographed so as to partially overlap with the digital camera 12 so that it can be combined with a small number of images over a wide area.

  The computer 14 extracts the shape of the ground building using the three-dimensional point cloud data obtained by the aviation laser surveying instrument 10 and the image data captured by the camera 12, and outputs vector data indicating the shape. The computer 14 includes a CPU 20, a ROM 22, a RAM 24, a hard disk (HDD) 26 as auxiliary storage, a keyboard 28, a mouse 30, an external interface 32, and a display device 34. The computer 14 displays various menus and images on the display device 34.

  The HDD 26 stores an operating system and a building extraction program for this embodiment. Although details will be described later, on the CPU 20, a schematic shape extraction program 20a, a detailed shape extraction program 20b, and a display / shaping program 20c operate.

  With reference to FIG. 2 to FIG. 5, processing on the computer 14 for extracting a schematic shape from three-dimensional point data obtained by the laser aerial survey apparatus and generating three-dimensional detailed shape data from the obtained schematic shape data will be described. To do.

  First, the three-dimensional point data obtained by the aviation laser surveying instrument 10 and the image data photographed by the camera 12 are taken into the computer 14 (S1). That is, a storage medium (not shown) or a file server for storing these data is connected to the external connection interface 34 of the computer 14 and transferred to the hard disk (HDD) 26. The external connection interface 34 is, for example, a USB (Universal Serial Bus).

  First, the approximate shape extraction program 20a operating on the CPU 20 extracts the approximate shape of the ground building from the three-dimensional point cloud data (S2), and the detailed shape extraction program 20b determines the result of the approximate shape extraction program 20a and the captured image data. The detailed three-dimensional shape is extracted or determined with reference to (S3). The display / shaping program 20c assigns the three components of the latitude, longitude, and elevation of the normalized normal vector obtained by the approximate shape extraction program 20a to the R value, G value, and B value of the color image, respectively. The normal vector is converted into a color image and displayed, and an image showing a three-dimensional detailed shape obtained by the detailed shape extraction program 20b is superimposed and displayed (S4). The operator corrects the three-dimensional detailed shape data with reference to the display image, and the display / shaping program 20c stores the correction result in the HDD 26 (S4).

  FIG. 3 shows a detailed flowchart of the schematic shape extraction process (S2). With reference to FIG. 3, the function of the schematic shape extraction program 20a will be described in detail.

  The schematic shape extraction program 20a first removes isolated noise from the three-dimensional point cloud data (S11). This is a process of removing points with different three-dimensional coordinate values that are unnatural with respect to surrounding point data. Since it can exist randomly on the latitude and longitude planes, generally, when the altitude value is isolated and far away from an adjacent point, it is determined as isolated noise and removed.

  The outline shape extraction program 20a generates an irregular triangular network (TIN) having point vertices for the point cloud data from which isolated noise is removed, and stores it in the HDD 26 as triangular network (TIN) data. (S12). Thereby, a surface can be expressed. TIN and its generation method itself are well known in the Geographic Information System (GIS).

  The schematic shape extraction program 20a calculates a normal vector using each vertex coordinate for each triangular plane represented by the TIN data (S13). The calculated normal vector is randomly positioned with respect to the ground plane, and it is inevitable that it is dense. Therefore, in order to facilitate the subsequent calculation processing, the approximate shape extraction program 20a uses the calculated normal vector at the required spatial density in the latitude and longitude directions, for example, at a constant distance interval of 50 cm or 1 m. A grid is formed (S14). In other words, the outline shape extraction program 20a performs vector synthesis on each grid at a constant interval by combining a normal vector of TIN within a fixed distance, or a TIN including the grid and a TIN normal vector adjacent thereto. Generate gridded normal vectors.

  As the number of random normal vectors used for vector synthesis increases, the size of the resultant normal vector, that is, the scalar value increases. Since the gridded normal vector should consider not the size but the direction, the outline shape extraction program 20a normalizes the gridded normal vector to a certain size (S15). The schematic shape extraction program 20a stores the obtained normalized normal vector in the HDD 26 for display / shaping processing (S16). Normalization facilitates comparison of the surface direction of each part.

  In order to recognize a surface larger than the grid unit as a single surface, the approximate shape extraction program 20a compares the normalized normal vectors with each other, and if the direction difference is within a certain tolerance, It is assumed that a flat surface is expressed, and these are grouped (S17). Thereby, for example, the roof of a building can be expressed or identified by a single or a small number of squares in a grid unit, and it becomes easy to separate and extract the outer wall where the building stands.

  The outline shape extraction program 20a extracts the outline of the building from each surface formed by grouping the normalized normal vectors (S18). Specifically, it can be considered that a surface whose normalized normal vector is horizontal or nearly horizontal to the ground, that is, a surface having a small height direction (z) component, represents a standing wall. Since it can be said that the outline of the building itself, the surface showing such a normalization normal vector is defined as the outer wall surface, and the outer wall surface is thinned to obtain the outer wall contour line.

  In addition, it can be considered that a plane whose normalized normal vector is at a certain angle or more with respect to the ground represents a flat or curved roof. You may determine with the outline of the roof surface or top surface of a building, and you may add to the outline determined with the z direction component.

  The schematic shape extraction program 20a further assigns an elevation value to the extracted contour line (vertex thereof) (S19). Specifically, the altitude value of the edge line on the roof side among the edge lines constituting the wall surface before thinning is given to each vertex of the extracted contour line. For each vertex of the extracted contour line, the elevation value of the adjacent vertex of the roof net or the triangular mesh data on the top surface side of the contour line may be extrapolated to determine the elevation value of the end point.

  The schematic shape extraction program 20a refers to the contour line data with the elevation value obtained in step S19, and determines the shape inside the contour line, that is, the three-dimensional shape of the roof portion of the building (S20). Specifically, the triangular mesh generated in step S12 is gridded with the same size and position as the normal vector grid (S17), and is closed by the contour line with the elevation value obtained in step S19. The grids are aggregated and the shape is converted into a three-dimensional vector. Thereby, shapes, such as a land roof, a curved roof, a roof of Yada flow, can be determined, and the shape can be expressed in vector.

  Finally, the outline shape extraction program 20a summarizes the contour line data with the elevation value generated in step S19 and the three-dimensional shape data obtained in step S20 in units of buildings, and saves them in the HDD 26 as outline shape data (S21). .

  FIG. 4 shows a detailed operation flowchart of the detailed shape extraction program 20b. With reference to FIG. 4, the operation of step S3 by the detailed shape extraction program 20b will be described in detail.

  The detailed shape extraction program 20b reads the captured image data including one or more buildings specified by the approximate shape extraction program 20a from the HDD 26, and determines the image coordinates of each captured image (S31). Although the image coordinates may be determined only by the captured image by the aerial photogrammetry technique, the captured image is superimposed on the approximate shape determined by the approximate shape extraction program 20a and displayed on the display device 34, thereby performing laser surveying. It is also possible to determine image coordinates using the result. For example, the coordinates (image coordinates) on the photographed image are determined by the operator using the mouse 30 or the like to instruct and input a known point at the spatial position of the approximate shape determined by the approximate shape extraction program 20a, for example. . Of course, by determining the image coordinates of a plurality of points on the captured image, it is possible to calculate the image coordinates of the positions not specified.

  The approximate shape extraction program 20a detects an edge as an adjacent line from the displayed image by a known edge detection technique (S32), and superimposes the extracted edge and the approximate shape determined by the approximate shape extraction program 20a on the captured image. To display. This display allows the operator to visually recognize the validity of the approximate shape and the finer shape by referring to the approximate shape on the display image. You can specify the shape of the details.

  On the screen of the display device 34, the operator uses the mouse 30 to determine an adjacent line serving as a base of a detailed contour that complements the schematic shape (S33). The adjacent line is a line that forms the outer shape of the building when viewed in individual captured images. Specifically, the operator identifies an edge adjacent to the approximate shape with reference to the approximate shape determined by the approximate shape extraction program 20a, and selects a line segment constituting the outer shape of the building from the edge by using the mouse 30. Specify sequentially. Furthermore, the operator can additionally specify a line segment that becomes an adjacent line from the edge of the location that was overlooked in the outline shape extraction.

  The operator applies steps S31 to S33 to a plurality of photographed images having different photographing directions, and extracts adjacent lines from the photographed images having different photographing directions. Then, the detailed shape extraction program 20b determines the contour line by applying the forward intersection method to the adjacent line thus extracted (S33). The determined contour line is additional shape data that refines the general shape, and forms detailed shape data as a whole. The detailed shape extraction program 20b adds the designated or corrected shape to the rough shape data and stores it in the HDD 26 (S34).

  The detailed shape extraction program 20b ends when extracting the detailed shapes of all the features including buildings that are of interest. In this way, the normal vector is gridded, whether or not the building outline is composed of horizontal or near horizontal components on the ground, and the building outline is extracted to efficiently determine the building outline. it can. Also, three-dimensional shape data indicating the building shape can be obtained in a short time.

  FIG. 5 shows a detailed operation flowchart of the display / shaping program 20c. With reference to FIG. 5, the operation of step S4 by the display / shaping program 20c will be described in detail.

  The normalized normal vector stored in the HDD 26 in step S16 includes three-direction components of latitude, longitude, and altitude. The display / shaping program 20c assigns the three components of the latitude, longitude, and elevation of the normalized normal vector obtained by the approximate shape extraction program 20a to the R value, G value, and B value of the color image, respectively. The normal vector (the spatial distribution thereof) is converted into an RGB image (S41).

  The display / shaping program 20c also converts the three-dimensional detailed shape extracted by the detailed shape extraction program 20b into an RGB image having the same coordinate scale and position as the RGB image of the normalized normal vector in step S41 (S42), Both are combined and displayed on the display device 34 (S43).

  The operator observes the image displayed on the display device 34 and manually corrects the detailed shape data with reference to the RGB image of the normalized normal vector (S44). If there is a correction, the display / shaping program 20c stores the corrected detailed shape data separately from the existing data or overwrites the existing data (S45). In the former case, the correction can be canceled afterwards. The detailed shape image is updated with the corrected detailed shape (S42), and is synthesized and displayed with the RGB image of the normalized normal vector (S43).

  The shaping process here also includes a process of correcting a shape expressed by a plurality of broken lines to an expression with fewer lines.

  When the correction is lost (S44), the operator ends the display / shaping program 20c.

  With the above processing, according to the present embodiment, since the normal vector is used for the outline outline determination, a steep rising portion can be found effectively. In addition, since the two-step process of determining the detailed shape by applying the approximate shape determined in this way to the captured image is applied, a fine outer shape can be determined in a short time, and overlooking can be reduced.

  In the above embodiment, the display / shaping process is applied to the detailed shape obtained through the detailed shape extraction process. However, if the approximate shape obtained by the approximate shape process has sufficient accuracy, the detailed shape is omitted. Obviously we can do it. In this case, display / shaping processing is executed only when shaping is necessary, and at that time, in step S42, the three-dimensional schematic shape is converted into an RGB image.

  Although the present invention can be realized mainly as a computer program that runs on the computer 14, it goes without saying that part or all of the elements can be realized by dedicated hardware.

  Although the invention has been described with reference to specific illustrative embodiments, various modifications and alterations may be made to the above-described embodiments without departing from the scope of the invention as defined in the claims. This is obvious to an engineer in the field to which the present invention belongs, and such changes and modifications are also included in the technical scope of the present invention.

10: Aviation laser surveying device 12: Digital camera 14: Computer 20: CPU
20a: Outline shape extraction program 20b: Detailed shape extraction program 20c: Display / shaping program 22: ROM
24: RAM
26: Hard disk (HDD)
28: Keyboard 30: Mouse 32: External interface 34: Display device

Claims (3)

An apparatus for extracting the shape of a building from 3D point cloud data obtained by measuring the building in 3D,
A triangular network generating means for generating an irregular triangular network from the three-dimensional point cloud data;
Normal vector calculating means for calculating a normal vector for each triangular network generated by the triangular network generating means;
Gridding means that grids the normal vectors into a grid of predetermined distance intervals;
Grouping means for grouping those in the same direction within a predetermined direction range according to the normal vector gridded by the gridding means;
Contour extraction that extracts the contour of the building from the surface composed of grids having normal vectors parallel to the ground within a predetermined error range among the normal vectors grouped by the grouping means. Means,
An altitude imparting means for imparting an altitude value to the contour line;
A building extraction apparatus comprising: three-dimensional vectorization means for converting a shape determined by the contour line into a three-dimensional vector. A method of extracting the shape of a building from 3D point cloud data obtained by three-dimensional measurement of the building,
A triangular network generation step for generating an irregular triangular network from the three-dimensional point cloud data;
A normal vector calculating step for calculating a normal vector for each triangular net generated in the triangular net generating step;
Gridding step of gridting the normal vector into a grid of a predetermined distance interval;
A grouping step for grouping those in the same direction within a predetermined direction range according to the normal vector gridded in the gridding step;
Contour extraction that extracts the contour of the building from the surface composed of grids having normal vectors parallel to the ground within a predetermined error range among the normal vectors grouped in the grouping step. Steps,
An elevation giving step for assigning an elevation value to the contour line;
A building extraction method comprising: a three-dimensional vectorization step of converting a shape determined by the contour line into a three-dimensional vector. A program for extracting the shape of a building from 3D point cloud data obtained by measuring the building in 3D,
A triangular network generation function for generating an irregular triangular network from the three-dimensional point cloud data;
A normal vector calculation function for calculating a normal vector for each triangular network generated by the triangular network generation function;
A grid function that grids the normal vector into a grid at a predetermined distance interval;
A grouping function for grouping the same direction within a predetermined direction range according to the normal vector gridded by the grid function;
Contour extraction that extracts the contour of the building from the plane composed of grids having normal vectors parallel to the ground within a predetermined error range among the normal vectors grouped by the grouping function Function and
An altitude giving function for giving an altitude value to the contour line;
A building extraction program characterized by realizing a three-dimensional vectorization function for converting a shape determined by the contour line into a three-dimensional vector.

Images