KR101392804B1 - Indoor 3d model extraction method based on point clouds from terrestrial lidar and recording medium thereof - Google Patents

Abstract

The present invention relates to a point cloud-based indoor space three-dimensional model extraction method acquired in a terrestrial environment, wherein the indoor space is scanned using the terrestrial environment; A step 120 of merging the point clouds scanned in various places of the indoor space to generate a point cloud for one indoor space; 130) of segmenting the point cloud of the merged indoor space; 140. The method of claim 140, further comprising: determining a plane using the divided point cloud; A step 150 of extracting a boundary line for the indoor space using the determined surface; And a step 160 of generating a three-dimensional model of the indoor space by using the remaining point clouds for the boundary and the remaining area where the boundary is not extracted. Dimensional model extraction method of the indoor space.
According to the present invention, a plane constituting an indoor space is automatically extracted from a three-dimensional point cloud acquired through laser scanning technology, and a three-dimensional model is extracted using a three-dimensional point cloud for the remaining indoor space Thus, the time and cost can be drastically reduced as compared with the conventional method. In addition, by expressing the three-dimensional indoor space using the automatically extracted plane and the unused three-dimensional point cloud for plane extraction, the detailed representation can be achieved without loss of information compared with the relatively small data size as compared with the existing method . In addition, if a technique of generating a three-dimensional indoor space model automatically or semiautomatically with a point cloud obtained by using the ground lidar is applied, it can be efficiently used for planning, designing and maintenance of an existing indoor space But it has an effect of contributing to the construction of an indoor space database in the field of urban planning, three-dimensional geographic information system (GIS) and building information modeling (BIS).

Description

BACKGROUND OF THE INVENTION Field of the Invention [0001] The present invention relates to a three-dimensional model extraction method for indoor space based on a point cloud obtained from a terrestrial Lada, and a recording medium on which a program for implementing the method is extracted.

The present invention relates to a point cloud-based indoor space three-dimensional model extraction method and a recording medium on which a program for implementing the same is acquired, which is obtained from a terrestrial Lada, and is based on a 3D point cloud obtained using a terrestrial Lada The present invention relates to a method for constructing a three-dimensional space model of an indoor space and expressing the indoor environment of the room.

Land Raida acquires 3D data with the same principle as air Raida survey. It can acquire high-density point cloud while saving time and manpower. Therefore, it is possible to construct indoor 3D model in geographic information system, And it is used for building deformation measurement, cultural property management, and so on.

The ground-based LIDA used in the indoor modeling acquires three-dimensional data by the time difference method, and is classified into the TOF (Time of Flight) method and the phase difference method in detail. Long-range laser scanners typically acquire three-dimensional data in TOF and can measure distances of several hundred meters or more, but they are relatively inaccurate and take a relatively long time to measure. On the other hand, the high-density laser scanner is based on the phase difference and has a shorter measurement distance than the TOF method, but it can acquire a high-density point cloud, and thus is used for three-dimensional indoor modeling of the inside of a building and facilities.

In the past, indoor 3D modeling focused on the acquisition or restoration of 2D information using photogrammetry technology or CAD drawings. However, existing 2D map information has restrictions on expressing spatial information about high - rise buildings and complex facilities. Therefore, there is a growing need for a technique for constructing an indoor three - dimensional model and expressing indoor space. However, in the field of spatial information, the 3D model building technology is predominant in the construction of the terrain model, the outdoor model of the civil engineering and the building structure, and the research on the development and application of the indoor 3D model building technology is insufficient.

In order to construct an indoor three-dimensional model, a point cloud obtained using a ground lidar can be used. However, most of the work is carried out by hand, so the work results may be different according to the skill of the worker, and the work time and cost are increased. Therefore, it is necessary to standardize the three-dimensional modeling process.

In addition, apart from building a 3D model, a well-designed interior space representation method is required for precise visualization of accurate indoor space. However, in order to solve this problem, it is necessary to consider the following three points: (1) indoor environment expression using a three-dimensional model constructed by the operator, (2) interior space expression using a point cloud as a raw data, Express space. The first method is not only time-consuming and costly, but is also a subjective result, and it loses all information about the actual shape. The second method can preserve original data, but it has limitations on the processing and visualization of 3D information due to its large capacity. The third method can take advantage of the two methods, but has the limitation that complex space and simple space can be simplified to the same level because they are expressed using one simplifying variable. Therefore, a method of expressing indoor space is needed to solve such limitations.

DISCLOSURE OF THE INVENTION The present invention has been made to solve the above problems of the conventional art, and it is an object of the present invention to provide a three-dimensional point cloud which is obtained by dividing a three-dimensional point cloud acquired for an indoor space by a segmentation method, A three-dimensional model extraction method based on a point cloud obtained from the ground level, which can extract a three-dimensional model of an indoor space efficiently by utilizing a three-dimensional point cloud for a space, and a method for recording a program And it is an object of the present invention to provide a medium.

In order to attain the above object, a point cloud-based indoor space three-dimensional model extraction method acquired in a ground-level domain of the present invention is a point cloud-based indoor space three-dimensional model extraction method, A step 110 of scanning the indoor space using the lidar; A step 120 of merging the point clouds scanned in various places of the indoor space to generate a point cloud for one indoor space; 130) of segmenting the point cloud of the merged indoor space; 140. The method of claim 140, further comprising: determining a plane using the divided point cloud; A step 150 of extracting a boundary line for the indoor space using the determined surface; And a step 160 of generating a three-dimensional model of the indoor space using the remaining point clouds for the boundary and the remaining area in which the boundary is not extracted.

In operation 120, the absolute coordinates of the point cloud for one indoor space may be registered after merging.

In operation 130, the entire point cloud of the indoor space is classified into point data belonging to the same plane by using the point cloud segmentation method to remove invalid segments. In operation 150, Segmentation points are binarized and the boundary of interior space can be extracted using Vectorization.

Also, a refinement grid may be used to remove the invalid segment, and an outline trace grid may be used for binary imaging.

In operation 160, the surface is extracted using the finally created boundary line, and the 3D point cloud for the remaining indoor space is directly used, or the 3D mesh generated from the 3D point cloud is used for 3D rendering software Dimensional space, so that a three-dimensional model of the indoor space can be created.

According to another aspect of the present invention, there is provided a recording medium recording a program for causing a computer to execute a point cloud-based indoor space three-dimensional model extraction method acquired in a terrestrial environment.

According to the point cloud-based indoor space three-dimensional model extraction method and the recording medium on which the program for implementing the method is extracted, the indoor space is obtained from the 3D point clouds obtained through the laser scanning technology, It is possible to drastically reduce the time and cost in comparison with the conventional method by automatically extracting the planes to be constructed and extracting the planes and extracting the three-dimensional model using the three-dimensional point cloud for the remaining indoor space, It is possible to alleviate the difference in the result depending on the characteristics of the product.

In addition, by expressing the three-dimensional indoor space using the automatically extracted plane and the unused three-dimensional point cloud for plane extraction, the detailed representation can be achieved without loss of information compared with the relatively small data size as compared with the existing method .

In addition, if a technique of generating a three-dimensional indoor space model automatically or semiautomatically with a point cloud obtained by using the ground lidar is applied, it can be efficiently used for planning, designing and maintenance of an existing indoor space However, it has an effect of contributing to the construction of an indoor space database in the field of urban planning, three-dimensional geographic information system (GIS) and building information modeling (BIM).

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a flowchart of a point cloud-based indoor space three-dimensional model extraction method acquired from a ground level according to an embodiment of the present invention; FIG.
Fig. 2 is a point cloud of a building scanned using a ground level scanner.
FIG. 3A is an exemplary view showing a point cloud of a building divided; FIG.
FIG. 3B is an exemplary view showing a boundary line of an indoor space extracted using a plane determined by a divided point cloud. FIG.
FIG. 4A is an exemplary view showing a three-dimensional mesh data and a boundary line extracted; FIG.
FIG. 4B is an illustration showing a state in which mesh data is formed using point clouds that are not used for boundary line extraction of a three-dimensional model; FIG.

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

FIG. 1 is a flowchart of a method of extracting a point cloud-based indoor space three-dimensional model acquired in a terrestrial lathe according to an embodiment of the present invention.

Referring to FIG. 1, first, an indoor space is scanned using a ground lidar (S110). The point cloud obtained from the terrestrial Lada is generally composed of points composed of relative coordinates to the indoor space of (x, y, z) and reflection intensity, and is stored in a file in text or binary format. In order to obtain accurate data of the indoor space, the ground lidar is installed in several places and the unscanned area is scanned so that the obstruction area is minimized.

Next, the point clouds scanned at various places in the indoor space are merged (S120). Specifically, a point cloud for one indoor space can be created by merging the point clouds after removing the noise included in the point cloud scanned and stored in various places in the indoor space using the ground lidar. In addition, it is possible to remove the noise included in the merged point clouds after merging the stored point clouds with various points in the indoor space, respectively.

The method may further include a process of merging the point clouds from which the noise has been removed, registering the absolute coordinates, or merging the point cloud, removing noise, and registering absolute coordinates of the point cloud with respect to one indoor space . Specifically, since a plurality of point clouds have relative coordinates according to the position of the ground lidar, a registration process is required, and it is necessary to register the absolute coordinates of the local or global coordinates. These registration and absolute coordinate registration methods are classified into direct method and indirect method. Of these, indirect method is used. Indirect techniques require a separate measurement process to assign absolute coordinates to the point cloud data and use natural artifacts or artifacts to match the point cloud data. Three matching targets are required between two point cloud data for matching, but three or more targets are used to minimize errors.

Next, segmentation is performed using the point cloud for the indoor space merged into one (S130). Representative partitioning methods include Region growing method, Random sample consensus (RANSAC) method, Parameter-based method, and 3D BSP (Binary Space Partitioning) tree.

The segmentation of the point cloud using the area expansion method determines the initial plane using the adjacent point data from the seed point. If the new point is within the statistical threshold from the initial plane, the point is divided into the group to which the seed point belongs . After the region expansion centered on one seed point is completed, the planar equations are calculated through least square adjustment using points belonging to the same group. Then, the new point is set as another seed point, and region expansion is continuously performed. This process is performed for all the points to be incorporated into one of the planes constituting the indoor space.

The partitioning technique based on the RANSAC methodology continuously selects a random set of points, uses them to construct a plane, and then separates points belonging to this plane from points not belonging to it. Since the constructed plane is determined by an initial random set, the methodology is continuously performed to finally select the corresponding plane and points where the points belonging to the plane become the largest. After separating the selected points from the whole set, the above method is continuously performed using the remaining points. Specifically, when dividing using the RANSAC method, the entire point cloud is classified into point data belonging to the same plane, and a refinement grid is used to remove invalid segments. The segmentation technique using RANSAC divides the point cloud based on the mathematical plane represented by the equation. Therefore, if it belongs to the other plane but is close to the plane, it is recognized as the same segment, so a refinement process is needed to remove the points that are misclassified.

A typical method of parameter-based partitioning is Hough transform. Each point in the three-dimensional space forms the planes in the parameter space. The planes on the parameter space corresponding to the points lying on the same plane in the three-dimensional space intersect at the same position in the space. Therefore, extracting the point at the intersection of the largest number of planes in the parameter space extracts the points lying in the same plane in the three - dimensional space. After extracting the points lying on one plane, the above-mentioned method is continuously performed using the remaining points.

The point cloud division using the 3D BSP (Binary Space Partitioning) tree is a binary tree which contains information for dividing the space, and manages the faces in the 3D space by separating the front and back of the specific reference plane. By dividing and processing the space, it is possible to quickly process what data is in a space by a binary search method. That is, the search is performed by distinguishing whether it is front, back, left, or right.

It is possible to perform the division using the division method other than the division method as described above, and it is obvious that it is not limited to the kind of the division method.

Next, the plane for the indoor space is determined using the divided point cloud (S140). Specifically, when segmentation is performed, a surface to which each point belongs can be generated.

Next, a boundary line for the indoor space is extracted using the determined surface (S150). A plane is generated through division, and a line on which the plane and the plane meet can be determined. The line generated by the surface and the face meet is referred to as a boundary line, and the boundary line for the entire indoor space can be determined. Specifically, when the point cloud of the indoor space is divided by using the RANSAC method, the segmented points are binarized using an Outline Trace Grid, and the points are binarized using the Vectorization technique The boundary line is extracted. The size of the grid can be set differently according to the characteristics of the indoor environment as in the refining process. When the boundary extraction process ends, the binary image is inverted to extract an outline of an empty area in a segment area such as a window. At this time, an outline is extracted only for an empty space having an area of a certain size or more, and the extracted outlines are recorded in a file in the form of an IFC file, which is a 2D or 3D CAD or BIM standard format.

The method may further include dividing the point cloud of the indoor space until the error is converged within the threshold value with respect to the surface where the error occurs, when the point cloud is compared with the point cloud using the boundary line and the indoor space S130) and determining the plane using the divided point clouds (S140).

Finally, when a boundary line is finally generated through the above operation, a 3D model of the indoor space is generated using the remaining point cloud for the boundary line and the remaining area in which the boundary line is not extracted (S160). Specifically, the surface is extracted using the finally created boundary line, the 3D point cloud is directly used for the remaining indoor space where the boundary line can not be extracted, or the 3D mesh generated from the 3D point cloud is used as the 3D authoring software Dimensional model of the indoor space can be created by the user directly constructing the three-dimensional graphic.

FIG. 3 is a view showing a point cloud of a building that is divided by using a surface determined by a divided point cloud; FIG. FIG. 4A is an exemplary view showing a state in which 3D mesh data and a boundary line are extracted, FIG. 4B is a view illustrating an example in which a boundary cloud is not used in boundary line extraction of a 3D model, FIG. 8 is an exemplary view showing a state in which mesh data is formed. FIG.

Referring to FIG. 2, a point cloud of a building scanned at various places in an indoor space is removed by removing noise and created a point cloud for an indoor space using a ground lidar.

Referring to FIG. 3A, after the noise of the point cloud of the indoor space scanned using the ground lidar is removed, the divided image is shown. The faces are determined by the divided point cloud.

Referring to FIG. 3B, a boundary line of the indoor space can be seen using the plane determined by the divided point cloud.

Referring to FIG. 4A, the plane of the indoor space is determined using the divided point clouds, and the shape of the 3D mesh data (red color) of the boundary line (white) extracted by the determined plane and the boundary line is not extracted can see.

Referring to FIG. 4B, the mesh data (red color) is formed using the remaining point cloud not used for boundary line extraction of the three-dimensional model.

Meanwhile, the point cloud-based indoor space three-dimensional model extraction method obtained in the above-mentioned ground-level method according to the present invention is implemented as a program and is implemented as a computer-readable recording medium (CDROM, RAM, ROM, floppy disk, hard disk, Magnetic disk, etc.).

According to the present invention, a plane constituting an indoor space is automatically extracted from a three-dimensional point cloud acquired through a laser scanning technique, a plane is extracted, and a three-dimensional model is extracted Thus, the time and cost can be drastically reduced as compared with the conventional method. In addition, if a technique of generating a three-dimensional indoor space model automatically or semiautomatically with a point cloud obtained by using the ground lidar is applied, it can be efficiently used for planning, designing and maintenance of an existing indoor space But it has an effect of contributing to the construction of an indoor space database in the field of urban planning, three-dimensional geographic information system (GIS) and building information modeling (BIS).

It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims and their equivalents. Obviously, such modifications are intended to be within the scope of the claims.

Claims (6)

A point cloud-based indoor space three-dimensional model extraction method,
A step 110 of scanning the indoor space using the ground lidar;
A step 120 of merging the point clouds scanned in various places of the indoor space to generate a point cloud for one indoor space;
130) of segmenting the point cloud of the merged indoor space;
140. The method of claim 140, further comprising: determining a plane using the divided point cloud;
A step 150 of extracting a boundary line with respect to the indoor space using the determined surface, and a step 150 of comparing the point cloud with respect to the set surface and the indoor space using the boundary line, if the error occurs, Comprising the steps of: dividing the point cloud in the indoor space; and performing the step 140 of determining the surface using the divided point clouds until the error converges to within the threshold value.
And a step 160 of generating a three-dimensional model for the indoor space by using the remaining point cloud that is reduced in size relative to the initial point cloud for the remaining area in which the boundary line has not been extracted. A three - dimensional model extraction method of indoor space based on point clouds obtained in. The method according to claim 1,
In operation 120,
And registering the absolute coordinates of the point cloud with respect to one indoor space after the merging. The method of extracting indoor space three-dimensional model based on point clouds obtained in the above-mentioned way. The method according to claim 1,
In operation 130,
The entire point cloud of the indoor space is classified into point data belonging to the same plane by using the point cloud segmentation method to remove invalid segments,
In operation 150, the segmented points are binarized and the boundary line for the indoor space is extracted using a vectorization technique. The point cloud-based indoor space Three - Dimensional Model Extraction Method. The method of claim 3,
(3) extracting a point cloud-based indoor spatial three-dimensional model obtained from the ground-based model using a refinement grid for eliminating the invalid segment and using an outline trace grid for binary imaging. Way. The method according to claim 1,
In operation 160,
The surface is finally extracted using the generated boundary line and the 3D point cloud is directly used for the remaining indoor space or the 3D mesh generated from the 3D point cloud is constructed by using the 3D rendering software And generating a three-dimensional model of the indoor space. The method of extracting a point cloud-based indoor space three-dimensional model acquired in the ground level. A recording medium on which a program for causing a computer to execute the method according to any one of claims 1 to 5 is recorded.

Images