KR101549155B1 - Method of automatic extraction of building boundary from lidar data - Google Patents

Abstract

The present invention relates to a method for extracting the straight line boundary of a structure using light detection and ranging (LiDAR) data. According to the present invention, the method comprises the following steps: dividing a peripheral area into eight directions around one point, searching for a point of each area one by one, and extracting boundary points of a structure in irregular point cloud data by using a convolution method which regards total nine points including a corresponding point as one area on a convolution; aggregating the boundary points based on an elevation value of the extracted boundary point, and the adjacency; classifying the aggregated boundary points into ground points and non-ground points; extracting a point which exists at the outermost of the classified ground points to simplify the boundary points; and calculating an angle between points which exist on both sides around a central point among consecutive three points of the simplified boundary points, and removing the point which has the angle greater than a threshold value to linearize the outline.

Description

METHOD OF AUTOMATIC EXTRACTION OF BUILDING BOUNDARY FROM LIDAR DATA FIELD OF THE INVENTION [0001]

The present invention relates to a method of extracting a straight line boundary of a structure using Lada data, and more particularly, to a method of extracting a straight line boundary of a structure using a Lada data, Extracting a boundary point of the structure from the irregular point group data using a line technique for searching one point for each area and considering nine points to one point as a line window area; Clustering boundary points on the basis of the elevation values and adjacencies of the extracted boundary points; Classifying the clustered boundary points into a ground point and a non-ground point; Extracting points existing at the outermost of the classified ground points to simplify the boundary points; Calculating an angle between points on both sides of the simplified boundary point centered on a center point of three consecutive points and removing a point whose size exceeds a threshold value to linearize the outline; And a linear boundary extraction method using the Lada data.

Since the modern map production by photogrammetry has been carried out, the production environment of satellite images and advanced laser equipments has been rapidly changed due to the development of surveying and mapping techniques, thereby making a lot of investment in improving precision and shortening work time , Especially LiDAR (Light Detection and Ranging) system, which is excellent in precision, efficiency and economical, is very suitable for obtaining precise elevation information for a large area and is a new surveying method .

Generally, LiDAR is mounted on an aircraft, acquires data, observes the time the laser beam travels from the sensor to the object, observes the distance to the object, and determines the three-dimensional coordinates of the object. The LiDAR data is useful for acquiring data in DSM (Digital Surface Model) format because it has altitude information and reflection intensity information of not only natural terrain but also artificial facilities. It is used for DEM (Digital Elevation Model) production for terrain part and contour extraction for artifacts such as buildings.

DSM generation for the representation of land and water in the mapping process using LiDAR data can be generated by using the original data with little conversion and DEM etc. is easily done by dividing by using the difference of elevation, For extraction, processing such as filtering, splitting, or sorting from raw data must be performed. Particularly, the outline of the divided building is mostly unclear and it is zigzag shape and the corners are inaccurate, so further processing such as generalization and normalization is necessary.

The study on building extraction using LiDAR, which has been introduced so far, can be roughly classified into a method using point data, a method of interpolating point data in a grid form, and a method of data processing based on a feature.

Since the point-based processing method uses all of the LiDAR viewpoints, it requires a large amount of computation, requires a large amount of data to be processed, and has a problem in that the complexity of the processing increases because the data distribution structure is not constant. Even though there is a disadvantage that data processing should be performed for each subregion, since it uses the point data as it is, it shows the best result in terms of accuracy. On the other hand, the data processing method through interpolation in grid form is fast, but geometric distortion caused by interpolation of data is generated because the LiDAR viewpoint is interpolated spatially, and the actual shape of the building is deformed at the boundary of the building Dimensional information is damaged. In addition, the feature-based data processing method compensates for the disadvantages of point-based processing methods that require all points to be processed, but it requires more complex and long processing time and requires user intervention in the grouping of surface patches and the identification of facilities .

Korean Patent Registration No. 10-0857529 (Extraction Method of Building Boundary from Raida Data)

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and it is an object of the present invention to extract the boundary points of a building using three-dimensional information of point data in irregular point group data acquired from aviation LiDAR, This paper proposes a method of 3D rendering of a building with a straight line of building boundaries.

In order to achieve the above object, according to the present invention, there is provided a method of extracting a straight line boundary of a structure using a Lada data, wherein the method comprises the steps of: Extracting the boundary points of the structure from the irregular point group data using a line technique in which a total of 9 points are considered as one line window area; Clustering boundary points on the basis of elevation values and adjacencies of the extracted boundary points; Classifying the clustered boundary points into ground points and non-ground points; Simplifying boundary points by extracting outermost points among the classified ground points; Calculating an angle between points existing on both sides of a point located at the middle of three consecutive points of the simplified boundary points and removing a point whose size exceeds a threshold value to linearize the outline; , The straightening of the outline is performed such that when there is at least one angle that is less than a critical angle between an intermediate point and a point located at both ends of the three consecutive points, If the angle between the point located at the middle of the three consecutive points and the points existing at both sides of the consecutive three points is greater than the critical angle, the angle formed by the extended lines of the line connecting the points is calculated, In this case, the remainder except the middle point is deleted, and the point located at the middle of the three consecutive points and the point If is greater than each also critical angle on the angle of all the extension of the large and all segments than the critical angle between the dots, leaving the closer that the angle of the critical angle between the point in the middle of the point and exists at the both rest is characterized in that the deletion.

In the present invention, the step of clustering the boundary points may include: inputting an identifier starting from a boundary point having no identifier; Determining whether or not a boundary point having no identifier exists; Determining whether the inclination degree with respect to the boundary point is within a threshold value, assigning the same identifier to the corresponding boundary point, and setting a starting point; Determining whether or not an identifier is assigned to all of the boundary points; And terminating the clustering of the crack boundary points.

In the present invention, the step of classifying the clustered boundary points into a ground point and a non-ground point includes estimating a minimum height value of a region included in the filter, assuming a size and height threshold value of the filter, And a local maximum value filtering technique for selecting an area higher than an assumed threshold value is used.

In the present invention, the step of simplifying the boundary point may include: searching for all boundary points having the same identifier; Searching boundary points closest to neighboring points based on the boundary points having the same identifier; And recognizing the boundary points of the outermost edges for each identifier.

delete

According to the present invention as described above, boundary points of a building are extracted by using three-dimensional information of point data in irregular point group data obtained from aviation LiDAR, and building boundary lines are straightened by using the adjacent three points of the extracted boundary points and the angle formed between them Make the 3D drawing of the structure possible.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a schematic flow chart showing a 3D rendering method of a structure using the Lada data according to the present invention; FIG.
Fig. 2 is an exemplary diagram showing the principle of a circuit; Fig.
Figure 3 is an exemplary diagram illustrating simplification by the Douglas-Pouquet algorithm;
Fig. 4 is an exemplary diagram showing an algorithm of the distance error a and the error aberration b; Fig.
5 is an aerial photograph of an area of interest in accordance with an embodiment of the present invention.
Figure 6 is LiDAR data of a target area according to an embodiment of the present invention.
Figure 7 is an illustration of a typical line window and buffer roll-on.
8 is a flowchart showing a circuit technique using a buffer polygon.
9 shows the result of applying the boundary detection algorithm.
10 is a flow chart showing a clustering process of boundary points.
Fig. 11 is a result of clustering of building boundaries.
12 shows the result of ground point extraction.
13 shows the elevation characteristics of the target area.
Figure 14 shows the results of simplification of boundary points of a building.
15 is an angle calculation at a point.
16 is a flowchart showing a straightening process of a building outline;
17 illustrates a point removal method of a linearization algorithm;
18 shows a process of performing a linearization algorithm.
19 shows the result of performing the linearization algorithm.
20 is an exemplary view showing an entrance of a building according to a three-dimensional analysis result of an aerial photograph and a LiDAR data of a target area according to an embodiment of the present invention;

BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, features and advantages of the present invention will become more apparent from the following detailed description taken in conjunction with the accompanying drawings, in which: FIG.

The present invention extracts boundary points of a building using three-dimensional information of point data in irregular point cloud data acquired from aviation LiDAR and straightens the building boundary by using the adjacent three points of the extracted boundary points and the angle between them, It is a technical characteristic that it is possible to make an image.

1. Data processing method

(One). Building Boundary Extraction Method.

In general, in the grid-like data such as image or DEM (Digital Elevation Model), the boundary includes a lot of information about the position, size, shape and texture of the object. In the case of the image data, The method of finding the pixel corresponding to the boundary exists at a point where a discontinuous change occurs from a high place to a low place.

In the image processing, a convolution technique is generally used for boundary detection. The line technique is an image processing technique that is mainly used for edge detection, image smoothing, or image sharpening in an image. It can be said to be a sum obtained by multiplying the pixels adjacent to the original pixel by the weighting factor, The rate is determined by a small matrix called a line mask. As shown in FIG. 2, a moving window called a line window is located at the center of each pixel, and a new output pixel is generated through operation with a line mask. The dimension of the matrix is generally set to 3 or 5 Use an odd number of dimensions.

In general, the boundary of the image contains a lot of information about the position, size, shape, and texture of the object, and it exists at a point where discontinuous changes occur from low to high contrast or from high to low The method of obtaining the pixel corresponding to the boundary is called boundary detection. The simplest way to detect a boundary is to use a homogeneity operator or a difference operator that can detect the boundary at a correct rate by determining the maximum value by subtracting a series of pixels. As a detection method, there is a method using a first order differential operator and a second order differential operator. In general, the first-order differential represents the slope of the tangent line. As a result, the first-order differential operator is a Roberts operator having noise-sensitive characteristics based on the gradient of light and darkness between pixels of the image, Sobel operators sensitive to horizontal or vertical boundaries. Also, Laplacian operator is a typical second order differential operator with the advantage that the detected boundary forms a closed curve.

(2). How to simplify building boundaries.

Simplification is one of the generalization types of vector structure data, such as smoothing, amalgamation, or collapse, and is used to define points that maintain the characteristics or shape of the original line. It is a process of eliminating unnecessary surplus points in expressing the characteristics of a line, which has advantages such as reduction of drawing time and storage space, faster screen diagram, conversion of data, and shortening of various processing time. However, there are many difficulties in selecting objects by eliminating as many unnecessary points as possible, while showing the overall characteristics of objects such as shapes, positions and features, which are numerical values, A Douglas-Peucker algorithm, a distance error algorithm, a correction algorithm, and a Lang algorithm.

The Douglas-Pouquet algorithm is a method of simplifying the line by reducing the number of points by seeing the line as a connection of points. Typically, without calculating the mathematical relation of neighboring coordinate pairs, There is a way to remove the rest (Tobler, 1964) and divide all the points into several groups, leaving only one point at random for each group and delete the rest. However, this method has the problem of losing the characteristic of the phenomenon itself because it removes the point without consideration of the characteristics of the line or contour. To solve this problem, there is a local processing method of analyzing the characteristics of neighboring points and deleting points.

The distance error algorithm is similar to the Douglas-Pouquet algorithm but differentiates in that simplification is performed using three successive vertices. The distance error algorithm compares the start point and the end point of the first three points and omits the second point when the perpendicular distance between the second point and the second point is less than the tolerance (McMaster, 1987) Since the shape of the complex spatial object can be reflected in the simplification, it is possible to minimize the change of the original spatial object shape.

The Lang algorithm (Lang, 1969) uses the angle algorithm instead of the distance algorithm (McMaster, 1987) in a similar way to the distance error algorithm in that the error algorithm is performed on the basis of three points It is necessary to set the number of points and the tolerance to determine the size of a certain section and to calculate the original geometrical characteristics of the linearity ≪ / RTI >

(3). Ground point classification.

Local maxima filtering (hereinafter referred to as "Local Maxima filtering") is a typical method for extracting ground points from which various features such as buildings are removed from vector-based data, which is a point cloud of irregular distribution such as LiDAR data.

Local maxima filtering is a method of dividing a building area in vector - based data. It has the advantage of less distortion and distortion of data because it can directly use LiDAR data, which is a point cloud with irregular format of original data. First, local maxima filtering is based on the simple fact that features such as buildings are higher than surrounding terrain surfaces, and ground points can be separated by using height differences.

However, the method of dividing the building area by setting the overall height threshold on the whole of the raw data is easy to apply in the flat area, but it can cause the problem in the undulating landform, so it reflects the change of the regional terrain The local maxima filtering is applied to apply the height threshold locally.

)을 계산한 후,

보다 높이 임계치

만큼 높은 점들을 건물점으로 간주한다. 즉 local maxima 필터는 일정한 범위의 필터에 포함된 지역의 최저 높이를 구한 후 그 최저 높이보다 임계값 높은 지역을 선택하는 건물로 분류하게 되며 사용되는 변수는 필터의 크기(

)와 높이 임계값(

)이 고려된다.Local maxima filtering first determines the minimum height value of the points in the filter

), ≪ / RTI >

Higher Threshold

Are regarded as building points. In other words, the local maxima filter finds the minimum height of a region included in a certain range of filters, and then classifies it as a building that selects a region with a higher threshold value than its minimum height.

) And the height threshold (

) Is considered.

는 건물 최소 높이값을 고려하여 결정하고 필터의 크기

은 건물을 포함할 수 있을 정도의 크기로 결정해져야 하며 필터의 경사(

)가 대상지역 경사도보다 크도록 선택되어야 한다. 즉 경사가 급한 언덕에 존재하는 건물을 분할해내기 위해서는 필터의 크기가 작을수록 유리하다.At this time, the size and height threshold of the local maxima filter are determined by using the prior knowledge of the building,

Is determined by considering the minimum height value of the building and the size of the filter

Must be sized to contain the building and the slope of the filter

) Must be selected to be greater than the target area gradient. In other words, the smaller the size of the filter is, the more advantageous it is in order to divide a building existing on a sloping hill.

Local maxima The limit of filtering occurs when there is a building with a relatively large area within the site. In order to divide such a building, the filter size must be increased. At the same time, the filter slope becomes smaller. Therefore, in a region where the slope of the ground is larger than the slope of the filter, the ground is divided into buildings.

2. Application and Discussion

(1) Target area and data acquisition.

The area to be applied by the algorithm according to an embodiment of the present invention is a member of Daejeon Metropolitan City and the current status of the area is shown in FIG. The LiDAR data used in the present invention was acquired at an altitude of about 1,400 m at a point density of about 4points / m ² using an ALTM 3070 laser scanner manufactured by Optech, Canada.

In addition, from the LiDAR data presented in the present invention, a boundary point extraction algorithm, a building boundary point clustering algorithm, and a building line outline linearization algorithm by simplifying a building boundary point are implemented using Microsoft's Visual Basic 6.0 and ESRI's MapObjects 2.2. Arc / Info 9.2 and ArcView 3.3 from ESRI have been used for this purpose.

(2). Building boundary points and clustering of LiDAR data.

(2-1). Building boundary point extraction.

In the present invention, the boundary extraction of the structure (building) is implemented by the boundary detection algorithm used in the image processing field.

However, in the case of regular grid-type data with a structure in which a grid of the same size is repeated such as an image or a DEM, it is possible to apply an image processing technique such as a boundary detection algorithm or filtering using a line technique. On the other hand, irregular Direct application of circuit techniques is not easy in data of vector form with distribution.

Therefore, in the present invention, a method of applying a circuit technique to irregular point data is utilized for boundary extraction in LiDAR data.

This technique utilizes a buffer polygon as shown in FIG. 7 (b) instead of the line mask used in the general line technique as shown in FIG. 7 (a), and buffers a certain area around an arbitrary point. And then extracts the nearest point from the original center point in each direction and then uses a total of 9 points including the center point to perform the 3-line window function. 8, and can be directly applied to LiDAR data, which is a vector format data. If two or more points are extracted from each direction, an effect similar to increasing the size of the line window can be obtained. Particularly, in the case of point cloud data in which points are irregularly distributed, even when a point searched in the buffer region due to the occlusion region is concentrated in one direction or a small number of points are searched, have.

The boundary extraction algorithm uses a prewitt operator with a clear boundary and a simple operation mask to extract clear boundaries of buildings among the first - order differential operators.

FIG. 9 shows a result of extracting a boundary of a building using boundary detection using an algorithm for applying a line technique of vector data. Here, the boundary of the ground outside the building or the building boundary point that is mistakenly classified as the building is removed from the boundary of the building boundary and simplification of the boundary point of the building. Therefore, no preprocessing process such as noise processing was performed.

(2-2). Clustering of building boundary points.

The boundary points extracted by the boundary detection algorithm based on the altitude difference simply have information on the boundary, but do not include information on the boundary of an object or the boundary of the same object. Therefore, in order to grasp only the boundaries of the building, it is necessary to provide an identifier capable of identifying the same object to these boundary points. For this purpose, the extracted boundary points are clustered by the method shown in FIG. 10 based on the elevation value and the adjacency.

If a boundary point is found, the adjacent boundary points are searched for through buffer analysis, and the gradient is calculated for the detected points. If the boundary points are within a threshold value, identifiers are assigned to the same cluster . The same process is repeated from the point where the identifier is given to search for the boundary points to which no other identifier is assigned and it is repeated until there is no boundary point having a gradient within the threshold value or the identifier is assigned to all the boundary points. In order to remove objects other than buildings such as roadside trees after clustering is completed, a cluster with a small number of objects is not recognized as a building. 11 is a result of clustering the boundary points extracted from the target area.

(3). Straightening of building outline by simplification of building boundary point.

(3-1). Ground point classification.

Since the method of extracting a building boundary according to the present invention utilizes points classified as a ground for eliminating unnecessary surplus points during linearization of building boundary points extracted from LiDAR data, the present invention uses a pre- The local maxima filtering is used to classify the ground and non-ground points.

Local maxima filtering is a method of dividing a building area in vector - based data. Especially, it is known that extraction results are very good when the size of a building is constant in a region with little fluctuation. 12 is a result of classifying the ground and non-ground points by local maxima filtering.

However, the subject area of the present invention includes a mixture of various high-altitude apartment buildings, ground surface materials, parked vehicles and underground parking lots, and various types of topographic objects having different elevation values. In particular, Since the other direction consists of a steep slope, if the ground and non-ground points are classified using the local maxima filter, which is not classified based on the lowest region in the filter, the ground may be misclassified as non-ground. The feature of the local maxima filter is also reflected in the application result of the present invention, so that the entrance of the underground parking lot and surrounding areas are classified as the non-ground surface as shown in FIG.

However, in the present invention, the landmark classification result is used for eliminating surplus points for simplifying the boundary points in the process of straightening the building boundary points. The classification accuracy of the landmark points influences the extraction of the digital map layer using the LiDAR data This result is used as it is.

(3-2). Simplify building boundary.

In general, the outline of a building is a straight line. In particular, the building layer of a digital map is composed of straight lines connecting the singular points of a building. However, in the boundary point extraction result, many boundary points are extracted more than necessary, and a thick zigzag shape is exhibited by irregular distribution characteristics of LiDAR data. Therefore, in order to straighten the boundary line of the building, it is necessary to simplify the building boundary point by eliminating the surplus point except the singular point that can express the characteristic of the building.

 In the present invention, since the LiDAR data has a very high point density, a simplification process is performed using the clue that the outermost points of the building boundary points extracted from the LiDAR data are likely to represent the outline of the building.

The simplification of the boundary point first grasps all boundary points with the same identifier from the whole data. When the recognition is completed, the boundary points nearest to the ground points existing in the vicinity are searched based on the boundary point having the corresponding identifier. At this time, the boundary points that are searched can be selected even if they are searched from other ground points, and the selected boundary points are stored so that they are not overlapped. Also, considering the adjacency between the boundary points, the distant boundary points are excluded even if they have the same identifier.

When the process is performed for all the identifiers, the boundary points of the outermost boundaries are grasped and the same process as shown in FIG. 14 can be obtained by performing this process in the corresponding region according to the embodiment of the present invention.

(3-3). Straightening the building outline.

If a building boundary outline point is selected, you must select outliers to create the outline of the building, such as a digital map form. However, since the extracted result includes all the points on the outer edge of the building evenly, it can not be used as the final boundary of the building because it is very irregular like a digital map like a linear outline.

Therefore, in the present invention, the order of the boundary points is given based on the adjacency of each point, and the linearization algorithm applying the angular error algorithm among the linearization algorithms is applied. The straightening algorithm applied in the present invention is a method of calculating an angle between points existing on both sides of a center point of three consecutive points and eliminating a point whose size exceeds a threshold value, Unlike the angle error algorithm, which is made by grasping the angle formed by two points, if the reference point continues to move, it is repeated until the angle within the threshold value is found at the starting point. If the outline of the building is the outline of the building The problem that the shape may become unstable can be solved.

The angle formed with the adjacent boundary points at each boundary point is calculated by using the x and y coordinates for each point as shown in FIG. 15, and FIG. 16 shows a straightening process of the building outline.

In particular, the algorithm proposed in the present invention considers the following three cases in the process of deleting points in order to find a singular point.

First, when there is one or more angles smaller than the critical angle, the remaining points except the critical angle exist are deleted (Fig. 17 (a)).

Second, if all the angles are larger than the critical angle in three consecutive points, calculate the angle formed by the extension lines of the line connecting the points, and if the angle is smaller than the critical angle, remove the rest except the middle point. (Fig. 17 (b)).

Third, if all the angles are larger than the critical angle and the angles on the extension line of all the line segments are larger than the critical angle, the remaining points are deleted. (Fig. 17 (c)).

This process is repeated until all angles are smaller than the critical angle, and when a certain number of points remain, the straightening process is completed.

FIG. 18 shows the process of straightening the outline points of one of the buildings in the target area in a stepwise manner, and shows a numerical map of the target area in order to grasp the exact change. FIG. The position of the node is shown together with the numerical map.

(3-4). Accuracy assessment and analysis of results.

In order to analyze the accuracy of the linearization results, the present invention uses a building layer of a 1: 5,000 digital map of the target area.

20 (b) and 20 (c), the three-dimensional analysis of the aerial photographs and the LiDAR data shows that the convex portions of the buildings displayed on the digital map as shown in FIG. 20 (a) Chapter 2 Working Method of Section 1 of the Digital Mapping Work Notified by Geographical Information Department of the Ministry of Land, Infrastructure, Transport and Tourism Geographical Information Department of the Ministry of Land, Transport and Maritime Affairs. Section 1 of Article 9 (Description) (Vector Editing), a straight line building is required to have only one point data at each corner and must be closed. Therefore, they are excluded from the accuracy evaluation because they are not consistent with the object of the present invention.

1: 5,000 As a result of the accuracy evaluation using the numerical map, a total of 13 buildings in the target area were extracted, and the plane position results showed a position error of 0.078 m to 1.466 m at the minimum as shown in Table 1 below.

Table 1 shows the error of the plane position (unit: m) of the extraction result based on the building layer. division 1st place 2nd place 3rd place 4th place Average Building 1 1.205 0.685 0.405 0.725 0.755 Building 2 1.082 0.301 1.319 1.207 0.977 Building 3 0.795 0.508 0.524 1.461 0.822 Building 4 1.253 1.112 0.276 0.839 0.870 Building 5 1.030 1.342 0.680 0.781 0.958 Building 6 1.188 0.335 0.435 1.100 0.765 Building 7 1.347 1.151 0.556 0.127 0.795 Building 8 0.203 1.116 1.090 1.105 0.878 Building 9 1.441 1.466 0.766 1.227 1.225 Building 10 1.111 0.824 0.494 0.365 0.698 Building 11 0.818 0.164 0.738 0.125 0.461 Building 12 1.013 0.989 0.078 0.983 0.766 Building 13 0.411 0.779 1.359 1.110 0.915 Overall average 0.837 Overall standard deviation 0.927

In addition, the standard deviation of the error with respect to the plane position of the result obtained through the algorithm applied in the present invention is 0.927m and the maximum error is 1.466m. According to Article 10 (Accuracy) of the numerical mapping work, It can be used to update the building layer of 1: 5,000 numerical map because it satisfies the specification of 1: 5,000 scale (standard deviation of plane position, maximum error of 2.0m).

4. Conclusion.

The present invention is to grasp the applicability of the numerical map to the building layer modification update by extracting the points located at the boundary of the building from the LiDAR data and connecting each adjacent point and straightening them.

First, we can extract boundary points of buildings using 3D information of point data in LiDAR irregular point cloud data and straighten the building boundary by using the adjacent three points of the extracted boundary points and the angle formed between them.

 Second, as a method to facilitate linearization of the building boundary, we classify the ground points from the LiDAR point cloud data and use the points classified to the ground to simplify the extraction of the points at the outermost boundary of the building boundary points extracted from LiDAR data .

Third, the standard deviations and the maximum errors of the boundary line extracted by the proposed algorithm were 0.927m and 1.466m, respectively, and 1: 5,000 It is possible to utilize for updating building layer of digital map.

Claims (5)

In the linear boundary extraction method of a structure using RDA data,
By using the circuit technique that divides the surrounding area into eight directions based on one point and searches one point for each area and considers nine points to the corresponding point as one line window area, the irregular point group data Extracting boundary points of the image;
Clustering boundary points on the basis of elevation values and adjacencies of the extracted boundary points;
Classifying the clustered boundary points into ground points and non-ground points;
Simplifying boundary points by extracting outermost points among the classified ground points;
Calculating an angle between points existing on both sides of a point located at the middle of three consecutive points of the simplified boundary points and removing a point whose size exceeds a threshold value to linearize the outline; ,
Wherein the straightening of the outline comprises:
If there is at least one angle at which an angle between a point located at the middle of the three consecutive points and a point existing at both sides is less than the critical angle,
If the angles between the points located at the center of the three consecutive points and the points existing on both sides are both greater than the critical angle, the angle formed by the extended lines of the connected line segments is calculated. In addition,
If the angle between the center point of the three consecutive points and the points existing on both sides of the point are larger than the critical angle and the angle on the extension line of all the line segments is larger than the critical angle, And removing the remainder. A method of extracting a straight line boundary of a structure using Rada data.
The method of claim 1, wherein clustering the boundary points comprises:
Starting from a boundary point having no identifier, and inputting an identifier;
Determining whether an inclination from a boundary point to which the identifier is input is within a threshold value, assigning the same identifier to the boundary point, and setting a starting point;
Determining whether or not an identifier is assigned to all of the boundary points; And
Terminating the clustering of the crack boundary points;
The method comprising the steps of: (a) extracting a straight line boundary of a structure using the Lada data;
The method of claim 1, wherein classifying the clustered boundary points into surface points and non-surface points comprises estimating a minimum height value of a region included in the filter, assuming a size and a height threshold of the filter, And a local maximum value filtering technique for selecting an area higher than a threshold value that is higher than a threshold value that is already higher than the minimum height is used.
2. The method of claim 1, wherein simplifying the boundary points comprises:
Searching all the boundary points having the same identifier;
Searching boundary points closest to neighboring points based on the boundary points having the same identifier; And
Identifying the outermost boundary point for each identifier;
The method comprising the steps of: (a) extracting a straight line boundary of a structure using the Lada data;
delete

Images