KR101115489B1 - Method and system for generating numerical modeling input materials having high revolution based on raster by using airborne lidar data - Google Patents

Abstract

PURPOSE: An aviation LiDAR(Light Detection And Ranging) material high definition numerical modeling input material creation method and system are provided to improve the reliability of a verification step by improving the quality of a manufacturing process. CONSTITUTION: An input unit(310) receives aviation LiDAR data from an LiDAR point group. A processing unit(300) calculates reference area from an LiDAR point group by allocating the aviation LiDAR data to lattices. The allocation probability of a specific class is calculated. The processing unit creates virtual LiDAR points within the lattices by calculating a searching area according to the reference area. When the probability is bigger than a threshold value, the processing unit allocates the lattice into the specific class.

Description

Method and system for generating numerical modeling input materials having high revolution based on raster by using airborne LiDAR Data}

The present invention relates to a method and system for generating high resolution numerical modeling input data from aerial LiDAR point cloud data.

Land cover is an important surface input for numerical simulations. Therefore, it is necessary to secure the accuracy and reliability of the input data generation process in numerical modeling for urban weather application. Aviation LiDAR (Light Detection And Ranging) technology consists of a GPS device to determine position values, an inertial navigation device (IMU) to collect rotation information, and a laser device to measure the relative distance between the plane and the ground. The data processing technique for surveying and utilizing to obtain the spatial position information (X, Y, Z) of the reflection point by calculating the arrival time of the laser pulse which is reflected back by scanning. Here, the reflection point from which the spatial position information is obtained is called a LiDAR point. By the way, the Lidar point is not uniformly obtained on the ground surface because the ground surface (eg, water) absorbs the laser signal or shakes as the plane moves. Airborne LiDAR data, including LiDAR point clouds, provide objective and detailed information on land cover, an important input for urban weather simulation.

1 is a view for explaining the classification process of the existing aviation LiDAR data.

Referring to FIG. 1, the aviation LiDAR data surface classification result accuracy evaluation is based on reference data 202 from the total classified data 201 based on the reference data 202. (commission) The ratio of the normal classification result except the error (209) is calculated as a percentage. Here, the classification result 201 should be included in the true value based on the reference data 202 (203) excluded (206) missing error 210 and should be excluded in the true value (204) included (205) ) Is an additional error 209.

2 is a view for explaining the surface classification method associated with the existing aviation LiDAR data.

Referring to FIG. 2, the surface classification of airborne LiDAR data is performed in stages. A point group corresponding to the ground surface is extracted through a roller technique from the LiDAR data 211 where no classification is performed (212), and all the point groups on the ground surface are classified as low vegetation (213). Subsequently, the point group with a height of 50 cm or more from the ground is classified as a shrub (Mid Veg.) From the data classified as vegetation (214), and the point group with a height of 2 m or more from the ground is classified as a low vegetation (215) from the result classified as a shrub (215). ) Classified as point buildings based on reflection characteristics and geometry of aerial LiDAR data from point tree points (216).

3 is a view showing a classification result according to the surface classification method of the existing aviation LiDAR data.

As shown in Figure 3, the size of the sample for the evaluation of the overall classification result is 100 and ground (Ground): 20 (221), low vegetation: 20 (222), shrub (Medium Vegetation) for each class When classified as 10 (223), High Vegetation: 20 (224), Building: 20 (225), the classification accuracy of each class excluding missing error and additional error is 90%. , Vegetation: 84%, shrubs: 81%, trees: 90%, buildings: 90%. In this case, the class indicates which of the plurality of items exist on the ground.

However, the overall classification accuracy of the aviation lidar data adds only the true value of each class classification step, which is 73%.

Overall classification accuracy: (15 + 15 + 7 + 18 + 18) / 100 * 100 = 73%

In addition, as the area to be classified has a three-dimensional structure surrounded by high-rise buildings and facilities as in a city, or even in a natural environment, the accuracy of classification is lower in a forest structure environment composed of multiple trees. Therefore, the classification accuracy that can be expected through the automatic classification technique using the air lidar data is about 60% to 70%.

Misclassification results of 30% to 40%, which cannot be handled by the rest of the automatic classification technology, must be manually moved to the true value by hand, which requires enormous cost and processing time. Performing these misclassifications directly requires more labor and costs than necessary. On the contrary, if the misclassification is not corrected, the false information production will increase in the general lattice process.

As a result, in order to effectively use the aviation LiDAR data for numerical simulation, the land cover information must be extracted from the point cloud data and rearranged into a raster data structure. There is a need to provide an environment that can directly satisfy the user's desired quality and algorithm that produces grid information quickly while minimizing many errors.

In order to solve the problems described above, the present invention applies high-resolution numerical modeling input data having a high probability of being true by applying an algorithm that analyzes the probability of land cover distribution according to the size of the grid required by the user from the aviation lidar data. An object of the present invention is to provide a method and system.

According to an embodiment of the present invention for solving the above-described problem, a system for generating high-resolution numerical modeling input data based on aviation LiDAR is a aviation including Light Detection And Ranging (LiDAR) point cloudes. Splitting the air lidar data including the LiDAR input group and the LiDAR point group into grids, and obtaining a representative area of one LiDAR point from the number of LiDAR points included in the LiDAR point group After calculating a search radius according to the representative area, generate virtual LiDa points with random numbers in each grid, and specify from the distribution of the virtual LiDa points within the search radius of LiDa points included in each grid. A processor that calculates a probability that can be assigned to a class and allocates the grid to a specific class when the probability is greater than a predetermined threshold. It includes.

In addition, according to another embodiment of the present invention, a method for generating high-resolution numerical modeling input data based on aviation lidar may include aviation lidar data including light detection and ranging (LiDAR) point cloudes. Obtaining a representative area of a lidar point from the number of lidar points included in the step of receiving, dividing the aerial lidar data including the lidar point group into the lattice points, and Calculating a search radius according to the representative area, generating virtual lidar points with random numbers in each grid, and from the distribution of the virtual lidar points within the search radius of lidar points included in each grid. Calculating a probability that can be assigned to a specific class; and if the probability is greater than a predetermined threshold, assigning the grid to a specific class. .

The present invention produces the numerical modeling input data of the quality which is significantly improved in accuracy and detail while the data construction time is significantly reduced compared to the numerical model data construction method performed by the existing manual work. This is because the aviation LiDAR misclassification results are filtered and the user can edit and output the final result in the process of applying the probability analysis algorithm in the process of lattling the classification error from the data. .

The method of generating numerical modeling input data based on the aviation lidar data has higher resolution and quality than the land cover data imported to generate numerical modeling input data by hand. It provides a method and system for solving the problems of building bias and accuracy of buildings and trees, which were identified as problems in the process of expanding the data production range in high resolution. To reduce and utilize redundant investment for unnecessary data production by sharing and managing the land cover map, which has been produced by various national institutions that need land cover information up to now, by sharing and managing the high quality land cover results produced through the method and system of the present invention. And the effectiveness of collaboration. Ultimately, the quality was improved through a new production system that summarized the manufacturing process of basic input data used for numerical modeling into scientific data and processes, thereby improving reliability in the process of verifying the analytical capability and predictive results through numerical modeling. .

1 is a view for explaining the classification process of the existing aviation LiDAR data.
2 is a view for explaining the surface classification method associated with the existing aviation LiDAR data.
3 is a view showing a classification result according to the surface classification method of the existing aviation LiDAR data.
4 is a block diagram of a grid-based high-resolution numerical modeling input data generation system to which a distribution probability analysis algorithm of aerial LiDAR data is applied according to an embodiment of the present invention.
5 is a block diagram illustrating a grid class analyzer 326 according to an embodiment of the present invention.
FIG. 6 is a diagram for describing an operation of the lattice class analyzer 326 of FIG. 5.
7 is a control flowchart illustrating an operation of a high resolution numerical modeling input data generation system according to an embodiment of the present invention.
8 is a flowchart illustrating a process in which a user optimizes and adjusts a distribution probability of a land cover class at an initial value in a state where a size of a grid is determined.
FIG. 9 is a diagram illustrating a process of producing final LC data with improved positional accuracy classification quality by combining land cover (LC) information generated from aerial lidar data with external LC information.
FIG. 10 is a diagram showing the results of systemically implementing a distribution probability analysis algorithm, editing and editing of grid data, and numerical modeling input data of aviation LiDAR data.

In the following description of the embodiments, detailed descriptions of related well-known functions or configurations will be omitted when it is determined that the detailed descriptions may unnecessarily obscure the subject matter of the present invention.

Hereinafter, with reference to the accompanying drawings will be described the configuration and operation of the present invention.

4 is a block diagram of a grid-based high-resolution numerical modeling input data generation system to which a distribution probability analysis algorithm of aerial LiDAR data is applied according to an embodiment to which the present invention is applied.

As shown in FIG. 4, the grid-based high-resolution numerical modeling input data generation system 300 to which the distribution probability analysis algorithm of the aviation LiDAR data to which the present invention is applied is a lidar data classification unit 302 and an input unit 310. , A processing unit 320 and an output unit 330. The LiDAR data classifier 302 classifies LiDAR point groups that enter each grid into LiDAR point groups by class, and provides the input unit 310 with LiDAR point groups classified by classes. In detail, the LiDAR data classification unit 302 extracts a point group corresponding to the ground surface from the non-classified LiDAR data through a roller technique, and classifies all the point groups on the ground surface as vegetation. Then, the Lidar data classification unit 324 classifies point groups of 50 cm or more from the ground as shrubs (Mid Veg.) From the data classified as vegetation. After classifying as, it is classified into point buildings which are assumed to be buildings based on the reflection characteristics and geometrical shape of aerial LiDAR data from the arbor point groups.

The lidar data classifier 302 is not an essential component of the present invention. The lidar points of the air lidar data may be classified into classes through the ground classification process in advance and provided to the grid-based high-resolution numerical modeling input data generation system 300.

The input unit 310 receives the aviation LiDAR data from the lidar data classifier 302 and provides the processing unit 320. The input unit 320 may also receive related data for performing a distribution probability analysis algorithm of the aviation lidar data according to the present invention, and if necessary, receive values set for analyzing the aviation lidar data from the user. have.

When the processing unit 320 receives the airborne LiDAR data from the input unit 310, the processing unit 320 performs a distribution probability analysis algorithm according to the present invention on the airborne LiDAR data. To this end, the processing unit 320 includes a grid-based data processor 322, a lidar data classifier 324, and a grid class analyzer 326.

The grid-based data processor 322 performs grid-based data processing on aviation lidar data, that is, aviation lidar data. In detail, the grid-based data processor 322 determines a spatial range of the air lidar data, that is, a boundary between an upper left and a lower right. In addition, the grid-based data processing unit 322 divides the airborne lidar data into grids having a predetermined size in a range where a column size does not exceed 250 in consideration of a spatial range of the airborne LiDAR data. Specifically, the grid-based data processor 322 determines through calculation so that the size of the grid has a maximum size between 1m and 20m, and expands the size of the determined grid horizontally and vertically from the grid on the upper left to both the LiDAR point group Determine the lattice matrix comprising a.

The grid class analyzer 326 lists lidar point groups for each lattice by class, and analyzes the lidar points for each lattice to allocate the lattice points. That is, the grid class analysis unit 326 lists lidar point groups included in each grid for each class, analyzes whether each grid represents a characteristic of the class, and performs class allocation for each grid. In this case, when one grid represents the characteristics of several classes, class allocation for the grid may be performed according to the priority among the classes.

Hereinafter, the operation of the grid class analyzer 326 will be described in detail.

을 계산한다. 격자 클래스 분석부(326)는 개개의 격자에 난수로 임의의 점들을 발생시켜 표본들 즉, 가상 라이다 점들을 확보하고 검색 반지름을

로부터 구하여 가상 라이다 점들이 각각의 항공 LiDAR 점의 검색 반경을 통해 들어올 확률을 개개의 클래스 별로 구한다. 구해진 개개의 클래스 확률 계산 결과로부터 50%를 임계치로 하여 각각의 클래스에 대한 격자 할당이 결정된다. First, the grid class analyzer 326 may represent an area represented by one aviation LiDAR point from the number of LiDAR points included in the entire LiDAR point group.

. The grid class analyzer 326 generates random points in each grid by random numbers to obtain samples, that is, virtual lidar points, and to determine a search radius.

From each class, the probability that the virtual Lidar points will come in through the search radius of each air LiDAR point is obtained. The lattice allocation for each class is determined with a threshold of 50% from the obtained individual class probability calculation results.

The configuration and operation of the grid class analyzer 326 will be described with reference to FIGS. 5 and 6.

5 is a block diagram of the grid class analyzer 326 according to an exemplary embodiment of the present invention, and FIG. 6 is a diagram for describing an operation of the grid class analyzer 326 of FIG. 5.

Referring to FIG. 5, as shown in FIG. 5, the lattice class analyzer 326 may include a LiDAR point area calculator 410, a virtual LiDAR point generator 420, a probability calculator 430, and a grid. The allocation unit 440 is included. Before describing the operation of the grid class analyzer 326, an existing grid class allocation method will be described with reference to FIG. 6 (a).

Referring to Fig. 6 (a), there is shown air laid data processed with grid data. Specifically, the case where the aviation LiDAR points are distributed in an arbitrary space (501) and the space is divided into 4 * 4 grids is shown.

Conventionally, when the number of points in each grid exceeds a threshold in a space where points classified into a specific class are distributed, the grid is determined as a specific class area. For example, if the number of four or more lidar points is set to the lattice allocation threshold, in Fig. 6, the five lattices 502 represented by the thick solid line are determined to be regions of a particular class.

The grid allocation technique using the conventional number adjustment is vulnerable to the phenomenon of clustering point groups in a grid. For example, in a two-row and two-column grid, the points are concentrated at the edges (503), and the number of points in the grid exceeds the threshold, that is, four, but the reference is not included in the class. As a result, it is hard to think that calculating only the number of points in the grid is a representative attribute of the grid. In addition, since the number of points entering each grid is changed each time the size of the grid is changed, it is inconvenient to change the threshold according to the size adjustment of the grid each time.

In order to solve this problem, the present invention applies an algorithm for determining whether to allocate a grid through a percentage ratio by calculating the points in the grid with a probabilistic analysis method. Specifically, the Lidar point only has position information of x, y, and z, and the size of the point is basically 0. However, by analyzing the area occupied by the points inside each grid with probability, we can calculate the probability of assigning the feature to each grid.

)을 계산한다. 구체적으로, 전체 격자 개수(501)를 항공 LiDAR 점들의 개수로 나누면 하나의 항공 LiDAR 점이 대표할 수 있는 면적이 계산된다. 하나의 항공 LiDAR 점이 대표할 수 있는 면적을 정확히 계산하기 위해 전체 격자로부터 항공 LiDAR 점이 하나도 없는 격자(504)는 계산에서 제외하였다. 하나의 점이 대표할 수 있는 면적을 계산하는 식은 다음 수학식 1과 같다.To this end, the lidar point area calculating unit 410 may represent an area represented by one aviation LiDAR point from the total number of LiDAR point groups.

). Specifically, when the total grid number 501 is divided by the number of aviation LiDAR points, an area that can be represented by one aviation LiDAR point is calculated. In order to accurately calculate the area that one aerial LiDAR point can represent, the grid 504 without any aerial LiDAR points from the entire grid was excluded from the calculation. The equation for calculating the area that can be represented by one point is shown in Equation 1 below.

은 점(point) 하나가 대표할 수 있는 면적이고,

는 도엽 내 전체 점의 수,

는 x축 격자수,

는 y축 격자수,

는 결측 격자 수 이다. here,

Is the area that a point can represent,

Is the total number of points in the lobe,

Is the number of grids in the x-axis,

Is the number of y-axis grids,

Is the number of missing grids.

Lidar point area calculation unit 410 The area that can be represented by one lidar point can be obtained, and the search radius (r) according to the area can be calculated and provided.

)을 대입하여 아래 수학식 2와 같이 구한다.The search radius (r) from the aviation LiDAR point is the area that one aviation LiDAR point can represent in the area (A) in the equation for obtaining the area of the circle.

) To obtain Equation 2 below.

을 갖는다.Accordingly, as shown in FIG. 6 (b), each Lida point is represented by the search radius 502 calculated from the representative area of the Lida point.

Has

Subsequently, the virtual lidar point generator 420 generates and distributes virtual lidar points, as shown in FIG. 6 (c). Virtual lidar points are randomly generated.

In addition, the probability calculator 430 estimates an acceptable probability of a specific class from the distribution of the points 512 in one grid 511 using Equation 3 below.

는 임의의 격자가 특정 클래스로 채택될 수 있는 확률이며,

는 격자 내부에 분포하고 있는 난수로 발생시킨 가상 라이다 점들로부터 각각의 항공 LiDAR 점 검색 반경 내부에 난수 발생된 가상 라이다 점들이 들어올 확률을 나타낸다. 그리고,

는 난수 발생을 통해 생성된 각각의 점이 각각의 항공 LiDAR 점들로부터의 검색반경을 통해 포함되는지 포함되지 않는지를 판단하는 함수로서 다음 수학식 4와 같다.here,

Is the probability that any lattice can be adopted as a specific class,

Denotes the probability that random randomly generated virtual lidar points from each of the aviation LiDAR point search radii are generated from random randomly distributed random points distributed within the grid. And,

Is a function for determining whether each point generated through random number generation is included or not included through a search radius from each air LiDAR point, as shown in Equation 4 below.

는 상기 각각의 가상 라이다 점들이 각각의 항공 라이다 점들로부터의 검색반경 내에 포함되는지 포함되지 않는지를 판단하는 함수로서, 난수 발생을 통해 생성된 가상 라이다 점이 검색반경 내부에 포함되는 것으로 판단되면 ‘1’ 값을 내보내고 검색반경 내부에 포함되지 않는 것으로 판단되면 ‘0’ 값을 내보낸다. 그리고,

,

는 LiDAR 점의 위치인 X축 좌표 및 Y축 좌표이고,

,

는 난수발생을 통해 랜덤(random)으로 격자 안에 들어가는 점의 위치인 X축 좌표 및 Y축 좌표이다. In other words,

Is a function for determining whether each of the virtual lidar points is included in the search radius from each of the air lidar points, and if it is determined that the virtual lidar points generated through random number generation are included in the search radius. If a value of '1' is exported and not determined to be included in the search radius, a value of '0' is exported. And,

,

Is the X-axis and Y-axis coordinates of the location of the LiDAR point,

,

Are X-axis coordinates and Y-axis coordinates, which are positions of points that randomly enter the grid through random number generation.

In other words, as shown in FIG. 6 (d), the probability calculator 430 determines the number and area of the virtual lidar points 516 located in the area 518 determined by the search radius of the lidar points in each lattice. Probability is calculated based on the number of virtual lidar points 515 located other than 518.

Subsequently, the grid allocator 440 assigns the grid to a specific class when the calculated probability is greater than a predetermined threshold.

As such, the present invention does not assume that the aviation lidar point is an arbitrary sample point, but on the contrary, the probability is obtained from the quantity occupied by the sample points included in the radial circle centered on the position of the lidar point, and the threshold is set. Determine class assignments within

Here, the grid class analyzer 326 provides the output of the grid assignment to the output unit 330. The output unit 330 visualizes and outputs the grid allocation result on the display. In this case, the output unit 330 may output the high-quality numerical modeling input data information according to the input structure of the numerical models to be used.

The operation of the high resolution numerical modeling input data generation system 300 is illustrated in FIG. 7.

7 is a control flowchart illustrating the operation of the high-resolution numerical modeling input data generation system 300 according to an embodiment of the present invention.

Referring to FIG. 7, the high-resolution numerical modeling input data generation system 300 first receives, in step 602, aeronautical lidar ground data, and acquires a boundary of lidar data in step 604, that is, aeronautical lidar ground data. Obtain the upper left and lower right of the.

The high resolution numerical modeling input data generation system 300 then determines the grid's starting position, grid size, and grid range (rows * columns) in step 606. Data related to the grid setting may be input from a user or set as a default. Data related to such a grid setting may change according to various conditions.

The high resolution numerical modeling input generation system 300 lists the lidar points entering each grid in step 608 after grid setting. Subsequently, the high resolution numerical modeling input data generation system 300 calculates an average area represented by one Lidar point from the total number of Lidar point groups in step 610. That is, the high resolution numerical modeling input data generation system 300 obtains a representative average area by dividing the total area of each grating by the total number of lidar points included in each grating. In this case, the grid without any aerial LiDAR points from the entire grid is excluded from the calculation to accurately calculate the area that can be represented by one aerial LiDAR point.

The high resolution numerical modeling input data generation system 300 then calculates a radius in step 612 to retrieve the points generated through random number generation from one Lidar point. In detail, the high resolution numerical modeling input data generation system 300 calculates a radius of a circular area having a representative average area of a Lidar point.

Subsequently, the high resolution numerical modeling input data generation system 300 randomly generates a virtual Lidar point in step 614. The high resolution numerical modeling input data generation system 300 generates a virtual lidar point, and then, in step 616, the virtual lidar points located in the area determined by the search radius of the lidar points for each lattice and the virtual positions located outside the area. Calculate the probability of representing the ratio of the number of Lidar points. In this case, the high resolution numerical modeling input data generation system 300 calculates the probability for the adoption of each grid to a specific class for the entire class.

The high resolution numerical modeling input data generation system 300 determines the probability of each grid, and then sets a threshold in step 618 to determine the class allocation of the grid. That is, if the probability exceeds a predetermined threshold, for example 50%, the class assignment is performed for the grid. The threshold may be changed by the user or by any condition.

In operation 620, the high resolution numerical modeling input data generation system 300 visually outputs a result of performing calculation according to an initial probability value. The grid resolution can then be changed in step 622 by the user or by some condition. The grid resolution may change if the grid size, or grid range, changes.

If the grid resolution is changed, the high resolution numerical modeling input data generation system 300 returns to step 606, sets up the grids, and recalculates the probability for adoption of each grid into a particular class.

As such, the present invention does not assume that the aviation lidar point is an arbitrary sample point, but on the contrary, the probability is obtained from the quantity occupied by the sample points included in the radial circle centered on the position of the lidar point, and the threshold is set. Determine class assignments within

8 is a flowchart illustrating a process in which a user optimizes and adjusts a distribution probability of a land cover class at an initial value in a state where a size of a grid is determined.

Referring to FIG. 8, the five classified land cover classification classes are allocated to the grid when the probability of a threshold value or more is displayed from the probability analysis result. That is, the initial threshold probability F is determined to be 50% to make the grid allocation (701). Then, visually inspect each class using image maps and other reference data from the results derived from the criteria assigned to the grid when showing probability above the threshold, and whether the classification performance needs to be adjusted by the user's probability threshold. Check (702). In this case, the user can directly correct and edit errors that cannot be corrected by the algorithm of the present invention to secure the numerical modeling input data quality based on the user's criteria.

If it is determined that the user does not need to adjust the probability threshold as a result of the initial evaluation (711), the result obtained through the initial probability is determined as the final probability threshold and the final result derived through the algorithm (704). When it is determined that the evaluation result 702 needs to adjust the probability threshold with the user (712), the thresholds for the five classes are adjusted between 0 and 100 through a grid control for controlling the user threshold (705). The adjusted threshold is reflected in the grid's allocation criteria through the View window (706). The reflected result is inspected based on the reference data (707). If the result of the adjustment does not optimize the classification result and it is estimated that additional user probability threshold adjustment is necessary (721), it returns to the user probability threshold adjustment window and recursively cycles until the threshold adjustment result evaluation is satisfied (722). At this stage, the View window must maintain the grid-specific assignment results until the calculation results are re-established. If the optimal threshold value is determined through the recursion process and no further threshold adjustment is required, the final result obtained through the algorithm is determined as the final result obtained through the algorithm (704).

FIG. 9 is a diagram illustrating a process of producing final LC data with improved positional accuracy classification quality by combining land cover (LC) information generated from aerial lidar data with external LC information.

Land cover codes required as data for numerical modeling are water (1), arborescent (2), other vegetation (3), pavement (4), and building (5) (801). The land cover grid data generated from the dual aeronautical lidar data uses only buildings (5) and arbor vegetation (2) without using low vegetation and shrub (Med. Veg) data. Treat as zero (802). Therefore, in order to generate the grid data including all the LC code information, the grid information of water (1), other vegetation (3) and pavement area (4) should be collected from externally verified grid data. Grid data coming from outside is usually LC grid data of the same resolution obtained from satellite images or from land cover or biotope thematic maps. The collection of data is performed via conditional statement 804. A query that finds non-tree vegetation (2) and buildings (5) grids from LiDAR LC (602) data is returned (LiDAR LC Code == 0). 806), and the lattice code of the false value (False) is replaced by the lattice code of the LiDAR LC (605) to generate the final LC (807) data.

The grid-based high-resolution grid data generated by applying the distribution probability analysis algorithm of aviation LiDAR data is output according to the generation protocol of numerical modeling input data of various sizes and purposes. If trees and buildings are used together in the same grid at the output stage, the buildings are first selected for output.

FIG. 10 is a diagram showing the results of systemically implementing a distribution probability analysis algorithm, editing and editing of grid data, and numerical modeling input data of aviation LiDAR data. By applying a file of aviation LiDAR data and storing the applied results as numerical modeling input data, menu 902 for setting the size of grid desired by user between 1m and 20m, and distribution probability analysis algorithm of LiDAR data are applied. A menu 903 that selectively turns classes on and off as a result of lattice, a menu 904 for determining probability thresholds for determining grid allocations for each class, and an edit menu for viewing and modifying information for each grid. 905, an edit menu that can collectively perform class conversion of all grids within a rectangular range defined by the user.

In the foregoing detailed description of the present invention, specific examples have been described. However, various modifications are possible within the scope of the present invention. The technical spirit of the present invention should not be limited to the above-described embodiments of the present invention, but should be determined by the claims and equivalents thereof.

302: lidar data classification unit 210: input unit
322: grid-based data processor 326: grid class analyzer
330: output unit

Claims (12)

A system for generating high resolution numerical modeling input data based on aviation lidar,
An input unit configured to receive air lidar data including LiDAR (Light Detection And Ranging) point cloudes classified by class;
The air lidar data including the lidar point group is divided into grids, a representative area of one lidar point is obtained from the number of lidar points included in the lidar point group, and a search radius is calculated according to the representative area. After calculating, generating virtual rider points with random numbers in each grid, and calculating the probability of assigning a specific class from the distribution of the virtual rider points within the search radius of the rider points included in each grid. And a processing unit for allocating the lattice to a specific class when the probability is greater than a predetermined threshold. The method of claim 1,
And an output unit configured to receive an allocation result from the processing unit to a specific class of the grids and to visualize and output the allocation result. The method of claim 1, wherein the processing unit
A grid-based data processor for determining a grid matrix by receiving the airborne lidar data, determining a spatial range of the airborne lidar data, and dividing the airborne lidar data into grids of a predetermined size;
After obtaining a representative area of one Lidar point from the number of Lidar points included in the Lidar point group, calculating a search radius according to the representative area, and generating virtual Lidar points with random numbers in each grid, A grid that calculates a probability that can be assigned to a specific class from the distribution of the virtual lidar points within the search radius of the lidar points included in each lattice, and assigns the lattice to a specific class if the probability is greater than a predetermined threshold High resolution numerical modeling input data generation system comprising a class analysis unit. The method of claim 3, wherein the grid class analysis unit
A rider point area calculator configured to calculate the representative area by dividing the number of grids by the number of rider points, and to calculate a search radius of the rider points based on the representative area;
A virtual lidar point generator for randomly generating the virtual lidar points;
Probability calculation unit for calculating a probability based on the number of the virtual lidar points located in the area determined by the search radius of the lidar points in each grid and the number of virtual lidar points located outside the area determined by the search radius Wow,
And a grid allocator for allocating the grid to a specific class when the calculated probability is greater than a predetermined threshold. The method of claim 4, wherein
The lidar point area calculator is a high-resolution numerical modeling input data generation system, characterized in that for calculating the representative area according to the following equation.

here,

Is the area that a point can represent,

Is the total number of points in the lobe,

Is the number of grids in the x-axis,

Is the number of y-axis grids,

Is the number of missing grids. The method of claim 4, wherein
The probability calculator is a high-resolution numerical modeling input data generation system, characterized in that for calculating the probability according to the following equation.

here,

Is the probability that any lattice can be adopted as a specific class,

Denotes the probability that the virtual lidar points enter the search radius of each aerial lidar point from the virtual lidar points generated by the random numbers distributed in the grid,

Is a function for determining whether each of the virtual lidar points is included within the search radius from each of the air lidar points. The method of claim 1,
And the predetermined threshold value can be changed by a user. A method for generating high resolution numerical modeling input data performed by a system for generating aviation lidar based high resolution numerical modeling input data,
Receiving air lidar data including LiDAR (Light Detection And Ranging) point cloudes classified by class of the input unit of the system,
A grid-based data processing unit of the system dividing aviation lidar data including the lidar point group into grids;
Calculating a search radius according to the representative area by obtaining a representative area of one lidar point from the number of lidar points included in the lidar point group by the lidar point area calculating unit of the system;
Generating, by the virtual lidar point generator, the virtual lidar points in a random number in each grid;
Calculating a probability that the probability calculator of the system can assign a specific class from the distribution of the virtual LiDAR points within a search radius of the LiDAR points included in each grid;
And allocating the grid to a specific class if the grid allocation unit of the system is greater than the predetermined threshold. The method of claim 8,
And outputting, by the output unit of the system, an assignment result of the grid to a specific class and visualizing and outputting the assignment result. The method of claim 8,
The representative area is a high resolution numerical modeling input data generation method, characterized in that calculated according to the following equation.

here,

Is the area that a point can represent,

Is the total number of points in the lobe,

Is the number of grids in the x-axis,

Is the number of y-axis grids,

Is the number of missing grids. The method of claim 8,
High probability numerical modeling input data generation method, characterized in that the probability is calculated according to the following equation.

here,

Is the probability that any lattice can be adopted as a specific class,

Denotes the probability that the virtual lidar points enter the search radius of each aerial lidar point from the virtual lidar points generated by the random numbers distributed in the grid,

Is a function for determining whether each of the virtual lidar points is included within the search radius from each of the air lidar points. The method of claim 8,
The predetermined threshold value may be changed by the user, high resolution numerical modeling input data generation method.

Images