KR100973052B1 - Automatic matching method of digital aerial images using lidar data - Google Patents

Abstract

PURPOSE: An automatic matching method of digital aerial Images using LIDAR data is provided to supply high accurate 3D spatial information by generating a DSM through an accurate ground point of a radiator and the ground point extracted from an image-matched aerial image. CONSTITUTION: An image point indicating a ground point is detected from a duplicated aerial image, and the accuracy of image matching is improved by using virtual LIDAR(Light Detection And Ranging). Through the extraction of vertical error of the ground point, the image matching is verified(S60). Through the matched aerial images and LIDAR data, a DSM(Digital Surface Model) is created(S30). The radiator is collected(S40), and the DSM is updated(S50).

Description

Automatic matching method of digital aerial images using LIDAR data

The present invention relates to an automatic matching technique for digital aerial imagery using lidar data, and more particularly, to increase the precision of extracting the corresponding conjugate point initial value for matching using the lidar data and to increase the convergence speed of the corresponding conjugate point true value. It is a high-speed and reliable matching technique. Also, it is applied to the automatic matching technique of digital aerial image using LiDAR data, which raises the reliability of image matching by further correcting the external marking elements of each aerial image using LiDAR data before image matching. It is about.

Recently, in order to effectively and quickly cope with natural and urgent disasters occurring on the ground, the demand for high-resolution three-dimensional spatial information about the target area, that is, DSM (Digital Surface Model) or orthoimage is increasing.

In order to provide such high-resolution three-dimensional spatial information quickly and with high reliability, the matching between aerial images must be precise and rapid.

Several methods have been proposed to date for image matching methods, but high speed and reliability methods have not been developed.

This is basically because of the errors of the equipments used to acquire aerial images, the errors coming from the interrelationship of the equipments, and the errors between the acquired images, that is, geometrical errors and the slope or ups and downs of the area where the aerial images are taken. The accuracy, i.e., the reliability, is lowered due to the reason, and the inaccuracy and the convergence process of the initial value extraction of the corresponding conjugate point when matching by the least-squares matching method are caused by the large number of iterations.

However, until now, it has not been suggested to solve these problems and to provide fast and reliable aerial image matching technique.

The present invention has been made to solve the above problems, using the lidar data to minimize the geometric inconsistency (error), and by using the rapid and reliable lidar data by minimizing the accuracy and number of repetition of the conjugate point extraction The purpose of this paper is to provide an automatic matching technique for digital aerial imagery.

In order to achieve the above object, the automatic matching technique of digital aerial image using LiDAR data of the present invention

In the method of matching the digital aerial images of the specific region by using the LiDAR data for the specific region,

Extracting an initial value of a corresponding conjugate point of a matched image with respect to a conjugate point of a reference image using the LiDAR data;

And gradually extracting the true value of the corresponding conjugate point by gradually changing the corresponding conjugate point of the extracted initial value.

And extracting an initial value for the corresponding conjugate point

A plane patch generation step of generating a plurality of plane patches interconnecting irregular three-dimensional ground points of the LiDAR data;

A projection point extraction step of extracting a projection point at which the conjugate point of the reference image is projected onto the plane patch;

And extracting a corresponding conjugate point initial value extracting a corresponding conjugate point to which the extracted projection point is projected onto the registration image.

Extracting the true value of the corresponding conjugate point

The true value of the corresponding conjugate point is extracted by gradually changing the corresponding conjugate point of the set initial value using a least square matching method.

Before extracting the initial value for the corresponding conjugate point

And a geometric matching step of correcting an external expression element of each aerial image by geometric matching of aerial images and rider data for a specific region.

The geometric matching step

A ground point extraction step of selecting a conjugate point and a corresponding conjugate point in each of the two aerial images, and extracting a corresponding ground point by substituting coordinates and external expression elements of the selected conjugate point and corresponding conjugate point in a collinear condition equation;

A plane patch generation step of generating a plurality of plane patches interconnecting irregular three-dimensional points of the LiDAR data;

Selecting an external planar patch adjacent to the extracted ground point, and correcting the external expression element by gradually changing the external expression elements of the two aerial images such that the ground point is located on the selected plane patch. Characterized in that made,

And selecting a plurality of conjugated points-corresponding conjugate points of the two aerial images to extract a plurality of corrected external expression elements, and finally correcting the average of the extracted external expression elements to the external expression elements of the corresponding aerial image.

In the external expression element correction step

It is characterized in that the external expression element is corrected only for the [conjugated point-corresponding conjugated point] in which the extracted ground point and the selected planar patch are within a certain distance.

According to the present invention having the above-described configuration, the geometric distortion (error of an external expression element) of aerial images is corrected using lidar data, and the image registration of the aerial images whose geometric distortion is corrected is not a virtual horizontal plane. Since the image is matched using the plane patch generated from the data, the image matching is high, and the DSM is generated using both the ground point extracted from the precisely matched aerial images and the precise ground point of the LiDAR data. The DSM generated by the DSM provides three-dimensional spatial information with high density and high precision.

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

1 is a flowchart illustrating a method of generating a DSM using aerial video and lidar data according to the present invention.

As shown in the figure, the geometric matching step (S10), the image matching step (S20), the DSM generation step (S30), and further comprises a LiDAR data correction step (S40) and DSM update step (S50). can do.

The geometric matching step S10 is a step of minimizing geometric inconsistency (error) between the aerial image and the lidar data. More specifically, by correcting the external expression elements of aerial images, geometric errors of equipment related to aerial imagery and equipment related to lidar data are reduced.

Geometric matching refers to matching between aerial imagery and rider data. This requires a matching element. The registration element of the image is the conjugate point, and the registration element of the LiDAR data is a plane patch.

Generally, digital cameras, laser scanners, GPS, and INS (or IMU) are attached to manned or unmanned aerial vehicles to acquire aerial images and rider data. The digital camera photographs the ground to acquire an aerial image, the laser scanner irradiates the laser to the ground and receives the reflected laser to acquire the rider data. The GPS and the INS (or IMU) are used for the digital camera and the laser scanner. It is used to obtain information about position and posture (shooting direction and irradiation direction).

The geometrical inconsistency between the aerial images and between the aerial images and the rider data is caused by the individual errors of the four equipments or the errors caused by the interrelationships of the four equipments. Therefore, it is necessary to minimize the geometric inconsistency between them to increase the accuracy of matching between aerial images and to obtain more precise coordinates of ground points from the matched images.

Since the Lidar data is a high-precision ground point coordinate, the internal focusing elements (the focal length of the camera lens, the alignment misalignment between the lens and the CD) and the external marking elements (the positional coordinates of the camera and the attitude of the camera) are taken. Coordinates) so that their geometric inconsistency is minimized. In particular, the correction of external expression factors is important.

Correction of the external expression element, that is, geometric matching step (S10) is the ground point extraction step (S12) from the aerial image, the plane patch generation step (S13) from the rider data, the step of gradually changing (correcting) the external expression element ( S15) is made.

In order to extract the ground point, the conjugate point and the corresponding conjugate point are respectively selected in the overlapped portions of the two aerial images (S11). Here, the conjugate point and the corresponding conjugate point are pixel coordinates of the aerial image in which the two images photograph the same point on the ground, and the conjugate point and the corresponding conjugate point select a point that is easily distinguished from other parts of the aerial image. By selecting a point that is easily identified, that is, an embossed point as a [conjugated point-corresponding conjugated point], the matching (matching) of these points does not deviate a few pixels, so that the coordinates of the ground point extracted are more accurate. And although only one [conjugated point-corresponding conjugated point] may be selected, it may be desirable to select several and perform subsequent steps so that more accurate correction of the external expression element is performed. And you can choose the Conjugated-corresponding conjugate point by hand, but for the sake of speed, if one pixel is noticeably different in RGB or Gray values compared to other pixels around it, or is clearly distinguished by other factors The pixel may be automatically selected as a conjugate point and a corresponding conjugate point.

After selecting [conjugated point-corresponding conjugate point] from the aerial image (S11), the ground point of the ground level is extracted by substituting the coordinates and external expression elements of the corresponding aerial image into the collinear condition equation.

After the ground point is extracted or before or simultaneously with the extraction, a plurality of plane patches are generated from the LiDAR data (S13). Here, the planar patch means a plane formed by interconnecting irregular three-dimensional ground points constituting LiDAR data.

Set up the adjacency of irregularly distributed ground points, create an initial small plane patch from a few clustered ground points, and gradually include the surrounding ground points to create the remaining plane patches. do. Each planar patch does not overlap with each other, the boundaries abut other planar patches, and the planes have plane approximation errors. The simplest flat patch is a form in which triangular planes (ie, flat patches) connecting three ground points are connected at an angle to each other in series.

After the ground point is extracted (S12) and the plane patch is generated (S13), the plane patch closest to the extracted ground point is selected (S14), and the external expression of the two aerial images is positioned on the selected plane patch. The external expression element is corrected (S15) by gradually changing the element.

In more detail,

In ideal cases (i.e. no geometric inconsistency, no inclination or undulation on the ground, the selection of corresponding points is correct, and the planar patch coincides with the ground) the extracted ground point will be located exactly on the planar patch. Since it can not be as shown in Figure 2 the extracted ground point is out of the plane patch. Thus, the geometric inconsistency is reduced by gradually adjusting the external expression elements of the two aerial images so that the extracted ground point is located on the plane patch.

The gradual adjustment (change) of the external expression elements is controlled by the convergence equation, giving the constraint that the ground point and the ground point exist on the plane patch, indicating the relationship between the ground point and the coordinates (ie, conjugate point) of the aerial image. An example of the convergence equation is an observation equation used for the minimal manufacturing method in the image matching step (S20).

Briefly explaining the adjustment of external expression elements through convergence equations, the new external expression elements are obtained by substituting the initial external expression elements of the two aerial images corresponding to the convergence equation, and substituting the obtained external expression elements into the convergence equation again. It is repeated to find again, so that the external expression element obtained when one parameter of the convergence equation converges to a specific value is the final corrected (adjusted) external expression element.

Rather than selecting only one [conjugated-corresponding conjugated point] to calibrate an external expression element, select multiple [conjugated-corresponding conjugated points] to obtain a number of corrected external-expressed elements, and average the final correction of the corresponding aerial image. It may be more desirable to use an external expression element. In addition, it may be desirable to obtain an average of the external expression elements that deviate from the threshold by applying a standard deviation among a plurality of external expression elements corrected by selecting several [conjugated points-corresponding conjugate points], and before correction of the external expression elements (S15). If the distance between the extracted ground point and the selected plane patch is out of a certain range, there is a big error in the selection of [conjugated point-corresponding conjugate point] or a large error in the ground point of the rider data forming the plane patch. It may be desirable not to correct the external expression element.

The image matching step (S20) is a process of finding an image point (that is, a conjugate point) representing the same ground point from overlapping aerial images. Lidar data has a reflection characteristic that it is difficult to accurately obtain information on a corner or corner of a building. It is a process to obtain dense three-dimensional ground coordinates (ie, DSM) that are necessary to complement the problem.

Image matching is based on the similarity of the brightness values to the local area (Aerial Based Matching), the shape based method based on the similarity of the shape between the extracted linear objects (Feature Based Matching), between the extracted objects Many methodologies have been proposed, which are classified as Relational Matching.

However, a method suitable for real-time spatial information generation, that is, a high speed and high reliability has not yet been developed. This is because the proposed methodologies mainly consider only multiple images as inputs, but even if the images were acquired for the same area, the projected images on the images could be changed by the difference of observation point or the slope or undulation of the surface. Because it is difficult to find.

The domain-based method is represented by correlation matching or least squares matching, and in order to obtain a fast and reliable matching result, the search range for finding a matching target point (that is, corresponding conjugate point) is maximized. You should reduce it and set the initial position for the search to be within 2 or 3 pixels of the actual registration position.

The least square matching method copies a reference region from one image (i.e., reference image) to a reference region of a constant size based on the center of the selected matching object (i.e., conjugate point) and another image (i.e., registration image). It is a method of estimating a transformation coefficient such that the sum of squares of the difference in brightness values between the matched regions generated by the mathematical and geometric transformation is minimized. In the present invention, the least square matching method is used as the matching method, and Lidar data is used.

)과 복사학적 변환모델(

)에 포함되는 각각의 미지수

와

을 추정하기 위해 수립되는 관측방정식은 수식1과 같다. 이는 하나의 영상(t)에서 i,j에 위치한 하나의 픽셀의 밝기값t(i,j)은 두 번째 영상(m)에서 이에 대한 기하학적 변환을 통해 대응되는 위치

(i,j)의 픽셀의 밝기값 m(

(i,j))을 복사학적 변환을 통해 변환한 밝기값

(m(

(i,j)))과 동일해야 한다는 것을 바탕으로 관측방정식으로 수립한 것이다. 여기서, e(i,j)는 밝기값의 차이를 나타내는 오차로, 이를 최소화하도록 최소제법을 적용한다. Based on the basic principle of least square matching,

) And the radiological transformation model (

) Each unknown in

Wow

The observation equation established to estimate is This means that the brightness value t (i, j) of one pixel located at i, j in one image t corresponds to the position of the second image m through geometric transformation thereof.

The brightness value m of the pixel of (i, j)

Brightness value converted from (i, j)) by radiological conversion

(m (

It is based on the fact that (i, j))) must be the same. Here, e (i, j) is an error representing the difference in brightness values, and a minimum method is applied to minimize this.

[Equation 1]

Since observation equations such as Equation 1 are generally nonlinear, the linearization of each unknown is required to apply the least square method. For linearization, an initial value for the unknown is required, and the closer the initial value is set to the true value, the closer the estimated value for the unknown converged by the least square method is to the true value. Therefore, the process of finding the initial value in the least-squares matching is a very important process that has a significant effect on the matching result.

와

는 각각 미지수

와

에 대한 초기값을 의미하고,

이고,

이다. 또한, 미분값

는 chain rule에 따라서 수식3처럼 계산된다. Equation 2 is an observation equation linearized with Equation 1, where

Wow

Are each unknown

Wow

The initial value for,

ego,

to be. In addition, the derivative value

Is calculated as Eq. 3 according to the chain rule.

[Equation 2]

[Equation 3]

는

로서 미지수와 미지수에 대한 초기값의 차이를 나타낸다. One observation equation such as Equation 2 can be established for individual pixels in the reference area. Equation 4 is obtained by collecting observation equations for all pixels and surface them as a vector equation. Here, y represents the left side of Equation 2 for all pixels as a vector, A represents the derivative of Equation 2 for all pixels as a matrix, and e represents the difference in brightness values of all pixels. Expressed as a vector of

Is

As the difference between the unknown and the initial value for the unknown.

[Equation 4]

에 대한 추정값 d

를 수식5처럼 계산하고, 이를

에 더하여 수식6처럼

에 대한 추정값을 구한다. 추정값

를 다시 초기값으로 설정하고, 이에 대해 선화화해서 수식5 및 수식6의 과정을

이 수렴될때까지 반복적으로 수행한다. Applying least square method to Equation 4

Estimate d for

Is calculated like Equation 5, and

In addition to Equation 6

Obtain an estimate for. Estimate

Is set to the initial value again, and the linearization is performed to repeat the process of Equation 5 and Equation 6.

Iterate until it converges.

[Equation 5]

[Equation 6]

The unknown of the least-squares matching method consists of coefficients related to geometric transformations and coefficients related to radiative transformations. In general, radiative transformations use a linearized transformation model of brightness values, and the associated coefficients are the overall brightness values of two images. Determine by comparing the difference. In many cases, the linear transformation model determined by the preprocessing minimizes the difference in brightness due to radiological causes and does not estimate the coefficients of the radiation transformation model individually through the least squares matching method. Finally, the coefficients related to geometric transformations are unknown and estimated using the least-squares matching methodology.

If so, the initial values for the unknowns associated with the geometric transformation should be set as precisely as possible so that the initial position of the corresponding conjugate point calculated by the transformation falls within a few pixels of the true value.

The existing method of determining the initial value of the corresponding conjugate point is generally to assume the ground as a horizontal plane with an average elevation for the target area. However, since this assumed horizontal plane has a lot of errors from the real world, the initial value of many corresponding conjugate points is out of the range of 2 or 3 pixels that can converge to the true value.

Accordingly, in the image matching step (S20), by using the LiDAR data instead of the virtual horizontal plane, the difference between the initial value and the true value of the corresponding conjugate point significantly reduced the success rate (precision) of the image registration. This will be described in more detail below.

First, a reference image and a matched image overlapping the reference image are selected from the aerial image. 3 illustrates the principle of an image registration method according to the present invention.

Projection points are extracted by sequentially projecting each of the pixels (ie, conjugate points) of the overlapped portion of the reference image to the planar patch of the LiDAR data (S21). At this time, the extraction of the projection point uses a ray tracing algorithm. Since the ray tracing algorithm is well known in the art, a detailed description thereof will be omitted.

When the projection point is extracted, the external reflection elements (that is, the position of the camera and the photographing direction of the camera) of the reference image are corrected, so that the extracted projection point is more accurate.

After extracting the projection point to the plane patch (S21), the projection point is projected again to the registration image to extract the initial value of the corresponding conjugate point (S22). The extraction of the corresponding conjugate point is obtained by substituting the coordinates of the projection point extracted in the collinear condition equation and the corrected external expression element of the registered image.

After extracting the initial value of the corresponding conjugate point (S23), a new corresponding conjugate point is extracted by substituting the initial value of the corresponding conjugate point extracted into the observation equation of the least square matching method, and the newly extracted corresponding conjugate point is set again as the initial value. Substitute the observation equations and extract new corresponding conjugate points. Repeat this step to gradually adjust the new corresponding conjugate point to extract the true value (S25).

This process is applied to all pixels of the reference image of the overlapped portion to match the aerial images.

The result of image registration obtained through the above process is not always successful due to various causes. Therefore, a matching verification step (S60) is required to verify the quality of the image matching result to confirm that the corresponding conjugate contact corresponding to each conjugate point generated through matching is correctly selected.

As a method of verifying the quality of the image registration result, the method of comparing the initial process of the corresponding conjugate point to the true value using the least square matching method, that is, how many times the repetition is performed, A method of comparing the estimated values of the variance elements representing the average difference in brightness values of the local areas around the corresponding conjugate points and a method of comparing the standard deviation of the corresponding conjugate point estimates (initial values) are used.

In addition to the above three methods, the present invention further proposes a method for inspecting the quality of the ground point and the vertical distance of the Lidar data plane patch obtained from the initial values of the conjugate point and the corresponding conjugate point.

If the geometric registration (ie correction of external expression elements) and image registration are accurate, the ground point obtained as the initial value of the conjugate point and the corresponding conjugate point will be very close to the flat patch. Assuming that the errors included in the LiDAR data and the corrected external expression elements are not large, the vertical error of the ground point can be used to verify the quality of the image registration result because it can be seen that it directly represents the reliability of the registration result.

In the registration verification step (S60) of verifying the work due to the vertical distance error, a step of extracting a corresponding ground point by substituting the initial value light external expression element of the conjugate point and the corresponding conjugate point into the collinear condition equation (S61) and the plane of the lidar data Selecting a plane patch close to the ground point extracted from a plurality of patches (S62), and extracting the vertical distance between the extracted ground point and the selected plane patch (S63). After extracting the vertical distance for all conjugate points, the quality of the match is determined by how many conjugate points the vertical distance goes beyond the reference distance.

A, b, and c of FIG. 4 are Lidar data, aerial image reference image, and registration image of Jeju city,

FIG. 5 is a comparison graph of a result of image registration using a conventional general virtual horizontal plane with FIG. 4 and a result of image registration using LiDAR data according to the present invention, and FIG. 5A is a graph showing a true value of a corresponding conjugate point. 5b is a comparison of the estimates of the distribution components, and FIGS. 5c and 5d are comparisons of standard deviations in the x and y directions of the corresponding conjugate point estimates, respectively. The vertical distance between the ground point and the virtual horizontal plane / plane patch is compared.

As shown in FIG. 5A, the present invention converges at a high speed, but the existing method is not only slow in convergence but also 6 times out of 22 target points when not converged in 100 iterations, as shown in FIGS. 5B, C, and D. As shown in FIG. 5E, the present invention shows a good result (quality) because the standard deviation of the estimated value and the estimated value of the variance factor is lower than that of the conventional method. In the previous method, only 6 out of 22 target points showed good results, and the rest failed matching.

After the image matching step S20 is completed, a digital elevation data (DSM), which is a three-dimensional spatial coordinate of a corresponding region (ie, a specific region) is generated using the matched aerial images and the LiDAR data (S30). .

The DSM generation step S30 includes a ground point extraction step S31, a grid generation step S32, an altitude value extraction step S33, and an altitude value correction step S34.

The ground point extraction step (S31) extracts the ground point by substituting the conjugate point of the reference image and the coordinate of the final value of the corresponding conjugate point of the extracted registration image and the corrected external expression element in the collinear condition equation. The ground points are extracted for all aerial images of the entire specific area.

The grid generation step (S32) divides the specific area at a predetermined interval, and includes both extracted ground points and irregular ground points of LiDAR data, and all grids should be assigned one or more ground points. And the spacing of the grid is appropriately set in consideration of the point density of the ground points.

The altitude value extracting step (S33) for each lattice obtains an average of altitude values for ground points existing within a predetermined distance from the center of each lattice, and assigns the altitude value to the lattice. Here, the predetermined distance will be at least half of the lattice length, preferably not more than four times the maximum lattice length.

In the altitude correction step (S34), a standard deviation is obtained from the extracted altitude values of the respective grids and the altitude values of the corresponding ground points to remove ground points outside the threshold, and the altitude values of the grids are extracted again from the remaining ground points. do. Thus, the precision of the altitude value is increased.

Through this process, the altitude value is assigned to each grid, and a DSM for a specific area is generated. That is, each lattice has a planar positional coordinate and a three-dimensional spatial coordinate (X, Y, Z) since the altitude value is assigned. And the DSM generated up to this point has considerable precision.

However, since the coordinates of the ground points of the LiDAR data are precise, the LiDAR data is composed of three-dimensional ground points distributed irregularly. Is inaccurate and has limitations.

Accordingly, the present invention further includes the step of correcting the lidar data (S40) and the step of updating the DSM (S50).

Since the DSM generated in step S30 has high-density and high-precision position coordinates and an altitude value, the ground point of the DSM is given to an area in which there is no ground point within a predetermined distance from the Lidar data (S40). Attention should be given to the provision of too many ground points, which can lead to distortion of LiDAR data.

After the plane patch is newly generated with the corrected LiDAR data (S13), steps S10 to S30 are performed to update the DSM (S50) to finally generate a higher precision and higher density DSM.

In the above description of the present invention has been described with respect to the automatic matching technique of the digital aerial image using the lidar data having a specific procedure with reference to the accompanying drawings, the present invention can be variously modified and changed by those skilled in the art, such modifications And modifications should be construed as falling within the protection scope of the present invention.

1 is a process flow diagram of a matching technique in accordance with the present invention.

2 is a view showing a method of correcting an external expression element using lidar data.

3 is a view showing a method of matching two aerial images using LiDAR data.

4 a, b, and c are rider data of Jeju city and two aerial images.

5 a, b, c, d, and e is a diagram comparing the quality of the image matched by the existing method and the present invention with FIG.

<Description of the symbols for the main parts of the drawings>

S10: geometric matching step S20: image matching step

S30: DSM generation method S40: Lidar data correction step

S50: DSM update step S60: Registration verification step

Claims (7)

In the method of matching the digital aerial images of the specific region by using the LiDAR data for the specific region, Extracting an initial value of a corresponding conjugate point of a matched image with respect to a conjugate point of a reference image using the LiDAR data; Gradually changing the corresponding conjugate point of the extracted initial value to extract the true value of the corresponding conjugate point; and automatic matching technique for the digital aerial image using Lidar data. The method of claim 1, wherein extracting an initial value for the corresponding conjugate point A plane patch generation step of generating a plurality of plane patches interconnecting irregular three-dimensional ground points of the LiDAR data; A projection point extraction step of extracting a projection point at which the conjugate point of the reference image is projected onto the plane patch; And a corresponding conjugate point initial value extracting step of extracting a corresponding conjugate point from which the extracted projection point is projected onto the registration image. The method of claim 1, wherein extracting the true value of the corresponding conjugate point And a true value of the corresponding conjugate point is extracted by gradually changing the corresponding conjugate point of the set initial value using a least square matching method. 4. A method according to any one of the preceding claims, wherein prior to the step of extracting an initial value for the corresponding conjugate point. And a geometric matching step of correcting an external expression element of each aerial image by geometric matching of aerial imagery and lidar data for a specific region. The method of claim 4, wherein the geometric matching step A ground point extraction step of selecting a conjugate point and a corresponding conjugate point in each of the two aerial images, and extracting a corresponding ground point by substituting coordinates and external expression elements of the selected conjugate point and corresponding conjugate point in a collinear condition equation; A plane patch generation step of generating a plurality of plane patches interconnecting irregular three-dimensional points of the LiDAR data; Selecting an external planar patch adjacent to the extracted ground point, and correcting the external expression element by gradually changing the external expression elements of the two aerial images such that the ground point is located on the selected plane patch. Automatic matching method of digital aerial image using Lidar data, characterized in that is made. The method of claim 5, In the ground point extraction step, a plurality of ground points corresponding to each [conjugation point-corresponding conjugate point] are extracted by selecting a plurality of conjugate points and corresponding conjugate points (hereinafter, referred to as '[conjugation point-corresponding conjugate point]') in two aerial images. The external expression element correction step corrects the external expression element by using each ground point extracted in the ground point extraction step, and averages the external expression elements corrected by each ground point. Automatic matching technique for digital aerial imagery of Lidar data, characterized in that for correcting. The method of claim 6, wherein the external expression element correction step An automatic matching method of digital aerial image using Lidar data, characterized in that the external expression element is corrected only for the extracted conjugate point and the selected plane patch within a predetermined distance.

Images