KR100855657B1 - System for estimating self-position of the mobile robot using monocular zoom-camara and method therefor - Google Patents

Abstract

The present invention relates to a system and method for estimating a magnetic position of a mobile robot using a monocular zoom camera, and to obtain a zoom-out (original) image and a zoom-in image obtained by the monocular zoom camera according to the position movement of the mobile robot, respectively. Determining scale invariant feature points and performing matching between the feature points, constructing a feature map by estimating the relative position of the mobile robot by reconstructing the matched feature points in three dimensions, and then estimating its own position. It is a technique to do.

Mobile Robot, Self Location Estimation, Feature Point, Mapping, Map Database

Description

System and method for magnetic position estimation of mobile robot using monocular zoom camera {SYSTEM FOR ESTIMATING SELF-POSITION OF THE MOBILE ROBOT USING MONOCULAR ZOOM-CAMARA AND METHOD THEREFOR}

1 is an overall configuration diagram of a magnetic position estimation system of a mobile robot using a monocular zoom camera according to an embodiment of the present invention.

2 is a flowchart illustrating a method of estimating a magnetic position of a mobile robot using a monocular zoom camera according to an embodiment of the present invention.

Figure 3 is an exemplary view showing the matching of the feature points in the zoom-in / zoom-out image of the method for estimating the magnetic position of the mobile robot using a monocular zoom camera according to an embodiment of the present invention.

Figure 4 is a flow chart showing the detailed steps of the map construction process in the magnetic position estimation method of the mobile robot according to an embodiment of the present invention.

5 is an exemplary diagram of a camera model used in a magnetic position estimation system of a mobile robot using a monocular zoom camera according to an embodiment of the present invention.

6 is a flow chart showing the detailed steps of the absolute position detection process in the magnetic position estimation method of the mobile robot according to an embodiment of the present invention.

The present invention relates to an apparatus and method for estimating a magnetic position of a mobile robot using a monocular zoom camera, wherein each image is obtained by using a zoom-out (original) image and a zoom-in image obtained by the monocular zoom camera according to the movement of the robot. It is a technology that detects scale invariant feature points, performs matching between feature points, constructs a feature map by estimating the relative position of the robot by reconstructing the matched feature points in three dimensions, and then estimates the position of the robot later. .

In the prior art related to the present invention, a method of extracting feature points or obtaining distance information for two front images of a robot using a stereo camera attached to a robot has been proposed. The error accumulates and the range of the error increases in magnetic position estimation when applied to a robotic system. In the case of the stereo camera, an expensive image acquisition device additionally needs to be added to synchronize images or acquire images. In addition, the setting for matching two camera lens center points has a difficult problem.

In addition, recently, a method of estimating a relative position by matching feature points with two images based on a distance difference according to the movement of the robot using a monocular camera has been proposed, but in this case, the robot needs to obtain two images. Acquisition time is delayed because other images are superimposed while moving, and accordingly, there is a problem that many error factors occur when acquiring images due to posture distortion or terrain imbalance due to robot movement.

An object of the present invention, which was devised to solve the above-described problem, is to detect a scale invariant feature point for each image by using a zoom-out (original) image and a zoom-in image obtained by a monocular zoom camera according to the movement of a robot. An apparatus for estimating a magnetic position of a mobile robot using a monocular zoom camera capable of obtaining a real distance between the mobile robot and the feature point by performing matching between feature points and deriving the real world coordinate points of the matched feature points in three dimensions. To provide a way.

Another object of the present invention, by deriving a specific magnification for the zoom-out (original) image and the zoom-in image obtained by the monocular zoom camera, it is not affected by the error factors that may occur due to the movement of the mobile robot when the two images are acquired. The present invention provides an apparatus and method for estimating a magnetic position of a mobile robot using a monocular zoom camera.

It is still another object of the present invention to construct a feature map using feature points for zoom-out (original) and zoom-in images acquired through a monocular zoom camera, and use rotation and movement parameters for the feature points. The present invention provides an apparatus and method for estimating a magnetic position of a mobile robot using a monocular zoom camera that allows the robot to estimate its absolute position.

According to an aspect of the present invention, there is provided an apparatus for estimating a magnetic position of a mobile robot using a monocular zoom camera, wherein the camera apparatus obtains image information about a mobile robot working environment, and detects feature points by using image information acquired through the camera apparatus. And a main processing device for deriving the position of the mobile robot using distance information between feature points, and a serial communication device.

Referring to the above-described components in detail with reference to Figure 1 as follows.

The camera device (or monocular zoom camera 100) performs a function of acquiring images of the work area of the mobile robot, and the serial communication device 200 includes the camera device 100 and the main processing device ( It performs a function as an information transmission path between 300, in another embodiment of the present invention, the serial communication device may use a USB interface device.

The main processing apparatus 300 stores the images initially received from the camera apparatus as a reference image, and constructs a map database by deriving feature points, matching points, and relative position and attitude information of the mobile robot. And comparing the image received from the camera device with the reference image in the constructed map database while driving the mobile robot to derive the relative position, the absolute position, and the attitude information of the mobile robot.

In other words, the main processing unit 300 constructs a map database as an initial setting operation for deriving a position and attitude according to the movement of the mobile robot. The image is zoomed out / zoomed through the camera device while driving the mobile robot. It stores the image data, the feature point of each image, the relative coordinate data of the feature point and the mobile robot posture information of the time point at which each image is obtained to build a map database for the work area of the mobile robot.

The main processing apparatus 300 for performing the above-described function includes an image receiver 302 for receiving a zoom-out / zoom-in image from the camera apparatus 100 and a specific magnification of the camera apparatus from the received images. A specific magnification detector 304, an internal parameter calculator 306 for calculating internal parameters for camera calibration using a zoom-out / zoom-in image received from a camera apparatus, and a zoom-out / zoom-in image using a Gaussian difference operator A feature point detector 308 that detects the scale invariant feature points of the frame and generates a feature point set F _i and F _{i +1} , and an external parameter calculator that calculates an external parameter representing the rotation and translation of the real world coordinate and the camera coordinate system. And the feature point from the four linear system of equations generated using the calculated external and internal parameters. A feature point relative coordinate detector 312 which derives coordinates and real world coordinates, calculates relative coordinates of the feature points using the derived real world coordinates, and the distance between the detected object and the mobile robot using the calculated relative coordinates. Using a distance calculator 314 for calculating a value, a map database 316 for storing data outputted from the components, and data on the position and attitude of the mobile robot, and using the data stored in the map database. An absolute position deriving unit 318 for deriving the absolute position of the mobile robot, and the control unit 320 for controlling the mobile robot and the camera device and the respective components.

Here, when the function of the control unit 320 is described in more detail, the initial zoom out / zoom-in image for the mobile robot work area received by the image receiver 302 is transferred to the specific magnification detector 304 to detect a specific magnification. The invariant feature points are detected from each image by passing to the feature point detector 308, and the relative coordinates based on the position of the mobile robot are determined using the feature point relative coordinate detector 312 with respect to the matching feature points in the two images. The map database 316 detects the detected feature points and the relative coordinate data of the feature points, and the position at which the initial zoom-out / zoom-in image is acquired and the posture at the time of image acquisition in the working area of the mobile robot. It performs a function of controlling each configuration of the main processing device 300, including the function to store in.

In the method of estimating the magnetic position of the mobile robot using the monocular zoom camera according to the present invention, which is composed of the above-described elements, as shown in FIG. 2, the main processing apparatus acquires images that are zoomed out / zoomed into the mobile robot workspace. And a map construction process (A) for constructing a map database by detecting feature points of each image and detecting relative coordinates based on the position of the mobile robot with respect to each of the detected feature points. Absolute position detection process for detecting a feature point of an image acquired while driving, calculating a rotation and movement parameter for a feature point matching the feature points in the map database, and detecting the absolute position of the robot using the rotation and movement parameter (B )

3 is a realistic view showing feature point matching of a zoom-out / zoom-in image according to an embodiment of the present invention, FIG. 3A is a zoom-out (original) image, FIG. 3B is a feature point of the zoom-out image, FIG. 3C is a zoom-in image, and FIG. 3D Denotes matching between the feature point of the zoomed-in image and the zoom-out image feature point of FIG. 3B.

The map building process A will be described step by step with reference to FIG. 4 (steps S2 to S14).

The main processing apparatus acquires the initial zoom out / zoom in image using the camera apparatus and detects the specific magnification of the zoom of the camera apparatus (S2).

Referring to step S2 in detail, the main processing apparatus acquires the image of the initial zoom-out state at the position A of the mobile robot using the camera apparatus, and moves the mobile robot by a specific moving distance to obtain the image of the zoom-out state at the position B. After the acquisition, the mobile robot is moved to position A again to obtain an image that is the same image as that obtained at position B, and the zoom-in ratio at the time of image acquisition is determined as a specific magnification value. By the specific magnification, the mobile robot can continuously acquire two images having a certain distance difference while driving. In other words, the images acquired in step S2 are not stored as a reference image in the map database, but are for deriving a specific magnification to have a certain distance by using the image of the second step.

The main processing apparatus calculates an internal parameter for calibrating the camera using the initial zoom-out / zoom-in image acquired using the camera apparatus (S4).

Referring to step S4 in detail, the main processing unit performs corrections for the camera device in the initial zoom-out state and the camera device zoomed in at a specific magnification, respectively, wherein the correction process of the camera device is performed by a computer image point and a real world point. In order to determine the correlation, the correct correction must be made to obtain the exact position of the three-dimensional image from the computer image coordinates, and conversely, to accurately estimate the two-dimensional computer image coordinates from the actual position of the three-dimensional image. do.

In order to perform the calibration process of the camera apparatus, the main processing apparatus first sets a camera model as shown in FIG. 5 on the assumption that the lens of the camera apparatus is an ideal shape.

FIG. 5 illustrates a process of converting real world coordinates of an object and image coordinates (referred to coordinates projected from two dimensions to two dimensions) by a lens in a pinhole camera model.

-M (X, Y, Z): object position in 3D real world coordinates

m ( u, v ): object position in image coordinates

C: optical center of camera coordinate system

-(x, y): image coordinate system with p center

-Distance between C and p: focal length f

이며, 각 벡터에 스케일벡터를 추가하여 표시하면,

이 된다. 실세계 좌표의 M을 영상좌표의 m으로 투영한 관계를 식으로 나타내면 다음의 [수학식1]과 같다.If m and M in Fig. 5 is represented as a vector,

If we add scale vector to each vector,

Becomes The relation of projecting M of real world coordinates to m of image coordinates is expressed by Equation 1 below.

[Equation 1]

R, t: ‘external parameter’, means rotation and transformation between real world and camera coordinate system

A: Camera's 'internal parameter matrix'

u ₀ , v ₀ : 'Principal point', the point where the principal plane intersects the optical axis

α, β: mean scale values in the image coordinate system

c: The distortion value between the image axes

The process of calculating the internal parameter matrix A from the Equation 1 is described as follows.

의 스케일 값 T를 단위항인 1로 하고, Z를 0으로 가정하며, 회전행렬 R의 i번째 열벡터를 r_i로 설정하여, [수학식1]로부터 다음의 [수학식2]를 도출한다.In [Equation 1]

The scale value T of is assumed to be 1 as a unit term, Z is assumed to be 0, and the i th column vector of the rotation matrix R is set to r _i to derive the following [Equation 2] from [Equation 1]: .

[Equation 2]

When [Equation 2] is arranged in the form of [Equation 1], it becomes as [Equation 3] below.

[Equation 3]

If the i-th column vector of H is referred to as h _i , Equation 3 becomes Equation 4 below.

[Equation 4]

Λ in [Equation 4] is an arbitrary scalar value, and since r ₁ and r ₂ satisfy orthogonality, the following Equations 5 and 6 are satisfied.

[Equation 5]

[Equation 6]

이라 하면, B는 다음의 [수학식7]과 같다.here,

In this case, B is represented by Equation 7 below.

[Equation 7]

In Equation 7, since B is symmetric, it can be defined as b as shown in Equation 8 below.

[Equation 8]

라 하면, 다음과 같은 [수학식9], [수학식10]가 도출되며,I-th column vector of H

Then, the following [Equation 9], [Equation 10] is derived,

[Equation 9]

[Equation 10]

[Equation 9], [Equation 10] is summarized as the following [Equation 11].

[Equation 11]

Since n images are used for camera correction, Equation 11 becomes Equation 12 below.

[Equation 12]

In an embodiment of the present invention, V in Equation 12 is a 2n × 6 matrix, and when the b value is estimated, the internal parameter matrix A can be obtained.

를 만족할 때까지 추정하며, B값을 정한 후에는 상기 [수학식7]을 이용하여 다음의 [수학식13]의 내부 파라미터 (u₀, v₀), (α, β), c를 산출한다.The main processing unit for matrix B

Equation (7) is calculated until B is satisfied, and after determining the B value, the internal parameters (u ₀ , v ₀ ), (α, β), and c of Equation 13 are calculated using Equation 7 above. .

[Equation 13]

Since the values calculated in Equation 13 do not provide an optimal solution, the main processor refines the values using the maximum likelihood estimation method. Here, the maximum likelihood estimation in n images with m points is derived by minimizing Equation 14 below.

[Equation 14]

는 영상 i에 대한 점 M_j의 사영(projection)이다.In [Equation 14]

Is the projection of point M _j for image i.

After calculating the internal parameters in step S4, the main processing apparatus generates the feature point set F _i and F _{i +1} by detecting scale invariant feature points of the image which is the initial zoom out / zoom using the Gaussian difference method (S6).

The sixth step is described in detail as follows.

는 다음의 [수학식15]와 같이 입력 영상

와 가우시안 함수

의 컨볼루션으로 표시된다.Since the present invention detects feature points (landmarks) at different distances and viewpoints, a Gaussian filter is used to have a feature that is invariant in scale space.

Is the input image as shown in [Equation 15] below.

And Gaussian functions

It is represented by the convolution of.

[Equation 15]

Here, the Gaussian difference method is a method of efficiently detecting stable positions of feature points using poles in difference between Gaussian functions convolved with an image, and is detected using Equation 16 below.

[Equation 16]

는 [수학식17]로 정의된다.The constant k usually uses a multiplier of 2 ^1/2 ,

Is defined by [Equation 17].

[Equation 17]

In the present invention, the main processing apparatus selects local local and local points from the generated Gaussian difference image as candidates of scale invariant feature points, and local local and local points each pixel of the Gaussian difference image of all levels. Determination is made by comparing with eight pixels in the vicinity, nine pixels corresponding to the adjacent upper and lower levels, and a total of 26 pixels.

For the image processing of the local image (L _{x , y} ) having a radius of 9 pixels at each feature point position, the main processor can correct the illumination of the candidate points found in the Gaussian difference image. Using the equations (18) and (19), the magnitude (m) and azimuth (θ) are calculated.

Equation 18

[Equation 19]

The main processing unit calculates a reference orientation of the feature point with respect to the peripheral region of the feature point detected by the scale invariant feature point detection method by the azimuth calculation method according to Equation 19, and calculates the reference orientation calculated from the orientation of the pixels of the peripheral region. The new orientation is calculated by subtracting the orientation, and the calculated new orientation is set to a canonical orientation value that is invariant to the rotational transformation of the image. Azimuth histogram is generated by quantizing the reference orientation invariant to the rotation change to eight angles to generate a feature vector for the feature point.In the embodiment of the present invention, since the histogram quantized at eight angles for the 4x4 region is used. Each feature point generates a feature vector of 4 × 4 × 8 = 128 dimensions, and the descriptor of the scale invariant feature points consists of the generated 128 feature vectors and coordinates, scales, and reference directions in the original image.

The main processor applies the internal parameters calculated in step S6 to [Equation 1] to calculate external parameters for the rotation and transformation of the feature point, and generates the generated 4 parameters using the calculated external and internal parameters. Projection coordinates of the real world are derived from the linear system of equations (S8).

A detailed description of the step S8 is as follows.

The main processor calculates an external parameter by applying the internal parameter calculated in the above process to Equation 1 for the relationship of projecting M of real world coordinates to m of image coordinates. Since one camera is used for the convenience of calculation and installation in the present invention, R and t have constant values, and external parameters according to the embodiment of the present invention are defined by Equation 20 below. In Equation 20, 'd' is a distance constant when zoomed in by a specific magnification, and R and t are external parameters of the camera and represent rotation and translation between the real world coordinate system and the camera coordinate system.

[Equation 20]

과

을 생성하고, 생성된 두 영상좌표를 [수학식20]에 대입하여 [수학식21]과 같은 4개의 1차 선형연립방정식을 도출한다.Two image coordinates with different external parameters for the same point M in the real world through matching between the feature points of the feature set F _i and F _{i +1}

and

Then, four linear linear equations such as Equation 21 are derived by substituting the two image coordinates into Equation 20.

[Equation 21]

(A ₁ : Internal parameters of the camera's zoom out state calculated by camera calibration)

(A ₂ : internal parameter of the zoom-in state of the camera calculated by camera calibration)

The main processing unit derives the projection coordinates (X, Y, Z, T) of the four unknowns from the real world using the singular value decomposition calculation method (SVD), and uses the camera center as the origin. The coordinates M (U, V, W) are derived by the relational expression of U = X / T, V = Y / T, and W = W / T. In an embodiment of the invention, 'T = 1'.

,

및 실세계좌표

의 관계를 나타내는 [수학식21]을 이용하여 상기 검출된 특징점들과 카메라사이의 실제거리 d를 도출한다(S10).The main processing unit has projection coordinates

,

And real world coordinates

The actual distance d between the detected feature points and the camera is derived by using Equation 21 representing the relationship (S10).

항뿐이므로 [수학식21]의 연립방정식을 풀면

항의 (X, Y, Z) 값, 즉 실세계좌표점이 검출되며, 이 점은 로봇을 기준으로 한 특징점들의 상대좌표점을 나타낸다. 상기 검출된 특징점의 실세계좌표점은 카메라의 중심점을 원점으로 하는 좌표계이므로, 카메라 원점을 이동로봇의 중심으로 가정하면, 이동로봇의 중심에서 특징점까지의 실제 거리(d)는

로 산출된다. 여기서, x_i, y_i, z_i는 i번째 특징점의 실세계 좌표(3D좌표)를 나타낸다.Referring to step S10 in detail, the unknown term in the system of equations [21]

Since it is only a term, solving the system of equations [21]

The (X, Y, Z) value of the term, i.e., the real world coordinate point, is detected, which represents the relative coordinate point of the feature points based on the robot. Since the real world coordinate point of the detected feature point is a coordinate system using the center point of the camera as an origin, assuming that the camera origin is the center of the mobile robot, the actual distance d from the center of the mobile robot to the feature point is

Is calculated. Here, x _i , y _i , z _i represent real world coordinates (3D coordinates) of the i-th feature point.

As described above, when the real world coordinate point of the feature point is derived, the actual distance between the mobile robot and the feature point is obtained.

The main processing apparatus stores the data output in steps S4 to S10 with respect to the zoom-out / zoom-in image obtained while driving the workspace of the mobile robot to build a map database (S12).

Here, the data for the map database includes a reference image (zoom out / zoom-in image acquired for constructing a map database of a mobile robot's workspace), feature points (including descriptors of feature points and relative coordinate information), and images. It is information about the posture (x, y coordinate values and azimuth of the mobile robot) of the acquired mobile robot.

After completing the map construction process by performing the steps S2 to S12 as described above, the absolute position detection process (B) of the main processing apparatus will be described step by step with reference to FIG. Step S24).

The main processing apparatus obtains the zoomed out / zoom-in image by using the camera device while the mobile robot is driving, and detects scale invariant feature points from the obtained zoomed-out / zoom-in image (S14 and S16).

The main processing apparatus detects a feature point that matches the feature points of the map database among the feature points that match each other in the zoom-out / zoom-in image (S18).

The main processing device, a relative coordinate (N _i) is detected, and (S20), the detected relative coordinates and the feature points of the map database based on the location of the mobile robot relative to the detected feature point in the first S18 step The relative coordinates M _k are compared to calculate rotation and movement parameters of the relative coordinates detected in step S20 (S22).

The main processing apparatus detects the absolute position of the mobile robot using the rotation and movement parameters calculated in the step S22 and the attitude information of the robot stored in the map database (S24).

라 하고, 이동로봇 주행중 획득되는 영상에서 상기 참조영상과 매칭된 특징점의 상대위치를

라고 할 때, 두 특징점은 다음의 [수학식22]를 만족한다.In detail, the relative position of the feature point in the reference image (referring to the image input during the map construction process, see step S2).

The relative position of the feature point matched with the reference image in the image acquired while driving the mobile robot

In this case, the two feature points satisfy the following Equation 22.

[Equation 22]

When the main processor calculates t _x , t _z , and θ in [Equation 22], the relative position and attitude of the mobile robot during driving corresponding to the relative position of the mobile robot in the map database are respectively detected. When the image is matched and the calculated t _x , t _z , θ is applied to the attitude P _k information of the robot stored in the map database, the absolute position of the mobile robot is detected. That is, when two or more matching points are derived from the feature points in the image input while driving and the reference points stored in the map database, the position and attitude of the mobile robot can be detected. .

The present invention detects scale invariant feature points for each image, performs matching between the feature points, using a zoom-out (original) image and a zoom-in image obtained by a monocular zoom camera as the robot moves. The actual distance between the mobile robot and the feature point can be obtained by deriving the real world coordinate points of the feature points and reconstructing them in three dimensions, thereby accurately measuring the relative position of the mobile robot from the feature point.

The present invention facilitates the process of acquiring the two images using the specific magnification by deriving a specific magnification for the zoom-out (original) image and the zoom-in image acquired through the monocular zoom camera, and the mobile robot when the two images are acquired. There is an effect that is not affected by the error factors that may occur due to the movement of the.

The present invention constructs a map database (or feature point map) by using feature points detected from a zoom-out (original) image and a zoom-in image acquired through a monocular zoom camera, calculates rotation and movement parameters for the feature points, and calculates distance. The information error can be reduced, and the magnetic absolute position of the robot can be estimated using the calculated parameters and the data of the map database.

As described above, specific preferred embodiments of the present invention have been described, but the present invention is not limited to the above-described embodiments, and various modifications by those skilled in the art to which the present invention pertains are described below. It should be said that they fall within the claims of the present invention as claimed.

Claims (8)

In the magnetic position estimation system of a mobile robot, A camera device for obtaining images of a work area of the mobile robot; Store the images initially received from the camera apparatus as a reference image, construct a map database by deriving feature points, matching points and relative position and attitude information of the mobile robot, and running the camera while driving the mobile robot. A main processor for deriving relative position, absolute position, and attitude information of the mobile robot running while comparing the image received from the apparatus with a reference image in the constructed map database; And And a serial communication device, which is an information transmission path between the camera device and the main processing device. According to claim 1, wherein the main processing unit, The zoom-out / zoom-in images acquired through the camera apparatus while driving the mobile robot are stored as reference image data, and feature coordinates of the respective images, relative coordinate data of the feature points, and mobile robot posture information of the point in time at which the images are acquired. Magnetic location estimation system of a mobile robot using a monocular zoom camera, characterized in that to build a database of the work area map of the mobile robot by storing a. According to claim 1, wherein the main processing unit, An image receiving unit which receives a zoom out / zoom in image from the camera device; A specific magnification detector for detecting a specific magnification of a zoom of the camera device from the received images; An internal parameter calculator configured to calculate an internal parameter for calibrating the camera by using the zoom-out / zoom-in image received from the camera device; A feature point detection unit for generating feature point sets F _i and F _{i + 1} by detecting scale invariant feature points of the zoom-out / zoom-in image using a Gaussian difference operator; An external parameter calculation unit for calculating an external parameter representing rotation and movement transformation of a real world coordinate system and a camera coordinate system with respect to the feature point; A feature point relative coordinate which derives the projection coordinates and the real world coordinates of the feature points from the four linear system equations generated using the calculated external and internal parameters, and calculates the relative coordinates of the feature points using the derived real world coordinates. Detection unit; A distance calculator configured to calculate a distance between the detected object and the mobile robot using the calculated relative coordinates; A map database for storing data on the position and attitude of the mobile robot; An absolute position deriving unit for deriving an absolute position of the mobile robot using data stored in the map database; And And a control unit for controlling the mobile robot and the camera device and the respective components. A system for estimating a magnetic position of a mobile robot using a monocular zoom camera. The method of claim 3, wherein the control unit, The initial zoom-out and zoom-in image of the mobile robot's work area received by the image receiver are transmitted to the feature point detector to detect scale invariant feature points in each image, and the feature point relative to the feature point matched between the initial zoom-out and zoom-in image. A relative coordinate is detected based on the position of the mobile robot using a coordinate detector, and the coordinate data of the detected feature points and the feature points, and the position where the initial zoom-out and zoom-in image are acquired in the working area of the mobile robot are obtained. And a position data of the posture at the time of image acquisition in the map database. The method of claim 3, wherein the control unit, With respect to the relative coordinate data of the image feature points input as the mobile robot travels, the external parameter rotated and moved based on the relative coordinate data of the feature points stored in the map database is extracted using the external parameter calculator, and And an absolute position of the mobile robot by deriving an absolute position of the mobile robot by mapping the extracted external parameter to data stored in the map database. In the magnetic position estimation method of a mobile robot using a monocular zoom camera, The main processing apparatus obtains zoomed out / zoomed images for the mobile robot workspace, detects feature points of each image, and detects relative coordinates based on the position of the mobile robot with respect to each detected feature point. Map construction process; And The main processing apparatus detects a feature point of an image acquired while the mobile robot is driving, calculates a rotation and movement parameter with respect to a feature point matching the feature points in the map database, and uses the rotation and movement parameter to determine an absolute position of the robot. Absolute position detection process for detecting the; magnetic position estimation method of a mobile robot using a monocular zoom camera comprising a. The method of claim 6, wherein the map construction process, (a1) acquiring, by the main processing apparatus, an initial zoom-out / zoom-in image using the camera apparatus and detecting a zoom-in specific magnification of the camera apparatus; (a2) calculating, by the main processing apparatus, an internal parameter for calibrating the camera by using an initial zoom-out / zoom-in image obtained by using the camera apparatus; (a3) generating, by the main processing apparatus, a feature point set by detecting scale invariant feature points of an initial zoom-out / zoom-in image using a Gaussian difference method; (a4) The main processing apparatus calculates the external parameters for the rotation and transformation of the feature point using the internal parameters calculated in the step (a2), and generates four using the calculated external and internal parameters. Deriving the projection coordinates of the real world from the linear system of equations; (a5) deriving, by the main processing apparatus, the actual distance between the detected feature points and the camera using a relational expression between the projection coordinates and the real world coordinates; And (a6) storing, by the main processing apparatus, data output in steps (a2) to (a5) with respect to the zoom-out / zoom-in image obtained while driving the work space of the mobile robot; A method for estimating the magnetic position of a mobile robot using a monocular zoom camera. The method of claim 6, wherein the absolute position detection process, (b1) acquiring, by the main processing apparatus, an image zoomed out / zoom in using the camera device while the mobile robot is traveling; (b2) detecting, by the main processing apparatus, scale invariant feature points from the zoom-out / zoom-in image; (b3) detecting, by the main processing apparatus, feature points that match feature points of the map database among feature points matched with each other in the zoom-out / zoom-in image; (b4) detecting, by the main processing apparatus, relative coordinates based on the position of the mobile robot with respect to the feature points detected in the step (b3); (b5) the main processor compares the relative coordinates detected in the step (b4) with the relative coordinates of the corresponding feature points of the map database, and calculates the rotation and movement parameters of the relative coordinates detected in the step (b4). step; And (b6) detecting, by the main processing apparatus, the absolute position of the mobile robot using the calculated rotation and movement parameters and attitude information of the robot stored in a map database. A method for estimating the magnetic position of a mobile robot using

Images