KR102006291B1 - Method for estimating pose of moving object of electronic apparatus - Google Patents

Abstract

A method of estimating a moving body pose of an electronic device is provided. A method for estimating a moving object pose of an electronic device, the method comprising the steps of: (A) determining whether the electronic device has position information and direction information of a moving object sensed by a GPS / IMU (Global Positioning System / Inertial Measurement Unit) sensor, (B) estimating initial pose information of the moving object from the received position information and direction information, (C) estimating initial pose information of the moving object from the received position information and direction information, (C) (D) comparing the initial pose information estimated in the semantic map, which is a reference for correcting the position and direction of the moving object, with the semantic comparison area And a step of estimating and correcting a pose of the moving object by comparing the set semantic comparison area with the image analysis information, wherein the semantic map, the semantic ratio Area, and image analysis information is a map and / or image classified into a number of classes that contain the road and rain.

Description

[0001] The present invention relates to a method for estimating a moving body pose of an electronic device,

The present invention relates to a method for estimating a moving object pose of an electronic device, and more particularly, to a method and apparatus for estimating a moving object pose of an electronic device by using a camera and a GPS / IMU sensor of relatively low- And a method of estimating a moving body pose of the apparatus.

In general, an unmanned mobile object such as an unmanned robot used in the battlefield checks the forward situation of the unmanned robot through the running image on the remote control device, and controls the traveling of the unmanned mobile object while ascertaining the moving route in the state diagram.

However, since it is difficult to operate the vehicle while confirming the driving image and the state map in a rapidly changing battlefield, it is a recent tendency to display the route plan and various situation information on the running image by applying the augmented reality technology, In order to match and display the image, a technique for accurately estimating the position of the unmanned mobile object itself is required.

For example, the position of an unmanned vehicle is estimated using a satellite navigation system. However, the existing satellite navigation system has a large error caused by environmental influences, so it is impossible to accurately calculate the position depending on the satellite navigation system alone.

In order to solve this problem, one of the proposed techniques is a method of estimating a pose of a mobile body at a sensor level using a high-precision GPS / IMU (Global Positioning System / Inertial Measurement Unit) And the map matching is performed by comparing the detected feature information with the acquired three-dimensional map information.

However, since the method using the high-precision GPS / IMU sensor relies on a high-precision sensor that forms a high price, it is difficult to apply the method to an unmanned mobile object with a general low-cost specification. In addition, the GPS-based method may increase the error in a manner that can not be controlled according to the reception situation (when it is not open-sky).

In addition, the method using the LiDAR sensor is disadvantageous in that it requires maintenance and repair of the 3D map.

Korean Patent Publication No. 10-2014-0016781

SUMMARY OF THE INVENTION [0006] In order to solve the above problems, an object of the present invention is to accurately estimate a position and a direction of a mobile object, that is, a pose of a mobile object, using a camera and a relatively low- And a method of estimating a moving body pose of an electronic device.

The solution of the present invention is not limited to those mentioned above, and other solutions not mentioned can be clearly understood by those skilled in the art from the following description.

According to an aspect of the present invention, there is provided an electronic device for estimating a moving object pose, comprising: a moving object having at least one camera and a GPS / IMU (Global Positioning System / Inertial Measurement Unit) A communication interface unit for communication; Wherein the at least one camera captures an image of at least one of the front and rear sides of the moving object, positional information and direction information of the moving object sensed by the GPS / IMU sensor, and semantic A memory for storing a pose estimation program for estimating a pose of the moving object using a map; And estimating initial pose information of the moving object from the position information and direction information by executing the stored pose estimation program, setting a semantic comparison area on the basis of the initial pose information in the semantic map, Wherein the semantic map, the semantic comparison region, and the image analysis information are included in the road information of the moving object, And / or an image classified into a plurality of classes including a non-path.

Wherein the processor comprises: an image analyzing unit for analyzing the photographed image and obtaining image analysis information classified into a road and a background in units of pixels; An initial pause estimating unit for estimating initial pose information of the moving object from the position information and the direction information of the moving object; A semantic comparison area setting unit for setting a reference point using the position information and the azimuth information among the estimated initial pose information in the semantic map and setting an area having a predetermined range based on the reference point as the semantic comparison area; A map matching unit for map-matching the image analysis information from an origin of the set semantic comparison area to identify location information and direction information of an area having the lowest error with the image analysis information in the semantic comparison area; And a final pose estimating unit for estimating a pose of the moving object by adding the initial pose information and the position information and direction information confirmed by the map matching unit.

The image analyzing unit determines and classifies the class of the road and the background of each pixel of the photographed image by applying a Semantic Segmentation algorithm using CNN (Convolutional Neural Network).

Wherein the map matching unit comprises: a coordinate system transform unit for transforming the image coordinate system of the acquired image analysis information into a world coordinate system; An error calculating unit for calculating an error in the origin and the moved positions by performing map matching while moving the image analysis information converted into the world coordinate system by a predetermined distance and / or angle from the origin of the semantic comparison area; And a confirming unit for confirming position information and direction information corresponding to the calculated lowest error among the errors calculated at the origin and the moved positions.

Wherein the error calculator sets different error weights according to class types classified for each pixel of the semantic comparison area and classified classes for each pixel of the image analysis information to calculate an error of the image analysis information and the semantic comparison area on a pixel- A class of the semantic comparison area is classified into a road, a parking lot and a background, and a class of the image analysis information is classified into a road and a background.

The error calculator calculates an error using the following equation.

는 상기 영상 분석 정보와 상기 시맨틱 비교 영역의 오차를 산출하기 위해 정의된 목적함수,

는 상기 위치정보와 방향정보, (x, y)는 월드 좌표계에서 픽셀의 위치,

는 상기 시맨틱 비교 영역,

는 상기 시맨틱 지도,

는 상기 영상 분석 결과,

는 상기 오차 가중치이다.here,

Is an objective function defined to calculate the error between the image analysis information and the semantic comparison area,

(X, y) represents the position of the pixel in the world coordinate system,

The semantic comparison area,

The semantic map,

The image analysis result,

Is the error weight.

According to another embodiment of the present invention, there is provided a method for estimating a moving body pose of an electronic device, the method comprising: (A) determining whether the electronic device has position information of the moving object sensed by a GPS / IMU (Global Positioning System / The direction information and at least one camera photographing at least one of the front and rear sides of the moving object from the moving object; (B) estimating initial pose information of the moving object from the received position information and direction information; (C) the electronic device analyzing the received photographed image to obtain image analysis information; And (D) the electronic device sets a semantic comparison area on the basis of the estimated initial pose information in a semantic map serving as a reference for correcting the position and direction of the moving object, A semantic comparison area, and the image analysis information are maps and / or maps classified into a plurality of classes including roads and non-roads, and the step of estimating and correcting the pose of the moving object by comparing the semantic map, It is a video.

Wherein the step (C) comprises: analyzing an image of at least one of the front and rear sides of the received moving object to obtain image analysis information classified into a road and a background in pixel units; and (D) (D1) the electronic device sets the position information and the azimuth information of the estimated initial pose information as reference points in the semantic map, and sets an area having a predetermined range on the basis of the reference point as the semantic comparison area ; (D2) the electronic device performs map matching of the image analysis information from an origin of the set semantic comparison area to check position information and direction information of an area having the lowest error with the image analysis information in the semantic comparison area ; And (D3) the electronic device finally estimating a pose of the moving object by adding the initial pose information and the position information and direction information confirmed in the step (D2).

Wherein the electronic device converts the image coordinate system of the image analysis information obtained in the step (C) into a world coordinate system, and converts the image analysis information converted into the world coordinate system from the origin of the semantic comparison area And calculates an error at the origin and the moved positions, maps a position corresponding to the calculated lowest error among the errors calculated at the origin and the moved positions, Check information and direction information.

In the step (D2), the electronic device sets different error weights according to the class class classified for each pixel of the semantic comparison area and the class class classified for each pixel of the image analysis information, The class of the semantic comparison area is classified into a road, a parking lot, and a background, and the class of the image analysis information is classified into a road and a background.

According to the present invention, the position or pose of the moving object can be accurately estimated at a low cost by using the GPS / IMU sensor having a relatively large error with the monocular camera.

Further, according to the present invention, it is possible to know precise information about the position and the viewing direction of the traveling body, and thus it is possible to accurately generate a control signal required for autonomous traveling.

Also, according to the present invention, it is possible to apply the present invention to a field where accurate pose prediction of a moving object such as navigation or remote control, which provides a service based on Augmented Reality, is essential.

The effects of the present invention are not limited to those mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the following description.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram illustrating a mobile body pose estimation system according to an embodiment of the present invention;
2 is a block diagram illustrating another processor in more detail in accordance with an embodiment of the present invention;
3 is a diagram showing an example of image analysis,
4 is a diagram showing a relationship between a semantic map, a semantic comparison area, and image analysis information,
FIG. 5 is a block diagram showing the map matching unit shown in FIG. 2 in more detail,
6 is a flowchart for explaining a moving body pose estimation method of an electronic device according to an embodiment of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS The above and other objects, features, and advantages of the present invention will become more readily apparent from the following description of preferred embodiments with reference to the accompanying drawings. However, the present invention is not limited to the embodiments described herein but may be embodied in other forms. Rather, the embodiments disclosed herein are provided so that the disclosure can be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

In this specification, when an element is referred to as being on another element, it may be directly formed on another element, or a third element may be interposed therebetween.

Where the terms first, second, etc. are used herein to describe components, these components should not be limited by such terms. These terms have only been used to distinguish one component from another. The embodiments described and exemplified herein also include their complementary embodiments.

Also, when it is mentioned that the first element (or component) is operated or executed on the second element (or component) ON, the first element (or component) It should be understood that it is operated or executed in an operating or running environment or is operated or executed through direct or indirect interaction with a second element (or component).

It is to be understood that when an element, component, apparatus, or system is referred to as comprising a program or a component made up of software, it is not explicitly stated that the element, component, (E.g., memory, CPU, etc.) or other programs or software (e.g., drivers necessary to run an operating system or hardware, etc.)

It is also to be understood that the elements (or elements) may be implemented in software, hardware, or any form of software and hardware, unless the context requires otherwise.

Also, terms used herein are for the purpose of illustrating embodiments and are not intended to limit the invention. In the present specification, the singular form includes plural forms unless otherwise specified in the specification. The terms "comprises" and / or "comprising" used in the specification do not exclude the presence or addition of one or more other elements.

Hereinafter, the present invention will be described in detail with reference to the drawings. In describing the specific embodiments below, various specific details have been set forth in order to explain the invention in greater detail and to assist in understanding it. However, it will be appreciated by those skilled in the art that the present invention may be understood by those skilled in the art without departing from such specific details.

In some instances, it should be noted that portions of the invention that are not commonly known in the description of the invention and are not significantly related to the invention do not describe confusing reasons for explaining the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail with reference to the accompanying drawings.

Each configuration of the electronic device 200 for the mobile body pose estimation shown in FIG. 1 indicates that it can be functionally and logically separated, and that each configuration is separated into separate physical devices or written in separate codes It will be readily apparent to one of ordinary skill in the art of the present invention.

In addition, the electronic device 200 for estimating the moving body pose may be installed in a predetermined data processing apparatus to implement the technical idea of the present invention.

The electronic device 200 for estimating the moving body pose according to the embodiment of the present invention may be implemented by a computer such as a desktop PC, a server, a laptop PC, a netbook computer, And may be one of all electronic devices capable of being installed and executed.

1 is a diagram illustrating a moving body pose estimation system according to an embodiment of the present invention.

Referring to FIG. 1, a moving body pose estimation system according to an embodiment of the present invention includes a moving object 100 and an electronic device 200 for estimating moving object pose.

The mobile device 100 may enter a place where it is difficult for the user to directly search the mobile device 100 by remote control of the electronic device 200 to take a surrounding image and transmit the photographed image to the electronic device 200. [ Examples of the moving body 100 include an unmanned moving body such as an unmanned robot or an unmanned vehicle.

The mobile unit 100 may include a camera unit 110, a GPS / IMU (Global Positioning System / Inertial Measurement Unit) sensor 120, and a communication interface circuit 130.

The camera unit 110 may include at least one camera that photographs at least one of the front and rear sides of the moving object 100 and outputs a photographed image. The front side of the moving body 100 is a direction in which the moving body 100 advances. The rear side is a direction opposite to the traveling direction of the moving body 100, It means direction to look at. If the camera unit 110 includes only one camera, it can be basically set to photograph the front of the mobile unit 100. [

The GPS / IMU sensor 120 senses the current position and direction of the mobile body 100 and can output positional information and directional information of the mobile body 100. The GPS / IMU sensor 120 used in the embodiment of the present invention is not an expensive high-precision sensor but may be a low-precision sensor having a relatively large error as compared with a high-precision sensor.

The GPS sensor of the GPS / IMU sensor 120 senses the current position of the mobile body 100.

The IMU sensor of the GPS / IMU sensor 120 is an inertial measurement device and provides a three-dimensional rotational posture of the mobile body 100. To this end, the GPS / IMU sensor 120 is provided with an acceleration sensor for detecting movement in three axes in the three-dimensional space, three-axis rotation of pitch, roll, and yaw (Pitch, roll, yaw) of the moving object 100. [0064] The moving object 100 may include a gyroscope sensor for detecting the direction of the moving object 100 (pitch, roll, yaw).

The communication interface circuit 130 is capable of remote communication with the electronic device 200. For example, the communication interface circuit 130 transmits at least one of the front and rear images photographed by the camera section 110 and the position information and direction information of the moving object 100 sensed by the sensor 120, (200) and receive a remote control signal from the electronic device (200).

Meanwhile, the electronic device 200 for estimating the moving body pose according to the embodiment of the present invention can estimate the position and the direction of the moving body 100 that can be moved by remote control or the autonomous traveling system, that is, Can be estimated. In addition, the electronic device 200 can remotely control the mobile unit 100 to remotely monitor various operations such as searching for the surroundings, shooting the surroundings, providing the photographed peripheral image, and providing the augmented reality of the target.

To this end, an electronic device 200 for estimating moving object pose according to an embodiment of the present invention includes a bus 210, a communication interface unit 220, a user interface unit 230, a memory 240, and a processor 250 .

The bus 210 connects the communication interface 220, the user interface 230, the memory 240 and the processor 250 to each other and transmits various signals such as control messages, status information, and / Circuit.

The communication interface unit 220 is a device for wireless communication between the mobile device 100 and the electronic device 200. For example, the communication interface unit 220 transmits a remote control command to the mobile unit 100, and receives a video image of at least one of the front and rear sides of the mobile unit from the mobile unit 100 and a GPS / IMU sensing information, It is possible to receive position information and direction information of the mobile terminal 100.

 The user interface unit 230 may transmit commands or data input from a user to another component (s) of the electronic device 200. [ In addition, the user interface unit 230 may display a plurality of screens on a monitor according to a command or data input from a user through a touch panel, a mouse, a keyboard, or the like.

For example, the user interface unit 230 may receive a command for a moving direction of the moving object 100 from a user or a command for displaying a photographed image and a target as an augmented reality received from the moving object 100, And can be implemented and displayed as a photographed image and a target augmented reality.

Memory 240 may include volatile memory and / or non-volatile memory. The memory 240 may store instructions or data related to the components, one or more programs and / or software, an operating system, etc. to implement and / or provide the operations, functions, and the like provided by the electronic device 200 .

In addition, the memory 240 may store a semantic map serving as a reference for correcting the position and direction of the mobile object 100. Semantic refers to meaningful information that enables a computer (electronic device 200 in the case of the present invention) to know.

The semantic map can be set manually or automatically by the user in advance, and is a map that is used as a reference in map matching to be described later for correcting the position and direction of the moving object 100. [ The semantic map is also known by the computer, so the term semantic is used.

In the memory 240, the photographed image received from the moving object 100 and the positional information and directional information of the moving object 100 may be matched and stored over time.

The program stored in the memory 240 may include at least one of the images of at least one of the front and rear sides of the mobile unit 100, the position information of the mobile unit 100 sensed by the GPS / IMU sensor 120, And a pose estimation program including various commands for estimating a pose of the mobile object 100 using the semantic map stored in the memory 240. [ The pose of the moving object 100 includes the position measured by the GPS sensor of the moving object 100 and the direction (pitch, roll, yaw) measured by the IMU sensor.

The processor 250 executes one or more programs stored in the electronic device 200 to control the overall operation of the electronic device 200.

For example, the processor 250 executes the pose estimation program stored in the memory 240 to estimate the initial pose information of the mobile 100 from the position information and direction information received from the mobile 100, The semantic comparison area is set on the basis of the initial pose information, the image analysis information is obtained by analyzing the captured image, and the pose of the moving object is finally estimated (or corrected) by comparing the set semantic comparison area with the image analysis information .

Here, the semantic map, the semantic comparison region, and the image analysis information may be maps and / or images classified into a plurality of classes including roads and non-roads.

More specifically, the semantic map stored in the memory 240 is displayed in the world coordinate system, and M (x, y) shown in FIG. 4, which will be described later, is exemplified. In the case of the present invention, the classes of the semantic map can be classified into roads, parking lots, and backgrounds, which, of course, can be further classified. In addition, the semantic map information has geographical location information defined based on the world coordinate system, and the semantic map can include topographic information (e.g., roads, buildings, trees, etc.) at each location.

Also, since the semantic comparison area corresponds to a part of the semantic map, the classes of the semantic comparison area can be classified into roads, parking lots, and backgrounds.

In addition, classes of image analysis information can be classified into roads and non-roads, that is, roads and backgrounds.

Class classification allows the user to manually set classes of roads, backgrounds, and parking lots in advance for a plurality of image objects. That is, when the user gives the pose estimation program a reference value for each class, the processor 250 can classify the road, the background, the parking lot, and the like into the image input in real time based on the reference value.

FIG. 2 is a block diagram illustrating the processor 250 according to an embodiment of the present invention in more detail. FIG. 3 is a diagram illustrating an example of image analysis. FIG. 4 is a diagram illustrating a semantic map, a semantic comparison region, Fig.

2, the processor 250 includes an image analysis unit 251, an initial pose estimation unit 253, a semantic comparison region setting unit 255, a map matching unit 257, and a final pose estimation unit 259 .

The image analysis unit 251 may analyze the image captured by the camera unit 110 and acquire image analysis information classified into a road and a background in pixel units. That is, the image analyzing unit 251 can classify the pixels into a plurality of classes including roads and obtain image analysis information through an image analysis process.

3, the image analyzer 251 performs a Semantic Segmentation algorithm using a CNN (Convolutional Neural Network) to the photographed image (for example, the forward image 310 of the moving object 100) It is possible to judge and classify each of the pixels of the photographed image belonging to the road and the background, and generate the image analysis information 320 of the classified class. In the image analysis information 320 shown in FIG. 3, a dotted line hatching means a road class, and a white space means a background class.

2, the initial pose estimation unit 253 can estimate the initial pose information of the moving object 100 from the position information and the direction information of the moving object 100 sensed by the GPS / IMU sensor 120 .

Since the initial pose estimating unit 253 is provided with the camera unit 110 and the GPS / IMU sensor 120 fixedly mounted on the moving object 100, the pose of the camera unit 110 is determined by the pose of the GPS / IMU sensor 120 Can be expressed by the rigid body motion relation. The relational expression is an expression for converting the coordinate system of the navigation sensor into the coordinate system of the camera reference based on the distance and the posture difference value between the GPS / IMU sensor 120 as the navigation sensor and the camera unit 110.

The semantic comparison area setting unit 255 sets a reference point using the position information and the azimuth information Yaw in the semantic map stored in the memory 240 and sets the area having a predetermined range on the basis of the reference point as a semantic Can be set as a comparison area.

Referring to FIG. 4, the semantic comparison area setting unit 255 sets S (x1, y1) corresponding to the position information and azimuth information of the initial pose information on the semantic map as a reference point and S (x1, y1) An area having a range of 12 meters left and right and 36 meters forward can be set as a semantic comparison area. At this time, the forward direction and the left and right directions in the reference point S (x1, y1) are based on the traveling direction of the moving object 100, and thus the front of the reference point can coincide with the forward direction of the moving object 100. [ It is a matter of course that the size of the semantic comparison area (left and right 12 meters, front 36 meters) is not limited to this example.

In FIG. 4, the dotted line hatching represents a road class (class 1), the hatched line represents a parking class (class 2), and the white space represents a background class (class 3).

The map matching unit 257 maps the image analysis information 320 from the origin of the semantic comparison area set by the semantic comparison area setting unit 255 and outputs the image analysis information 320 in the semantic comparison area It is possible to confirm the position information and the direction information of the low region. Since the semantic comparison area is an area to be compared with the image analysis information 320a defined on the basis of the world coordinate system, the image analysis information 320a is made only within a range (i.e., a semantic comparison area) set based on initial pose information.

5 is a block diagram showing the map matching unit 257 shown in FIG. 2 in more detail.

5, the map matching unit 257 may include a coordinate system converting unit 257a, an error calculating unit 257b, and an identifying unit 257c according to an embodiment of the present invention.

The coordinate system conversion unit 257a can convert the image coordinate system of the image analysis information 320 acquired by the image analysis unit 251 into a world coordinate system. That is, the coordinate system conversion unit 257a converts the image analysis information 320a shown in FIG. 3 to the image analysis information 320a shown in the right side of FIG. 4 by further using the pitch and roll information from the GPS / IMU sensor 120 ). ≪ / RTI > That is, the front image is converted into the image analysis information 320a in the form of a top view, that is, a front view.

이 이미지 좌표계의 점

에 어떻게 대응되는지 표현할 수 있다. 따라서, 좌표계 변환부(257a)는 영상 분석 정보(320)를 Homography를 통해 월드 좌표계를 기준으로 정의할 수 있다. 한 평면을 다른 평면에 투영(Projection) 시켰을 때 투영된 대응점들 사이에서는 일정한 변환 관계가 성립하며, 이 변환 관계를 Homography라고 한다.In general, given a camera pose and camera calibration matrix, a point in the world coordinate system

The point in this image coordinate system

Can be expressed as to how they correspond to each other. Accordingly, the coordinate system conversion unit 257a can define the image analysis information 320 based on the world coordinate system through the homography. When one plane is projected on another plane, a certain transformation relation is established between projected corresponding points, and this transformation relation is called homography.

The error calculator 257b performs map matching while moving the image analysis information 320a converted to the world coordinate system by a predetermined distance and / or angle from the origin of the semantic comparison area to calculate an error in each of the origin and the moved positions .

4, the semantic comparison area has a range set based on the reference point S (x1, y1). In this case, the shape of the semantic comparison area has the same shape as that of the dotted line rectangle 320a. The error calculating unit 257b calculates an error by matching the image analysis information 320a from the origin with one corner in the semantic comparison area as an origin, and calculates an error in the direction set by the set distance (for example, 0.3 meters) The analysis information 320a is moved to calculate an error, and the error is calculated by moving in the set direction by the set distance again.

In addition, the error calculator 257b can calculate the error by moving a predetermined distance from the origin and rotating by a set angle (for example, by 0.1 degree, +/- 3 degrees).

The error calculator can calculate the error using the following equation (1).

는 영상 분석 정보(320a)와 시맨틱 비교 영역의 오차를 산출하기 위해 정의된 목적함수,

는 위치정보와 방향정보로서 위도(북거) 오차, 경도(동거) 오차 및 방위각 오차를 의미한다. 또한, (x, y)는 월드 좌표계에서 픽셀의 위치,

는 시맨틱 비교 영역,

는 시맨틱 지도,

는 영상 분석 결과(320a),

는 오차 가중치 또는 코스트(cost)를 의미한다.In Equation (1)

Is an objective function defined to calculate the error of the semantic comparison area with the image analysis information 320a,

Means latitude (tolerance) error, hardness (coincidence) error, and azimuth error as position information and direction information. (X, y) is the position of the pixel in the world coordinate system,

The semantic comparison area,

A semantic map,

The image analysis result 320a,

Denotes an error weight or cost.

Equation (1) is an objective function defined to measure the difference between the image analysis information 320a defined in the world coordinate system and the semantic comparison region, i.e., an error. In other words, in Equation (1), the position and the angle when the sum of the costs become minimum in the semantic comparison area mean the error of the mobile unit 100. The map matching unit 257 can optimize the objective function to predict the final pose of the mobile 100 more accurately.

At this time, the error calculating unit 257b sets different costs (i.e., error weights) according to the class types classified for each pixel of the semantic comparison area and the class types classified for each pixel of the image analysis information 320a, The error between the image analysis information 320a and the semantic comparison area can be calculated.

[Table 1] shows an example of the cost set according to the class of the semantic comparison area and the class of the image analysis information 320a.

)는 맵 매칭 시 동일한 위치에 해당하는 시맨틱 비교 영역(또는 시맨틱 지도(

))의 픽셀과 영상 분석 정보(320a)의 픽셀의 클래스가 동일할수록 낮게 설정될 수 있다. 또한, [표 1]에서 코스트가 클수록 동일한 위치에 해당하는 시맨틱 비교 영역의 픽셀과 영상 분석 정보(320a)의 픽셀의 클래스 간의 유사도는 낮아짐을 알 수 있다.Referring to Table 1,

) Is a semantic comparison area (or a semantic map (

) And the pixel class of the image analysis information 320a are the same. Also, in Table 1, it can be seen that the greater the cost, the lower the similarity between the pixel of the semantic comparison area corresponding to the same position and the class of the pixel of the image analysis information 320a.

In Table 1, the parking space in the class of the semantic map is classified as the road at the coordinates corresponding to the image analysis result 320a, and can be classified as the background, so 0.5 is set as the cost.

In the case where the image analysis information 320a is mapped at the origin of the semantic comparison area in consideration of misdetection and non-detection of the image analysis result 320a, when the class of the pixel at the same position is the background and the road, respectively 0.9, and 0.4 for background and background, respectively. These costs can be changed. If the semantic comparison area and the image analysis information 320a are respectively the background and the road, it may mean that the actual background is erroneously classified as a road (false detection) when the semantic division algorithm is applied. .

The confirmation unit 257c can confirm the position information and the direction information corresponding to the calculated lowest error among the errors calculated at the origin and the moved positions in the semantic comparison area. That is, the confirmation unit 257c confirms the rotated angle at the lowest error position as direction information corresponding to the lowest error calculated position.

2, the final pose estimating unit 259 adds the initial pose information of the moving object 100 estimated by the initial pose estimating unit 253 and the position information and direction information confirmed by the identifying unit 257c, (I.e., compensate) the pose of the pacemaker 100.

For example, if it is confirmed that the error at the position shifted to the right by 0.3 m after rotating 1 degree with respect to the origin of the semantic comparison area in the confirmation unit 257c is the lowest, 1 (x, y) = 0.3 , 0) can be used to correct the position information and direction information of the initial pause information.

When the position information and the direction information of the mobile body 100 are corrected by the above-described operation, that is, when the final pose is estimated, the processor 250 refers to the final pose information to generate a control signal required for autonomous travel of the mobile body 100 Can be generated.

In addition, the processor 250 can implement a forward image and a target, which are photographed by the moving object 100 at a more accurate position on the map, using the final pose information as an augmented reality, Can be accurately predicted.

Meanwhile, when the photographed image received from the moving object 100 is four images of the front, back, left, and right chambers of the moving object 100, the processor 250 performs the semantic division algorithm And the map matching is performed after the semantic comparison area is set, thereby correcting the error of the moving object 100 for each image. In addition, the position of the target can be more accurately displayed in augmented reality in consideration of the error of the moving object 100 corrected for each image.

6 is a flowchart for explaining a moving body pose estimation method of the electronic device 200 according to the embodiment of the present invention.

6 can be implemented by the electronic device 200 described with reference to Figs. 1 to 5, detailed description thereof will be omitted.

6, the electronic device 200 includes at least one of the position information and the direction information of the mobile unit 100 sensed by the GPS / IMU sensor 120 of the mobile unit 100, and at least one of the front and rear sides of the mobile unit The photographed image can be received from the mobile 100 (S610).

The electronic device 200 analyzes the received shot image to obtain image analysis information (S620), and estimates initial pose information of the mobile terminal 100 from the received position information and direction information of the mobile terminal 100 S630). In step S620, the electronic device 200 analyzes the image of at least one of the front and rear sides of the moving object 100, and obtains image analysis information classified into a road and a background in units of pixels.

The electronic device 200 sets the semantic comparison area on the basis of the initial pose information in the semantic map serving as a reference for correcting the position and direction of the mobile object 100 and compares the set semantic comparison area with the image analysis information, It is possible to estimate and correct the pose of the object 100 (S640).

In step S640, the electronic device 200 sets the position information and the azimuth information of the estimated initial pose information as the reference point in the semantic map, and sets the area having the predetermined range on the basis of the reference point as the semantic comparison area (S642),

Then, the electronic device 200 can map the image analysis information from the origin of the semantic comparison area set in step S642 and check the location information and direction information of the area having the lowest error with the image analysis information in the semantic comparison area ( S644). In step S644, the electronic device 200 converts the image analysis information into a world coordinate system, and then performs map matching. Map matching is performed at each position with reference to [Equation 1] and [Table 1] can do.

The electronic device 200 can estimate the pose of the mobile unit 100 by adding the initial pose information of the mobile unit 100 estimated in step S630 and the position information and the direction information confirmed in step S644 in step S646.

Meanwhile, the method of estimating moving body pose of an electronic device according to the present invention may be provided in a recording medium readable by a computer by tangibly embodying a program of instructions for implementing the same, have.

That is, the moving body pose estimation method of the electronic device according to the present invention can be implemented in a form of a program that can be performed through various computer means, and can be recorded on a computer-readable recording medium, A data file, a data structure, and the like, alone or in combination. The computer-readable recording medium may be any of various types of media such as magnetic media such as hard disks, optical media such as CD-ROMs and DVDs, and optical disks such as ROMs, RAMs, flash memories, And hardware devices specifically configured to store and execute program instructions.

Accordingly, the present invention also provides a program stored on a computer for controlling the electronic device or on a computer-readable recording medium carried on the electronic device to implement a method of estimating moving body pose of the electronic device.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the present invention is not limited to the disclosed exemplary embodiments, but, on the contrary, It will be apparent to those skilled in the art that numerous modifications and variations can be made to the present invention without departing from the scope of the present invention. Accordingly, all such modifications and variations are intended to be included within the scope of the present invention. Accordingly, the true scope of the present invention should be determined by the technical idea of the appended claims.

100: Moving body 110:
120: GPS / IMU sensor 130: Communication interface circuit
200: electronic device 210: bus
220: communication interface unit 230: user interface unit
240: memory 250: processor

Claims (5)

A method for estimating a moving body pose of an electronic device,
(A) the electronic device displays positional information and direction information of the moving object sensed by a GPS / IMU (Global Positioning System / Inertial Measurement Unit) sensor and an image of at least one of the front and rear sides of the moving object Receiving from the moving body;
(B) estimating initial pose information of the moving object from the received position information and direction information;
(C) the electronic device analyzing the received photographed image to obtain image analysis information; And
(D) the electronic device sets a semantic comparison area on the basis of the estimated initial pose information in a semantic map serving as a reference for correcting the position and direction of the moving object, and sets the semantic comparison area and the image analysis information And estimating and correcting a pose of the moving object,
The semantic map, the semantic comparison region, and the image analysis information are maps and / or images classified into a plurality of classes including roads and non-roads,
The step (C)
Wherein the electronic device applies a semantic segmentation algorithm using a CNN (Convolutional Neural Network) to an image of at least one of the front and rear sides of the received moving object, Background, and image analysis information obtained by classifying,
The step (D)
(D1) the electronic device determines, as a reference point, position information and azimuth information among the estimated initial pose information in the semantic map, and sets an area having a predetermined range based on the reference point as the semantic comparison area ;
(D2) the electronic device performs map matching of the image analysis information from an origin of the set semantic comparison area to check position information and direction information of an area having the lowest error with the image analysis information in the semantic comparison area ; And
(D3) a step of the electronic device finally estimating a pose of the moving object by adding the initial pose information and the position information and direction information confirmed in the step (D2) Estimation method. delete The method according to claim 1,
The step (D2)
Wherein the electronic device converts the image coordinate system of the image analysis information obtained in the step (C) into a world coordinate system, and converts the image analysis information converted into the world coordinate system into a distance and / And calculates an error at the origin and the moved positions. Then, the position information and the direction information corresponding to the calculated lowest error among the errors calculated at the origin and the moved positions are checked And estimating a moving body pose of the electronic device. The method of claim 3,
In the step (D2), the electronic device includes:
Wherein the error analysis unit calculates an error between the image analysis information and the semantic comparison area in units of pixels by setting different error weights according to class types classified for each pixel of the semantic comparison area and classified classes for each pixel of the image analysis information,
Wherein the class of the semantic comparison area is classified into a road, a parking lot, and a background, and the class of the image analysis information is classified into a road and a background. 5. The method of claim 4,
The electronic device estimates a moving body pose of an electronic device, which calculates an error using the following equation:

here,

Is an objective function defined to calculate the error between the image analysis information and the semantic comparison area,

(X, y) represents the position of the pixel in the world coordinate system,

The semantic comparison area,

The semantic map,

The image analysis result,

Is the error weight.

Images