KR101047150B1 - Compensation device and method for physical convenience between laser scanner and video camera, Compensation measurement device and method for physical convenience - Google Patents

Abstract

The present invention is a tracking system for fusing a laser scanner and a video camera to track and recognize the behavior of an object (person), an apparatus and method for correcting the physical convenience between the laser scanner and the video camera, an apparatus for measuring a physical bias correction value and It is about a method.

An apparatus for compensating physical convenience between a laser scanner and a video camera of a tracking system according to an embodiment of the present invention includes a candidate boundary determining unit for extracting a candidate region from a foreground image image from which a background region is removed from an image image, and within the candidate region. A clustering unit for classifying the laser scan data included into a plurality of clusters, a cluster selecting unit for selecting a cluster most likely to be an object among the plurality of clusters classified by the clustering unit, the candidate region and the cluster selecting unit And a physical convenience correction unit configured to correct the physical bias by moving at least one position of the image image and the laser scan data so as to reduce an error of the selected cluster.

Tracking, Objects, Laser Scanners, Video Cameras, Physical Convenience, Clusters, Real Time

Description

Apparatus and method for compensating physical convenience between laser scanner and video camera and measuring apparatus and method for measuring physical comfort value {a calibration of compensate horizonal bias between a laser scanner and a video camera}

The present invention relates to an apparatus and method for correcting physical convenience between a laser scanner and a video camera in a tracking system in which a laser scanner and a video camera are fused to track and recognize an object (person) 's behavior. The invention also relates to an apparatus and method for measuring the physical bias correction between a laser scanner and a video camera.

In many fields, such as security-related fields such as intelligent video surveillance systems, intelligent robots capable of interacting with humans, and intelligent home appliances, technology for tracking people or recognizing human behavior is applied.

As a conventional behavior recognition technique, there is a paper published by Bobick, A.F et al. ["The Recognition of Human Movement Using Temporal Templates", IEEE Transactions on pattern Analysis and Machine Intelligence, 2001). In this paper, a technique for recognizing behavior using a motion history image (MHI) that weights a person's silhouette in front over time is disclosed. However, since this technique uses a fixed video camera, behavior recognition cannot be performed unless the direction of human movement in the image captured by the video camera coincides with the direction learned by the behavior recognizer.

As a conventional behavior recognition technique for solving this problem, a paper published by Weinland, D. et al. ["Free Viewpoint Action Recognition using Motion History Volumes", Computer Vision and Image Understanding, 2006). The technique disclosed in this paper generates a three-dimensional model by matching images input from four cameras, so that behavior recognition is free from viewpoint. However, since the process of matching the four camera data requires a complicated calculation, there is a problem that real-time operation is impossible. In addition, the prior art has a problem that the behavior recognition is possible only when all four cameras capture the object, and even one camera cannot perform the action recognition in the state of not capturing the object. In addition, since four cameras must be installed in all spaces to be perceived, there is a problem that it takes too much cost to build a behavior recognition system.

Accordingly, the inventors of the present invention developed a technology for tracking a person's position and recognizing a behavior by fusing a laser scanner and a video camera and applying it to the Korean Patent Office as 10-2009-0022584. This tracking system technology, which combines a laser scanner and a video camera, uses laser scan data and video images acquired from laser scanners (usually installed on the floor) and video cameras (commonly installed on the ceiling) installed at different locations. To recognize the behavior by tracking the subject (person). At this time, the laser scan data and the video image can be synchronized only when the laser scanner and the video camera are installed at the same angle. The horizontal angle of the user is physically distorted during the process of installing the laser scanner and the video camera, respectively. Occurs. Due to this physical bias, a conventional tracking system causes an error in calculation, thereby causing a problem that the position of the object (person) cannot be tracked robustly.

As a solution to this problem, Q.Zhang and R.Pless used an external assistant, which was proposed in the paper ["Extrinsic calibration of a camera and laser range finder", International Conference on Intelligent Robots, 2004, pp.2301-2306]. The method can be used. However, such a prior art can be calibrated for physical convenience only by installing an external aid such as a checker board, and the user must manually calibrate each time a physical bias occurs between the laser scanner and the video camera. There is a problem that is difficult to apply.

An object of the present invention is to fuse a laser scanner and a video camera to automatically track the horizontal position of a physically generated object between the laser scanner and the video camera without a separate external aid in a tracking system that tracks the position of an object such as a person or an object. An apparatus and method for calibrating are provided.

Another object of the present invention is to provide an apparatus and method for measuring a horizontal bias correction value physically generated between a laser scanner and a video camera.

An apparatus for compensating physical convenience between a laser scanner and a video camera of a tracking system according to an embodiment of the present invention includes a video camera that continuously acquires an image image including an object and a background, and a distance according to an angle with respect to the object and a background. A physical convenience correction device between a laser scanner and a video camera of a tracking system having a laser scanner that continuously acquires laser scan data by measuring a candidate region from a foreground image image from which the background region is removed from the image image. A candidate boundary determination unit, a clustering unit for classifying the laser scan data included in the candidate area into a plurality of clusters, and a cluster selection for selecting a cluster most likely to be the object among a plurality of clusters classified in the clustering unit; And selected from the candidate area and the cluster selection unit. Moving at least one of the position of the video image and the laser scan data so as to reduce the error of the house and is characterized in that parts including physical comfort correction to correct the physical convenience.

In addition, the physical convenience correction method between the laser scanner and the video camera of the tracking system according to an embodiment of the present invention, a video camera that continuously acquires the image image including the object and the background, the angle according to the object and the background A physical convenience correction method between a laser scanner and a video camera of a tracking system having a laser scanner that continuously acquires laser scan data by measuring a distance, wherein the candidate region is selected from a foreground image image from which the background region is removed from the image image. A candidate boundary determination step of extracting, a clustering step of classifying the laser scan data included in the candidate area into a plurality of clusters, and a selection of a cluster most likely to be the object among the plurality of clusters classified in the clustering step; A cluster selection step, the candidate region and the cluster selection group And a physical convenience correction step of correcting the physical bias by moving at least one position of the image image and the laser scan data so as to reduce an error of a selected cluster in the system.

An apparatus for measuring physical convenience correction values between a laser scanner and a video camera of a tracking system according to an embodiment of the present invention includes a video camera that continuously acquires an image image including an object and a background, and a distance according to an angle with respect to the object and a background. An apparatus for measuring a physical convenience correction value between a laser scanner and a video camera of a tracking system having a laser scanner that continuously acquires laser scan data by measuring a measured value of the image, the foreground image image of which the background area is removed from the image image, A candidate boundary determination unit for extracting candidate regions from the correction values, a clustering unit for classifying the laser scan data included in the candidate regions into a plurality of clusters, and a plurality of clusters classified in the clustering unit, most likely to be the object. A cluster selection unit for selecting a high cluster, and the candidate And a correction value calculator which updates the set correction value of the set physical piece so as to reduce an error of a region selected by the cluster selector and transmits the correction value of the set physical piece to the candidate boundary determiner.

In addition, the method for measuring the physical convenience correction value between the laser scanner and the video camera of the tracking system according to an embodiment of the present invention, a video camera that continuously acquires the image image including the object and the background, and at an angle to the object and the background A method of measuring physical correction value between a laser scanner and a video camera of a tracking system having a laser scanner that continuously acquires laser scan data by measuring a distance according to a distance, the foreground image image having a background region removed from the image image, and a setting A candidate boundary determination step of extracting a candidate region from a physical convenience correction value, a clustering step of classifying the laser scan data included in the candidate area into a plurality of clusters, and a possibility of the object among the plurality of clusters classified in the clustering step A cluster selector that selects this highest cluster And will be in the candidate region and the cluster selecting step characterized by updating the correction value setting physical comfort to reduce the error of the selected cluster, including a correction value calculation step of performing repeat determining the candidate boundary.

As described above, according to the present invention, the positional deviation between the laser scanner and the video camera is corrected in real time, so that the position of the person can be strongly tracked.

Hereinafter, an apparatus and method for calibrating physical convenience between a laser scanner and a video camera, and an apparatus and method for measuring physically corrected values according to an embodiment of the present invention will be described in detail.

FIG. 1 is a block diagram showing an example of a tracking system in which a video camera and a laser scanner are fused to which the present invention is applied. The tracking system includes a video camera 11 for continuously acquiring an image image including an object and a background, a laser scanner 12 for continuously obtaining laser scan data by measuring a distance according to an angle with respect to the object and a background, A foreground image extractor 13 extracting the foreground image image by removing the background region from the image image, a foreground scan data extractor 14 extracting the foreground distance data by removing the background region from the laser scan data, and a foreground image It includes a motion tracker 15 for tracking the movement of the object from the image and the foreground scan data.

The physical bias correction device between the laser scanner and the video camera of the present invention is provided to the motion tracking unit 15 after correcting the physical bias between the laser scanner and the video camera by using the foreground image image and the scan data.

FIG. 2 is a block diagram showing a device 20 for calibrating the physical convenience between the laser scanner and the video camera of the present invention.

First, the operation of the tracking system of FIG. 1 will be described. The video camera 11 continuously photographs video images. The foreground image extracting unit 13 extracts the foreground image image by removing the background region from the photographed image image, and compares the photographed image image with a previously stored background image to remove the same region as the background image from the photographed image image. Only the foreground image is left. 3 is a view showing an image image a and a foreground image image b from which a background area is removed from the image image a.

The laser scanner 12 continuously scans the distance data according to the angle and outputs the laser scan data. The foreground scan data extracting unit 14 extracts the foreground scan data by removing the background area from the laser scan data. The principle that the foreground scan data extractor 14 removes the background area from the laser scan data is the same as the principle of the foreground image extractor 13 removing the background area from the image image.

The physical bias correction device 20 between the laser scanner and the video camera of the present invention corrects the physical bias between the laser scanner and the video camera by using the foreground image image and the laser scan data.

The physical bias correction device 20 between the laser scanner and the video camera of the present invention classifies a candidate boundary determination unit 21 for determining the boundary of the candidate region from the foreground image image, and classifies the laser scan data located in the candidate region into a plurality of clusters. The clustering unit 22, a cluster selecting unit 23 for selecting a cluster that meets a predetermined condition among a plurality of clusters classified in the clustering unit 22, and a direction for reducing an error between the candidate region and the region of the selected cluster. It includes a physical convenience correction unit 24 for correcting the angle of the foreground image image or the foreground distance data.

)은 수학식 1과 같고 후보영역의 오른쪽 경계각(

)은 수학식 2와 같다.The candidate boundary determiner 21 determines a candidate region in which the object exists in the laser scan data from the position information of the object such as an object or a person in the foreground image image.

) Is the same as Equation 1, and the right boundary angle (

) Is the same as Equation 2.

는 비디오카메라에서 촬영된 영상이미지의 폭이고

은 전경영상이미지의 물체의 왼쪽 픽셀 위치값이고

은 해당 물체의 오늘쪽 픽셀 위치값이다.

는 후보영역을 확장하기 위한 확장인자로서, 전경영상이미지로부터 얻어진 물체의 왼쪽 경계와 오른쪽 경계를 각각 확장하여 후보영역이 결정되도록 한다. bias는 설정 물리적편의보정치로서, 초기값은 0으로 설정되고 후술하는 물리적편의보정치 측정방법에 따라 실제 물리적 편의를 보정하기 위한 보정값으로 수렴된다.Where fov is the viewing angle of the video camera,

Is the width of the image taken by the video camera

Is the left pixel position value of the object in the foreground image.

Is the pixel position of today's object.

Is an extension factor for extending the candidate region, and the candidate region is determined by extending the left boundary and the right boundary of the object obtained from the foreground image image, respectively. The bias is a set physical bias correction value. The initial value is set to 0 and converged to a correction value for correcting the actual physical bias according to the method for measuring the physical bias correction value described later.

4 is a diagram showing candidate areas determined by the candidate boundary determination unit 21 of the present invention. When two people exist in the foreground image image, two candidate areas are detected, and FIG. 4 illustrates a state in which the two candidate areas overlap.

The clustering unit 22 classifies the laser scan data into a plurality of clusters using a neighbor neighborhood clustering algorithm in the candidate area. The neighbor setting boundary condition used in this nearest neighbor clustering technique is shown in Equation 3 below.

는 각도 (

)에서의 레이저스캔데이터로서 거리값을 의미한다.

과

는

과

사이의 임의의 각도값이다.

는 근접이웃설정경계값으로서, 이 발명의 실시예에서는 사람의 평균 보폭인 2.5피드(feet)로 설정한다.here,

Is the angle (

Means distance value.

and

Is

and

Any angle value in between.

Is the neighborhood setting boundary value, which is set to 2.5 feet, which is the average stride length of a person in the embodiment of the present invention.

That is, according to Equation 3 and the nearest neighbor clustering technique, this clustering is performed by classifying the regions in which the difference between the laser scan data at any two angles obtained from the laser scan data among candidate regions is smaller than the neighboring neighbor set boundary value. Through the section 22 a plurality of clusters are created.

The cluster selector 23 selects a cluster most likely to be an object (object and / or human) among the plurality of clusters bundled in the clusterer 22. Among the plurality of clusters bound by the clustering unit 22, the background may be included as one cluster. Therefore, the cluster selector 23 applies the plurality of clusters to criteria such as the population number of the cluster, the difference from the background model, the difference from the center angle of the candidate region, and the like, and is most likely the object (object and / or person). Choose a high cluster

That is, the discrimination function for selecting the most suitable cluster among the plurality of clusters is shown in Equation 4.

는 수학식 5와 같다. arg는

가 최대인 i값을 m값으로 리턴한다. 즉, 군집을 구성하는 개체수가 많을수록 그 군집이 채택될 가능성이 높다.Here, k means the number of clusters, Ci means the i th cluster,

Is the same as Equation 5. arg is

Returns the maximum value of i as m. In other words, the larger the number of individuals constituting a cluster, the more likely that cluster is to be adopted.

는 각도에 따른 배경 모델(

) 중 가장 먼 거리값을 의미하는 바, 이를 수식으로 표현하면 수학식 6과 같다.

는 각 각도(θ)에서 측정된 레이저스캔데이터(

)와 배경모델(

) 간 차이값을 의미하는 바, 이를 수식으로 표현하면 수학식 7과 같다.

는 후보영역의 경계각의 폭을 의미하는 바, 이를 수식으로 표현하면 수학식 8과 같다.

는 후보영역의 경계각의 중심각을 의미하는 바, 이를 수식으로 표현하면 수학식 9와 같다.In Equation 5,

Is the background model based on the angle (

) Means the farthest distance value, which is expressed by Equation 6 below.

Is the laser scan data measured at each angle

) And background model (

) Means a difference value, and it is expressed by Equation 7 below.

Denotes the width of the boundary angle of the candidate region, which is expressed by Equation (8).

Denotes the center angle of the boundary angle of the candidate region, which is expressed by Equation (9).

에 따르면 군집에 포함된 각도(θ)가 후보영역의 경계각의 중심각에 가까울수록

는 커지고, 군집에 포함된 각도(θ)가 후보영역의 경계각의 중심각으로부터 멀수록

는 작아진다.That is, the equation (5)

According to, the closer the angle θ to the center angle of the boundary angle of the candidate region is,

Becomes larger, and the angle θ included in the cluster is farther from the center angle of the boundary angle of the candidate region.

Becomes smaller.

이 δ 이상이면

에

가 곱해지기 때문에

는 매우 커지고,

이 δ 이상이 아니면

에

가 곱해지기 때문에

는 작다. 즉,

는 레이저스캔데이터가 배경과의 거리가 일정거리 이상 떨어지면 그 값은 커진다.together,

Is greater than or equal to δ

on

Is multiplied by

Is getting very large,

Is not greater than δ

on

Is multiplied by

Is small. In other words,

The value increases when the laser scan data drops more than a certain distance from the background.

)은 가구들에 의해 변화될 수 있으며, 따라서 배경모델을 학습할 필요가 있다. 배경모델은 전통적인 흑백영상 배경모델을 생성하는 기법 과 유사하게 생성되는 바, 배경모델의 학습을 위한 수식은 아래의 수학식 10과 같다.Background model

) Can be changed by the households, so it is necessary to learn the background model. The background model is generated similarly to the method for generating a background image of a traditional black and white image. The equation for learning the background model is expressed by Equation 10 below.

In Equation 10, α is a learning rate of foreground data, β is a learning rate of background data, and β is generally set to a value 100 times larger than α.

Two pairs of footprints 41 and 42 in the two dashed triangles in FIG. 4 mean a cluster selected through the discriminant functions of Equations 4 to 9 in the candidate areas a and b. According to the present invention, even if candidate regions overlap, an appropriate cluster can be selected.

The physical convenience correction unit 24 adaptively corrects the physical bias in the horizontal direction between the laser scanner and the video camera by using a result of determining a cluster most suitable for an object by applying a discrimination function to one candidate area. 5 is a diagram showing the principle of the physical convenience correction unit 24 to correct the physical convenience.

The foreground image image or the laser scan data in a direction of reducing the difference between the center angle of the candidate area determined by the candidate boundary determiner 21 and the center angle of the cluster selected by the cluster selector 23. Move it. 6 is a diagram illustrating a process of correcting a physical bias in a horizontal direction between a laser scanner and a video camera by moving a foreground image image, (a) shows an image image, and (b) shows a foreground image image. , (c) is a diagram illustrating the correction of the physical bias with respect to the foreground image image.

FIG. 7 is a diagram illustrating an apparatus for measuring correction values of physical pieces between a laser scanner and a video camera according to the present invention.

The physical convenience correction value measuring apparatus includes a candidate boundary determination unit 71 that determines the boundary of the candidate region using the foreground image image and the set physical convenience correction value, and a clustering unit 72 that classifies the laser scan data located in the candidate region into a plurality of clusters. ), And a cluster selection unit 73 for selecting a cluster that meets a predetermined condition among a plurality of clusters classified by the clustering unit 72, and using the error between the candidate region and the region of the selected cluster. And a correction value calculator 74 provided to the candidate boundary determiner after the update.

The functions of the candidate boundary determination unit 71, the clustering unit 72, and the cluster selecting unit 73 are the functions of the candidate boundary determining unit 21, the clustering unit 22, and the cluster selecting unit 23 of FIG. 2, respectively. same.

The correction value calculator 74 updates the set physical bias value in the direction of reducing the center angle of the candidate region and the center angle of the selected cluster.

Where ρ is the learning rate and MidAng (m) is the center angle of the selected cluster.

는 군집을 이루는 각도(θ) 중 최소값이고,

는 군집을 이루는 각도(θ) 중 최대값이다.here,

Is the minimum of the clustering angles θ,

Is the maximum value of the angle θ of the cluster.

In this way, the correction value calculator 74 updates the correction value of the set physical convenience and provides it to the candidate boundary determination unit 71, and the operation of the candidate boundary determination unit 71 to the correction value calculation unit 74 for the foreground image image of the next frame. Repeatedly, the set physical bias correction value converges to the actual physical bias correction value.

[Experiment result]

In order to evaluate the performance of the proposed system, relative physical bias corrections from the origin were measured by rotating the horizontal angle of the video camera by 3 degrees, and the experimental results are shown in Table 1. In this experiment, measurements were performed five times, and Table 1 below records the average values.

Learning rate (ρ) Angle 6 3 -3 -6 0.1 6.013 2.951 -3.139 -6.178 0.5 5.981 2.986 -3.080 -6.186

As can be seen from Table 1, according to the present invention, the average error between the set bias of the set physical bias and the correction value of the actual physical bias is 0.085 degrees. The degree of convergence with time of the correction value of the set physical convenience is shown in FIG. 8. In FIG. 8, LR means a learning rate ρ. When the learning rate is 0.5, it can be seen that the convergence time is less than 10 frames.

Therefore, when the variation amount of the set physical bias correction value (bias) is less than or equal to the threshold value, the set physical bias correction value is set as the actual physical bias correction value, and the physical bias between the video camera and the laser scanner can be corrected using this.

The technical spirit of the present invention has been described above with reference to the accompanying drawings, but this is by way of example only for describing the best embodiment of the present invention and not for limiting the present invention. In addition, it is a matter of course that various modifications and variations are possible without departing from the scope of the technical idea of the present invention by anyone having ordinary skill in the art.

1 is a block diagram showing an example of a tracking system incorporating a video camera and a laser scanner to which the present invention is applied;

2 is a block diagram showing a device for compensating physical convenience between the laser scanner and the video camera of the present invention;

3 is a view showing an image image (a) and a foreground image image (b) from which a background area is removed from the image image (a);

4 is a diagram showing candidate areas determined by the candidate boundary determination unit 21 of the present invention;

5 is a diagram showing the principle of the physical convenience correction unit 24 to correct the physical convenience,

6 is a diagram illustrating a process of correcting a physical bias in a horizontal direction between a laser scanner and a video camera by moving a foreground image image;

7 is a view showing an apparatus for measuring correction value of physical convenience between a laser scanner and a video camera according to the present invention;

8 is a graph showing the degree of convergence over time of the correction value of the set physical convenience according to the present invention.

Claims (20)

The laser scanner and the video of the tracking system having a video camera for continuously acquiring an image image including the object and the background, and a laser scanner for continuously acquiring laser scan data by measuring a distance according to an angle with respect to the object and the background. In the compensation device for physical convenience between cameras, A candidate boundary determination unit extracting a candidate region of the object from the foreground image image from which the background region is removed from the image image; A clustering unit classifying the laser scan data included in the candidate area into a plurality of clusters; A cluster selection unit for selecting a cluster most likely to be the object among a plurality of clusters classified in the clustering unit; A laser scanner comprising a physical convenience correction unit for correcting the physical bias by moving at least one of the image image and the laser scan data so as to reduce an error between the candidate region and the cluster selected by the cluster selector; Physical convenience compensation device between video cameras. 2. The apparatus of claim 1, wherein the clustering unit groups the regions in which the distance difference between the laser scan data at any two angles belonging to the candidate region is smaller than a threshold value into clusters. .  The method of claim 1, wherein the cluster selection unit is likely to be the object by applying a selection measure of at least one of the population of the cluster, the difference from the background model, and the difference from the center angle of the candidate region. Physical convenience correction device between the laser scanner and the video camera, characterized in that the selection of the highest cluster. The apparatus of claim 3, wherein the cluster selector learns the background model by using a distance difference between the background model and the laser scan data. 5. The method of claim 1, wherein the physical convenience correction unit moves the position of at least one of the image image and the laser scan data such that a difference angle between the center angle of the boundary of the candidate region and the center angle of the boundary of the selected cluster is reduced. Features correction device between the laser scanner and the video camera. The laser scanner and the video of the tracking system having a video camera for continuously acquiring an image image including the object and the background, and a laser scanner for continuously acquiring laser scan data by measuring a distance according to an angle with respect to the object and the background. In the method for correcting physical bias between cameras, A candidate boundary determination step of extracting a candidate region of the object from the foreground image image from which the background region is removed from the image image; A clustering step of classifying the laser scan data included in the candidate area into a plurality of clusters; A cluster selection step of selecting a cluster most likely to be the object among a plurality of clusters classified in the clustering step; And a physical convenience correction step of correcting the physical bias by moving at least one of the image image and the laser scan data to reduce the error between the candidate region and the cluster selected in the cluster selection step. Physical convenience correction method between video camera and video camera. The method of claim 6, wherein in the clustering step, physical bias correction between the laser scanner and the video camera is characterized by grouping the regions in which the distance difference between the laser scan data at any two angles belonging to the candidate region is smaller than a threshold value. Way.  The method of claim 6, wherein the cluster selection step is the possibility of being the object by applying at least one selection measure of the population, the difference from the background model, and the difference from the center angle of the candidate region to the plurality of clusters. A method of correcting physical convenience between a laser scanner and a video camera, characterized by selecting the highest cluster. 10. The method of claim 8, wherein the cluster selection step trains the background model using a distance difference between the background model and the laser scan data. The method of claim 6, wherein the correcting of the physical piece comprises moving at least one of the image image and the laser scan data such that a difference angle between the center angle of the boundary of the candidate region and the center angle of the boundary of the selected cluster is reduced. A physical convenience correction method between the laser scanner and the video camera, characterized in that. The laser scanner and the video of the tracking system having a video camera for continuously acquiring an image image including the object and the background, and a laser scanner for continuously acquiring laser scan data by measuring a distance according to an angle with respect to the object and the background. In the measurement device of the correction value of physical convenience between cameras, A candidate boundary determination unit extracting a candidate region of the object from the foreground image image from which the background region is removed from the image image and the set physical convenience correction value; A clustering unit classifying the laser scan data included in the candidate area into a plurality of clusters; A cluster selection unit for selecting a cluster most likely to be the object among a plurality of clusters classified in the clustering unit; Compensation value measurement between the laser scanner and the video camera, characterized in that it comprises a correction value calculation unit for updating the set physical deviation correction value to be transmitted to the candidate boundary determination unit so that the error between the candidate region and the cluster selected by the cluster selection unit is reduced. Device. 12. The method of claim 11, wherein the clustering unit measures physically corrected correction values between the laser scanner and the video camera, characterized in that the areas of the laser scan data at any two angles belonging to the candidate area are smaller than a threshold. Device.  12. The method of claim 11, wherein the cluster selector is likely to be the object by applying at least one selection measure of the population, the difference from the background model, and the difference from the center angle of the candidate region to the plurality of clusters. A device for measuring the correction value of physical convenience between a laser scanner and a video camera, characterized by selecting the highest cluster. The apparatus of claim 13, wherein the cluster selector learns the background model by using a distance difference between the background model and the laser scan data. 12. The laser scanner and the video camera of claim 11, wherein the correction value calculator updates the set physical piece correction value such that a difference angle between the center angle of the boundary of the candidate region and the center angle of the boundary of the selected cluster is reduced. Measurement device for correction of physical convenience. The laser scanner and the video of the tracking system having a video camera for continuously acquiring an image image including the object and the background, and a laser scanner for continuously acquiring laser scan data by measuring a distance according to an angle with respect to the object and the background. In the method of measuring the correction value of physical convenience between cameras, A candidate boundary determination step of extracting a candidate region of the object from the foreground image image from which the background region is removed from the image image and the set physical convenience correction value; A clustering step of classifying the laser scan data included in the candidate area into a plurality of clusters; A cluster selection step of selecting a cluster most likely to be the object among a plurality of clusters classified in the clustering step; And a correction value calculation step of updating the set physical piece correction value so as to reduce the error between the candidate region and the cluster selected in the cluster selection step, and repeatedly performing the candidate boundary determination step. How to measure the correction value. 17. The physical convenience correction value of claim 16, wherein the clustering step groups the regions in which the distance difference between the laser scan data at any two angles belonging to the candidate region is smaller than a threshold value. How to measure.  17. The method of claim 16, wherein the cluster selection step is a possibility of being the object by applying at least one selection measure of the population, the difference from the background model, and the difference from the center angle of the candidate region to the plurality of clusters. A method for measuring the correction value of physical convenience between a laser scanner and a video camera, characterized by selecting the highest cluster. 19. The method of claim 18, wherein in the cluster selection step, the background model is trained using a distance difference between the background model and the laser scan data. The laser scanner and the video camera of claim 16, wherein the calculating of the correction value comprises updating the correction value of the set physical piece so that a difference angle between the center angle of the boundary of the candidate region and the center angle of the boundary of the selected cluster is reduced. How to measure the correction value of liver liver.

Images