KR102043366B1 - Method for measuring trajectory similarity between geo-referenced videos using largest common view - Google Patents

Abstract

The present invention relates to a method for determining similarity between geo-referenced videos and, more specifically, to a trajectory similarity measurement method between geo-referenced videos using a maximum common view. For example, in order to solve the problem of the prior art that conventional similarity measurement algorithms, such as Hausdorff and Longest Common SubSequence (LCSS), have limitations because it is difficult to achieve video classification with semantically similarity simply by considering similarities between movement trajectories, the present invention gives higher similarity values to objects moving while looking at a common space, even in slightly different trajectories, by considering the field of view (FoV) of a geo-referenced video. Therefore, image processing can be done more quickly and accurately than conventional LCSS algorithms, such as grouping using the point of interest, at which a subject of interest such as a photographed person is looking or clustering of the geo-referenced video acquired by a mobile device according to a photographed space.

Description

Method for measuring trajectory similarity between geo-referenced videos using largest common view}

The present invention relates to a method for determining similarity between geo-referenced video (hereinafter also referred to as "GeoVideo"), and more specifically, for example, Hausdorff and the longest common. Conventional similarity measurement algorithms such as Longest Common SubSequence (LCSS) have solved the problems of the prior art, in which the classification of images having semantically similarity is difficult to be achieved simply by considering similarities between movement trajectories. The present invention relates to a method of measuring trajectory similarity between spatial reference images using a maximum common view configured to further extend and improve an existing similarity measurement algorithm by considering a gaze view (FoV) of a spatial reference image.

In addition, in the present invention, in order to further extend and improve the existing similarity measuring algorithm, even if a slightly different trajectory is moved in consideration of the field of view (FoV) of the spatial reference image, the user moves while looking at the common space. It is configured to give a higher similarity value to one object, such as grouping by using the point of interest that the subject of interest, such as the photographed person is viewing, clustering the spatial reference images obtained by the mobile device according to the photographed space, etc. The present invention relates to a method for measuring trajectory similarity between spatial reference images using a maximum common view, which is configured to perform the same image processing more quickly and accurately than the conventional LCSS algorithm.

Recently, with the development of miniaturization technologies such as GPS, accelerometer and gyro sensor, generation of geo-referenced video (Geo Video) acquired by mobile devices such as smart phones, dash cams (black boxes), etc. is gradually increased. It is increasing.

That is, generally, the spatial reference image refers to an image including geographic position at which the image is captured, information about which direction the image is taken, and geographic and positional information such as a photographing angle of the camera. .

Specifically, while the existing general video includes only video frames, for example, video captured by a smartphone, a dashcam (black box) video attached to a general vehicle, or installed in a specific place. Like video obtained by CCTV, GeoVideo includes location information such as latitude and longitude of the video, information about which direction the video was taken, and the camera's shooting angle. Contains information.

In addition, since the spatial reference image can be easily combined with spatial information through GPS or spatial coordinates, using the spatial information of the spatial reference images, spatial interests and regions / points are discovered. Can be used as the base data for.

In addition, in order to utilize the spatial information of the spatial reference images as described above, a process of classifying the images having similarities with each other by determining the similarity between the spatial reference images is required. For this purpose, for example, Hausdorff Similarity measurement algorithms, such as (Hausdorff) and Longest Common SubSequence (LCSS), are widely used.

Here, as an example of the prior art using the similarity measuring algorithm as described above, for example, according to Korean Patent Laid-Open Publication No. 10-2015-0071781, the image input unit for receiving an input image; An object movement trajectory generation unit for generating an object movement trajectory by tracking an object in the input image received by the image input unit; A trajectory model generation unit for calculating a similarity using the Hausdorff distance between the moving trajectories of the objects generated by the object moving trajectory generation unit and generating a trajectory model according to the direction of the moving trajectory of the object; And a movement trajectory analysis unit for analyzing a movement trajectory of the target in the analysis target image received by the image input unit by using the trajectory model generated by the trajectory model generator and determining whether the target is in normal behavior according to the movement trajectory analysis result of the target. Therefore, a description has been made of a trajectory transformation-based moving trajectory modeling apparatus configured to detect abnormal behavior by analyzing the progress direction and the direction of the trajectory.

Further, another example of the prior art using the similarity measurement algorithm as described above, for example, according to Korean Patent Publication No. 10-1789979, extracting Harris corners from the visible image and the infrared image; Classifying Harris corners into a plurality of corner subsets according to the gradient direction of each corner and calculating a weighted-Husdorf distance based on the Hausdorff distances of the corner subsets. A description has been made of a method for calculating the Hausdorff distance based on gradient direction information that is configured to accurately calculate and improve matching accuracy of multi-sensor images.

As described above, in the related art, various technical contents for measuring the similarity between spatial reference images have been presented, but the contents of the prior art as described above have the following problems.

More specifically, the existing similarity measurement algorithms such as Hausdorff and LCSS generally only consider similarities between movement trajectories, that is, distances and angles between the lines constituting the trajectory, There was a limitation in that information such as the eye direction and the camera angle of the camera was not considered at all.

That is, for example, even in the case of the same trajectory, images having opposite directions should be judged to have relatively low similarities. However, the conventional similarity measuring algorithms described above consider only distances between two trajectories. There is a problem that the closer the distance is, the higher the similarity is.

Also, for example, assuming a case where an image of a plurality of mothers and children's groups moving hand in hand in a shopping mall is taken, the existing algorithms as described above merely measure the similarity of the trajectories, and thus the same group is generated. It could be useful to find, but it was also difficult to apply to image processing that clusters or classifies photographed people or animals, such as classifying mothers and children, which are more essential characteristics.

Therefore, as described above, in order to solve the problems of similarity measurement algorithms, such as Hausdorff and LCSS, which had limitations in considering only the similarity between movement trajectories, the gaze view (FoV) of the spatial reference image In consideration of the above, it is desirable to provide a similarity measuring method of a new configuration that is configured to further extend and improve the existing similarity measuring algorithm, but an apparatus or method that satisfies all such requirements has not been proposed until now.

[Preceding technical literature]

1. Korean Patent Publication No. 10-2015-0071781 (2015.06.29.)

2. Korean Registered Patent Publication No. 10-1789979 (2017.10.19.)

The present invention is intended to solve the problems of the prior art as described above, and therefore, the object of the present invention is to limit the image classification with semantically similarity simply by considering only the similarity between the moving trajectories. In order to solve the problems of similarity measurement algorithms such as Hausdorff and LCSS, the maximum common view is configured to further extend and improve the existing similarity measurement algorithm by considering the gaze view (FoV) of spatial reference image. An object of the present invention is to provide a method for measuring trajectory similarity between spatial reference images.

In addition, another object of the present invention is to move while looking at the common space even if slightly different trajectories are moved in consideration of the gaze view (FoV) of the spatial reference image in order to further expand and improve the existing similarity measurement algorithm as described above. It is configured to give a higher similarity value to one object, such as grouping by using the point of interest that the subject of interest, such as the photographed person is viewing, clustering the spatial reference images obtained by the mobile device according to the photographed space, etc. It is an object of the present invention to provide a method of measuring trajectory similarity between spatial reference images using a maximum common view configured to allow the same image processing to be performed more quickly and accurately than the conventional LCSS algorithm.

In order to achieve the above object, according to the present invention, the similarity (similarity) between the spatial reference images is determined in consideration of the field of view (FoV) of the geo-referenced video (GeoVideo) In the method for measuring the trajectory similarity between spatial reference images using a maximum common view configured to be performed by a computer or dedicated hardware, the processing includes: each gaze view for two spatial reference images (GeoVideo). Calculating a common view weight (CVW) for calculating a common view weight (CVW) for the FoV; A similarity distance calculating step of calculating a maximum common view based similarity distance with respect to the two spatial reference images using the common view weight CVW; Calculating a similarity function using the similarity distance to calculate a maximum common view based similarity function for the two spatial reference images; And calculating a distance function (Distance) for calculating a common view subsequence distance function for the two spatial reference images by using the similarity function. A method of measuring trajectory similarity between the two is provided.

The calculating of the common view weight CV may be performed such that a process of calculating the common view weight CVW between the View fov _i and the View fov _j is performed using the following equation. It is done.

(Where View (fov _i ) and View (fov _j ) are functions representing the gaze view area generated by fov _i and fov _j , respectively, and fov _i = (p _i , r _i , θ _i , δ _i ) , fov _j = (p _j , r _j , θ _j , δ _j ), p _i and p _j are the camera positions at which the spatial reference frames obtained by GPS are taken, and r _i and r _j are respectively The maximum distance that can be included in the image taken by the camera, θ _i and θ _j are the angles of view of the camera expressing the direction of the camera with respect to the north, respectively, and δ _i and δ _i are respectively photographed by the camera lens. Indicates the maximum horizontal angle)

The similarity distance calculating step may be configured to perform a process of calculating a maximum common view based similarity distance for the two spatial reference images using the following equation.

(Where, when two spatial reference images GVa and GVb have m and n fovs, respectively, A = {fov ₁ , ..., fov _m )} and B = {fov ₁ ,... , fov _n )}, Head (A) = {fov ₁ ,... , fov _m-1 )} and Head (B) = {fov ₁ ,... , fov _n-1 )}, where σ is a constant representing the minimum time limit)

The calculating of the similarity function may be performed such that a process of calculating a maximum common view based similarity function between A and B of the two spatial reference images is performed using the following equation.

Further, the distance calculation step may be configured to perform a process of calculating the distance function by using the following equation.

Further, the measuring method, each step calculates said common view weight (CVW), and the View (fov _i) and in place of the View (fov _j), with respect to the View (fov _i) and the View (fov _j) Calculation by calculating the common view weights CVW using the functions representing the gaze view area for the derived minimum bounding triangle (MBT) area (ViewTraingle (fov _i ) and ViewTraingle (fov _j )) It is characterized in that it is configured to simplify.

In addition, according to the present invention, there is provided a computer-readable recording medium having recorded thereon a program configured to execute a trajectory similarity measuring method between the spatial reference images described above on a computer or dedicated hardware.

As described above, according to the present invention, the maximum common view is configured to give higher similarity values to the moved objects while looking at the common space even though a slightly different trajectory has been moved in consideration of the eye view (FoV) of the spatial reference image. By providing a method for measuring the trajectory similarity between spatial reference images using an image, grouping by using a point of interest viewed by a subject of interest, such as a photographed person, or clustering spatial reference images acquired by a mobile device according to a photographed space Such image processing can be performed more quickly and accurately than the conventional LCSS algorithm.

In addition, according to the present invention, by considering the gaze view (FoV) as described above, by using the point of interest that the point of interest is grouped or clustered according to the photographed space can be configured to be carried out quickly and accurately By providing a method for measuring trajectory similarity between spatial reference images using the maximum common view, Hausdorff of the prior art had a limitation in that image classification with semantically similarity was difficult to be achieved simply by considering similarity between moving trajectories. And solve the problems of similarity measurement algorithms, such as LCSS, and can further extend and improve the existing algorithm.

1 is a diagram illustrating an example of a spatial reference image (GeoVideo) having trajectory similarity.
2 is a diagram schematically showing an example of a FoV model in GeoVideo.
FIG. 3 is a diagram schematically illustrating a concept of simplification of view (fov _i ) region calculation using ViewTraingle (fov _i ).
4 is a diagram schematically showing the overall configuration of a method for measuring trajectory similarity between spatial reference images using a maximum common view according to an embodiment of the present invention.

Hereinafter, a detailed embodiment of a trajectory similarity measuring method between spatial reference images using the maximum common view according to the present invention will be described with reference to the accompanying drawings.

Here, it should be noted that the contents described below are only one embodiment for carrying out the present invention, and the present invention is not limited to the contents of the embodiments described below.

In addition, in the following description of the embodiments of the present invention, the same or similar to the contents of the prior art, or the part judged to be easily understood and implemented at the level of those skilled in the art, the detailed description for simplicity of explanation Note that omit.

That is, the present invention, as will be described later, similarity measurement algorithms such as Hausdorff and LCSS of the prior art that it was difficult to achieve the image classification having a semantic similarity simply by considering only the similarity between the movement trajectories In order to solve the problem, a method of measuring the trajectory similarity between spatial reference images using the maximum common view configured to further extend and improve the existing similarity measuring algorithm considering the eye view of the spatial reference image (FoV) will be.

In addition, the present invention, in order to further expand and improve the existing similarity measurement algorithm, as described later, even if the trajectory slightly moved in consideration of the gaze view (FoV) of the spatial reference image, the object moved while looking at the common space Images to be grouped by using a point of interest viewed by a subject of interest, such as a photographed person, or clustering spatial reference images acquired by a mobile device according to a photographed space. The present invention relates to a method for measuring trajectory similarity between spatial reference images using a maximum common view, which is configured to be processed more quickly and accurately than the conventional LCSS algorithm.

Subsequently, a detailed description of a trajectory similarity measuring method between spatial reference images using the maximum common view according to the present invention will be described with reference to the drawings.

First, referring to FIG. 1, FIG. 1 is a diagram illustrating an example of a spatial reference image (GeoVideo) having trajectory similarity.

In FIG. 1, geovideo 1, geovideo 2 and geovideo 3 show examples of different geovideos with similar trajectories, for example, where three people move in the same direction at an attraction. It can be assumed to shoot video.

Here, in the existing distance-based similarity algorithms such as Hausdorff and LCSS, only the distance between the trajectories is taken into consideration. Therefore, the most similar trajectory to geovideo 1 is determined to be geovideo 2, but considering the video eye view (FoV), geovideo 1 and More semantic similarity is geovideo 3.

Also, for example, in the case of images of a large number of mothers and children's groups moving hand in hand in a shopping mall, the existing algorithm is useful for finding the same group, while mothers and children are more essential characteristics. It is very difficult to distinguish between them, but considering the visual view (FoV) of the image, it is possible to distinguish the group of mothers and children using the common FoV passed by two objects while looking at different stores according to their interests. It is expected to be.

More specifically, the most common algorithms that can be applied to measure similarity between GeoVideos are distance-based similarity measurement algorithms, such as Euclidean distance, Hausdorff distance-based algorithm, or Longest Common SubSequence (LCSS). There are methods using dynamic time warping (DTW), and the common problem of these trajectory similarity measures is considering only the distance and angle between the lines constituting the trajectory, and the visual direction of the image or the camera angle. This is not considered.

Here, the closest to which FoV can be applied is the LCSS algorithm using a common subsequence concept. The LCSS algorithm is a condition that constitutes a common subsequence, and the distance limit between two points on two trajectories. Use the value ε and the time limit value δ.

However, the LCSS algorithm has a problem that when the estimated distance to be photographed by the camera is set to the distance limit value ε, the value 1 is unconditionally returned when the distance between the FoV points of the two cameras is within ε.

That is, as shown in FIG. 1, even if the geovideo data is within a certain distance ε or even the same trajectory, it may be an image including the opposite space, in which case a value of 0 should be returned.

Accordingly, in the present invention, in order to improve the problems of the prior art as described above, a method of extending the LCSS to return a weight as much as a common view between two geovideo for calculating the similarity between geovideo data.

More specifically, referring to FIG. 2, FIG. 2 is a diagram schematically showing an example of a FoV model in GeoVideo.

That is, a general video can be represented as an ordered collection of video frames, whereas Geovideo consists of spatial reference frame information for each frame along with a general video frame.

Here, when a set of spatial reference frame information is called GV, one spatial reference video information GV = {gv _i | ∀gv _i ∈GV}, which is a video frame having a spatial reference value in each video frame gv It can be expressed as an ordered collection of these.

That is, given a spatial reference frame gvi, gv _i = (f _i , fov _i ), where f _i is information for pointing to a specific video frame on the actual video, and fov _i = (p _i , r _i , θ _i , δ _i ).

Also, in FIG. 2, p _i is a location where a spatial reference frame acquired by GPS is captured as a camera location, r _i is a maximum distance that can be included in an image photographed by the camera as a blind distance, and θ _i is The camera viewing angle is an angle expressed by the north facing the camera, and δ _i is a camera lens horizontal angle, which means a horizontal maximum angle that can be photographed by the camera lens.

Here, in general, the camera lens has a vertical maximum angle as well as horizontal, it should be noted that it is omitted here for simplicity.

In addition, a function for deriving the gaze view area generated by certain fov _i and fov _j is called View (fov _i ) and View (fov _j ), respectively, between two View (fov _i ) and View (fov _j ). The function CVW for obtaining the common view weight (CVW) of may be expressed by Equation 1 below.

[Equation 1]

Here, in the above Equation 1, | View (fov _i )) ovView (fov _j ) | is the area of the space generated by the union operation for the two spaces, and | (View ( The result of fov _i )) ∩View (fov _j ) | means the area of space created by the intersection operation on two spaces.

If there is no common area between two View (fov _i ) and View (fov _j ), CVW (fov _i , fov _j ) will be 0. If there is a common view, it will be weighted by the ratio of the common view.

Further, assuming that there are two spatial reference video information GVa and GVb, and A and B have m and n fovs derived from GVa and GVb, respectively, A = {fov ₁ ,... , fov _m )} and B = (fov ₁ ,... , fov _n )}, and Head (A) is Head (A) = {fov ₁ ,... , fov _m-1 )}.

In addition, when the minimum time limit value σ is given, the maximum common view based similarity distance algorithm LCVSσ (A, B) for A and B may be defined as Equation 2 below.

[Equation 2]

Here, in Equation 2 above, the constant σ is a constant for controlling how far apart two geovideos can be used for matching until in time.

In addition, when a minimum time limit value σ is given, a similarity function between A and B of FoV of two geovideos may be defined as in Equation 3 below.

[Equation 3]

That is, in the above Equation 3, since the maximum value that LCVSσ (A, B) can return is min (m, n), the basic Similarity function has LCVSσ (A, B) smaller than m and n. Divided by number.

This Similarity function can be extended with some modifications, such as extended to support spatial shift in LCSS, and, given the minimum time limit σ, a common view sub-based on this Similarity function. The sequence distance function Distance may be defined as shown in Equation 4 below.

[Equation 4]

In the above Equation 4, since LCVSσ (A, B) and LCVSσ (B, A) are the same, the distance function Distance operates symmetrically regardless of the order.

Subsequently, referring to FIG. 3, FIG. 3 schematically illustrates a concept of simplification of view (fov _i ) region calculation using ViewTraingle (fov _i ).

More specifically, the most computational cost in the algorithm described above is that View (fov _i ) and cross-domain computation View (fov _i ) View (fov _j ) and summation operation View (fov _i ) ∪View (fov _j )

Accordingly, in the present invention, in order to minimize such a calculation cost, as shown in FIG. 3, it may be configured to derive a minimum bounding triangle (MBT) region rather than an arcuate region using ViewTriangle (fov _i ). have.

Therefore, it is possible to implement a method for measuring trajectory similarity between spatial reference images using the maximum common view according to the embodiment of the present invention, whereby, according to the present invention, even if a slightly different trajectory is moved, the common space can be implemented. It is configured to give a higher similarity value to the moved objects while viewing the video, thereby grouping by using a point of interest viewed by the subject of interest, such as a person photographed on the video, or by recording spatial reference videos acquired by the mobile device. It can be very usefully applied to image processing such as clustering according to space.

More specifically, referring to FIG. 4, FIG. 4 is a diagram schematically illustrating an overall configuration of a method for measuring trajectory similarity between spatial reference images using a maximum common view according to an exemplary embodiment of the present invention.

In more detail, as shown in FIG. 4, a method of measuring trajectory similarity between spatial reference images using a maximum common view according to an embodiment of the present invention is divided into two categories. First, two spatial reference images are geovideo. A common view weight (CVW) calculating step (S10) for calculating a common view weight (CVW) for each field of view (FoV) and a maximum common view based similarity distance for two spatial reference images are calculated. A maximum common view based similarity distance calculation step (S20), a similarity function calculating step (S30) for calculating a similarity (Similarity) function for two spatial reference images, and two spatial based on the similarity (Similarity) function described above And calculating a common view subsequence distance function (Distance) for calculating a common view subsequence distance function (Distance) for the reference image.

Here, the common view weight (CVW) calculating step (S10), as described above with reference to the above [Equation 1], the common view weight between the two views (fov _i ) and View (fov _j ) ( CVW) can be configured to be performed.

In addition, the maximum common view-based similarity distance calculation step (S20), as described above with reference to [Equation 2], when the minimum time limit value σ is given A and B of two spatial reference images Processing to calculate the maximum common view based similarity distance for may be configured to be performed.

In addition, the similarity function calculating step (S30), as described above with reference to [Equation 3], when the minimum time limit value σ is given, the similarity function between A and B of two spatial reference images May be configured to be performed.

Further, the distance calculation step (S40), as described above with reference to [Equation 4], when the minimum time limit value σ is given, the distance function (Distance) using the similarity function May be configured to be performed.

Here, the processing of each step (S10 ~ S40) can be configured to be performed through a computer or dedicated hardware, for this purpose, the present invention, to the computer to execute the processing of each step as described above It may be configured in the form of a program to be configured or a recording medium on which such a program is recorded.

Accordingly, as described above, a trajectory similarity measuring method between spatial reference images using the maximum common view can be implemented.

In addition, by implementing the method of measuring the trajectory similarity between the spatial reference images using the maximum common view as described above, according to the present invention, slightly different in view of the eye view of the spatial reference image (FoV) Even if the trajectory is moved, a method of measuring trajectory similarity between spatial reference images using the maximum common view configured to give higher similarity values to the moved objects while looking at the common space can be reported by an interested person such as a photographed person. Image processing such as grouping by using a point of interest, or clustering spatial reference images obtained by the mobile device according to the photographed space may be performed more quickly and accurately than the conventional LCSS algorithm.

In addition, according to the present invention, by considering the gaze view (FoV) as described above, by using the point of interest that the interested point of view or grouping according to the photographed space can be configured to be quickly and accurately By providing a method for measuring trajectory similarity between spatial reference images using the maximum common view, Hausdorff of the prior art had a limitation in that image classification with semantically similarity was difficult to be achieved simply by considering similarity between moving trajectories. And solve the problems of similarity measurement algorithms, such as LCSS, and can further extend and improve the existing algorithm.

As described above, the details of the method for measuring the trajectory similarity between spatial reference images using the maximum common view according to the present invention have been described through the embodiments of the present invention. However, the present invention is not limited, and therefore, it is obvious that various modifications, changes, combinations, and substitutions may be made by those skilled in the art according to design needs and various other factors. I will call it work.

Claims (7)

A process for determining similarity between spatially-referenced images in consideration of a field of view (FoV) of geo-referenced video (GeoVideo) is configured to be performed by a computer or dedicated hardware. In the method for measuring trajectory similarity between spatial reference images using maximum common view,
The processing is
Calculating a common view weight (CVW) for calculating a common view weight (CVW) for each gaze view FoV for two spatial reference images GeoVideo;
A similarity distance calculating step of calculating a maximum common view based similarity distance with respect to the two spatial reference images using the common view weight CVW;
Calculating a similarity function using the similarity distance to calculate a maximum common view based similarity function for the two spatial reference images; And
And a distance function calculating step of calculating a common view subsequence distance function for the two spatial reference images using the similarity function. How to measure trajectory similarity of.
The method of claim 1,
The common view weight (CVW) calculating step,
A method of measuring trajectory similarity between spatial reference images, wherein a process of calculating a common view weight CVW between View (fov _i ) and View (fov _j ) is performed by using the following equation.

(Where View (fov _i ) and View (fov _j ) are functions representing the gaze view area generated by fov _i and fov _j , respectively, and fov _i = (p _i , r _i , θ _i , δ _i ) , fov _j = (p _j , r _j , θ _j , δ _j ), p _i and p _j are the camera positions at which the spatial reference frames obtained by GPS are taken, and r _i and r _j are respectively The maximum distance that can be included in the image taken by the camera, θ _i and θ _j are the angles of view of the camera expressing the direction of the camera with respect to the north, respectively, and δ _i and δ _i are respectively photographed by the camera lens. Indicates the maximum horizontal angle)
The method of claim 2,
The similarity distance calculating step,
And a process for calculating a maximum common view based similarity distance for the two spatial reference images is performed using the following equation.

(Where, when two spatial reference images GVa and GVb have m and n fovs, respectively, A = {fov ₁ , ..., fov _m )} and B = {fov ₁ ,... , fov _n )}, Head (A) = {fov ₁ ,... , fov _m-1 )} and Head (B) = {fov ₁ ,... , fov _n-1 )}, where σ is a constant representing the minimum time limit)
The method of claim 3, wherein
The similarity function calculating step,
And a process for calculating a maximum common view based similarity function between A and B of the two spatial reference images is performed using the following equation.

The method of claim 4, wherein
The distance function calculation step,
A method for measuring trajectory similarity between spatially-referenced images according to the following equation, wherein the processing for calculating the distance function is performed.

The method of claim 5,
The measuring method,
In step calculates said common view weight (CVW), and the View (fov _i) and the View (fov _j) Alternatively, the View (fov _i) and the View (fov _j) respectively deriving the minimum bounding triangles (minimum bounding triangle with respect to the (B) calculating the common view weight CVW using the functions representing the gaze view region for the MBT region (ViewTraingle (fov _i ) and ViewTraingle (fov _j )). A method for measuring trajectory similarity between spatial reference images.
A computer-readable recording medium having recorded thereon a program configured to execute a method of measuring the trajectory similarity between spatial reference images according to any one of claims 1 to 6 on a computer or dedicated hardware.

Images