System and method for extracting the location of a moving object

Abstract

The present invention relates to a method for extracting the actual position coordinates of a moving object in a modified image by automatically correcting a preset irregular triangular network (TIN) and map coordinate matching information even when the image is modified, such as by rotation, reduction, enlargement, or change in resolution of the camera image. The method comprises the steps of: storing a camera input image at an arbitrary point in time as a reference image; storing together as reference information TIN-related information, in which an irregular triangular network (TIN) is generated based on the image coordinates of reference points selected by an administrator in the reference image, and information in which the image coordinates of the reference points are matched to actual position coordinates on a map; comparing the current input image of the camera with the reference image to estimate the position information of the vertices constituting the TIN that must be projected onto the current input image; and projecting each vertex point of the estimated TIN onto the current input image to correct the reference information to reference information corresponding to the current input image. The method is characterized by including the step of identifying a triangle containing a representative pixel of a moving object in a current input image projected with the estimated TIN, and calculating and outputting the actual position coordinates of the moving object using the image coordinates of the representative pixel, the image coordinates of each vertex constituting the identified triangle, and actual position coordinate information matched to the image coordinates of each vertex.

Description

Technology Field

[0001] The present invention relates to position tracking of a moving object based on camera images, and more specifically, to a moving object position extraction system and method that automatically extracts the actual position coordinates of a moving object from a changed image by automatically correcting a preset Triangulated Irregular Network (TIN) and map coordinate matching information even when the image is changed, such as by rotation, reduction, enlargement, or change in resolution of the camera image. Background Technology

[0003] Smart City Integrated Operation Centers operated by local governments utilize a solution that extracts the location of moving objects (hereinafter referred to as "moving objects") using security cameras (CCTV).

[0004] Camera image-based moving object location extraction solutions require camera attitude data; however, since local governments often lack this information, applying a solution that extracts the location of moving objects using existing security cameras requires incurring new time and economic costs to separately measure the camera attitude data.

[0005] Therefore, there is a need for a new technology capable of accurately extracting the location of a moving object based on camera images without acquiring attitude information from existing security cameras, and furthermore, without using position extraction techniques utilizing vanishing points or lines that require extensive computation.

[0006] As an invention created in response to this need, there is the ‘Moving Body Position Extraction System and Method (Republic of Korea Registered Patent Publication No. 10-2737650)’ which was previously filed and registered by the applicant of the present application.

[0007] The prior patented invention described above matches a reference pixel on a road edge in a CCTV image with actual location coordinates on a map, generates an irregular triangular network (TIN) for the road section using this, and then extracts the actual location of the moving object through matrix operations or interpolation operations using the coordinates of the representative pixel of the moving object and the reference pixel.

[0008] However, the prior registered patent invention cited above is based on the assumption that the coordinate system of the camera image is fixed, and since the position of the moving object is calculated based on the reference image and TIN generated at the time of initial setting, if the CCTV rotates or zooms with pan-tilt-zoom movements, or if camera replacement, resolution change, aspect ratio change, or image rotation by the operator occurs, the geometric relationship between the reference image and the current image changes.

[0009] In other words, if a change in the image coordinate system occurs as described in the example above, the previously generated TIN and reference pixel matching information cannot be used as is, and it is necessary to designate a new reference point and regenerate the TIN. Since these tasks are typically performed manually, they require a significant amount of time and effort. Because this unnecessary expenditure of time poses a major constraint on real-time or near-real-time location tracking services, a new solution is required to address this issue. Prior art literature

[0011] Republic of Korea Published Patent Application No. 10-2023-0099239 Republic of Korea Registered Patent Application No. 10-2737650 The problem to be solved

[0012] Accordingly, the present invention is an invention conceived in accordance with the aforementioned necessity, and the main objective of the present invention is to provide a system and method for extracting the position of a moving object that can generate a corrected TIN corresponding to a new image without recalibration by automatically correcting the matching information between the existing irregular triangular network (TIN) and image reference pixels and map coordinates, even if a change in the image coordinate system occurs due to the rotation, zooming, or resolution change of the camera image.

[0013] Furthermore, another objective of the present invention is to provide a system and method for extracting the position of a moving object that can continuously and accurately extract the actual position coordinates of the moving object using an automatically corrected TIN without separately measuring internal and external parameters of the camera or performing complex camera calibration procedures.

[0014] In addition, the present invention provides a system and method for extracting the position of a moving object, wherein error evaluation and threshold criteria are introduced during the TIN correction process, so that when the image transformation is small, it is processed solely through automatic correction, and positional accuracy is secured through minimal auxiliary operations, such as reassigning some reference points, only when the image transformation is large or the error exceeds the threshold. means of solving the problem

[0016] A moving body position extraction system according to one embodiment of the present invention for achieving the aforementioned purpose is a system in the form of an application program that is installed and executable on a computer system,

[0017] A reference information setting unit that stores a camera input video image as a reference image, and together stores TIN-related information generated by creating a Triangulated Irregular Network (TIN) based on the image coordinates of reference points selected from the reference image, and information that matches the image coordinates of the reference points to actual location coordinates on a map as reference information;

[0018] A TIN position estimation unit that estimates position information of vertices constituting the TIN to be projected onto the current input image by comparing the current input image of the camera with the reference image, and

[0019] A reference information correction unit that projects each vertex point of the estimated TIN onto the current input image to correct the reference information into reference information corresponding to the current input image, and

[0020] It is characterized by including a moving body position coordinate extraction unit that identifies a triangle containing a representative pixel of a moving body in a current input image projected with the estimated TIN above, and calculates and outputs the actual position coordinates of the moving body using the image coordinates of the representative pixel, the image coordinates of each vertex constituting the identified triangle, and actual position coordinate information matched to the image coordinates of each vertex.

[0021] In a moving object position extraction system comprising the above-described configurations, the reference information correction unit comprises:

[0022] For some of the above reference points, the error between the actual position coordinates before and after reference information correction is calculated, and the estimated TIN is determined to be valid only when it is below a set threshold, thereby selectively activating the moving body position coordinate extraction unit.

[0023] Furthermore, in a moving body position extraction system comprising the aforementioned configurations, the TIN position estimation unit, for each vertex constituting the TIN,

[0024] Another feature is that the position information of each vertex constituting the TIN is estimated by extracting an image of a rectangular region centered on one vertex of the TIN from the reference image, extracting common feature points between the image of the rectangular region and the current input image through a feature point extraction algorithm, extracting the pairs most similar to each other among the extracted feature points as matched points through a feature point matching algorithm, calculating a position coordinate transformation matrix based on the information of the matched points, and calculating the corresponding coordinates in the current image by performing operations on the coordinates of one vertex of the TIN with the position coordinate transformation matrix.

[0025] Furthermore, in the aforementioned moving object position extraction system, the reference information correction unit is,

[0026] For some of the above reference points, if the error between the actual position coordinates before and after reference information correction exceeds a set threshold when calculated, the reference information may be re-corrected through re-estimation of the position coordinate transformation matrix, additional feature point matching, or reassignment of some reference points.

[0027] Meanwhile, a method for extracting the position of a moving object according to another embodiment of the present invention is a method executed by a processor in a computer system that processes camera input images, wherein

[0028] A step of storing a camera input video at an arbitrary point in time as a reference image, and together storing TIN-related information in which an irregular triangular network (TIN) is generated based on the image coordinates of reference points selected by an administrator in the reference image, and information in which the image coordinates of the reference points are matched to actual location coordinates on a map as reference information;

[0029] A step of estimating position information of vertices constituting the TIN to be projected onto the current input image by comparing the current input image of the camera with the reference image, and

[0030] A step of projecting each vertex point of the estimated TIN onto the current input image to correct the reference information into reference information corresponding to the current input image, and

[0031] The method is characterized by including the step of identifying a triangle containing a representative pixel of a moving object in a current input image projected with the estimated TIN, and calculating and outputting the actual position coordinates of the moving object using the image coordinates of the representative pixel, the image coordinates of each vertex constituting the identified triangle, and actual position coordinate information matched to the image coordinates of each vertex.

[0032] In addition, in the above-described method for extracting the position of a moving object, the step of estimating the position information of the vertices constituting the TIN comprises, for each vertex constituting the TIN,

[0033] A step of extracting an image of a rectangular region centered on one vertex of the TIN from the reference image above, and

[0034] A step of extracting common feature points between the image of the rectangular area and the current input image through a feature point extraction algorithm, and

[0035] A step of extracting the pairs of feature points that are most similar to each other as matched points through a feature point matching algorithm, and

[0036] A step of calculating a position coordinate transformation matrix based on the information of the matched points above, and

[0037] The method is characterized by estimating the position information of each vertex constituting the TIN, including the step of calculating the corresponding coordinates in the current image by performing an operation on the coordinates of one vertex of the TIN with the position coordinate transformation matrix.

[0038] In some cases, for some of the above reference points, the error between the actual position coordinates before and after reference information correction can be calculated, and the estimated TIN is determined to be valid only if it is below a set threshold, thereby performing the step of calculating and outputting the actual position coordinates of the moving body. Effects of the invention

[0040] According to the means for solving the technical problem described above, the moving body position extraction system and method according to an embodiment of the present invention are,

[0041] Despite changes in the video coordinate system, such as video rotation, zooming, and resolution changes of security cameras like CCTV, it has the advantage of continuously and accurately extracting the actual location coordinates of moving objects without additional manual work by automatically correcting the reference information generated during initial setup (TIN-related information, video coordinates of reference points, and the actual location coordinates on the map matched thereto).

[0042] In addition, the present invention does not require separate measurement of camera attitude information (internal / external parameters) and is compatible with existing TIN-based position extraction technology, thus having the advantage of being applicable to an extended range without significantly modifying existing moving object position extraction systems.

[0043] Furthermore, through error evaluation and recalibration control functions, the quality of automatic calibration can be quantitatively managed, and accuracy can be maintained with minimal manual intervention only when errors are large, which has the effect of significantly improving maintenance efficiency in actual operating environments.

[0044] In addition, by reducing the time required to reset TINs and reference points in situations of environmental change, such as CCTV replacement, reinstallation, or changes in pan / tilt / zoom patterns, it offers the advantage of providing stable location-based services in various application services, including smart cities, traffic monitoring, public safety, and population flow analysis. Brief explanation of the drawing

[0046] FIG. 1 is an example diagram of the peripheral configuration of a moving body position extraction system according to one embodiment of the present invention. FIG. 2 is an example diagram of the configuration of the moving object position extraction system (200) in FIG. 1. FIG. 3 is a flowchart illustrating a method for extracting the position of a moving body according to an embodiment of the present invention. FIGS. 4 to 8 are example images for further explaining a method for extracting the position of a moving body according to an embodiment of the present invention. Specific details for implementing the invention

[0047] The following detailed description of the present invention refers to the accompanying drawings, which illustrate specific embodiments in which the present invention can be practiced in order to clarify the objects, technical solutions, and advantages of the present invention. These embodiments are described in sufficient detail to enable a person skilled in the art to practice the present invention.

[0048] Furthermore, throughout the detailed description and claims of the invention, the word "comprising" and its variations are not intended to exclude other technical features, additions, components, or steps. Other objects, advantages, and characteristics of the invention will become apparent to a person skilled in the art, in part from this description and in part from the practice of the invention. The following examples and drawings are provided as examples and are not intended to limit the invention. Moreover, the invention encompasses all possible combinations of the embodiments set forth herein. It should be understood that various embodiments of the invention differ from one another but need not be mutually exclusive. It should also be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the following detailed description is not intended to be taken in a limiting sense, and the scope of the invention is limited only by the appended claims, including all equivalents to those claimed therein, provided they are properly described.

[0049] Meanwhile, unless otherwise indicated in this specification or clearly contradictory to the context, items referred to in the singular encompass the plural unless otherwise required by the context. Furthermore, in describing the present invention, if it is determined that a detailed description of related known configurations or functions could obscure the essence of the invention, such detailed description will be omitted.

[0050] First, FIG. 1 illustrates a configuration diagram of a mobile object location extraction system (200) according to an embodiment of the present invention. The mobile object location extraction system (200) according to an embodiment of the present invention receives road section (video) images from a plurality of security cameras (100) and calculates, stores, or displays the actual location coordinates of the mobile object (meaning map coordinates on a map) in real time. This location extraction system (200) can be built and executed in the form of a collection of hardware or control program (or application program) data on a security control server or a security computer system. Although security cameras are exemplified below, it is preferable to understand this as a concept that includes all image acquisition devices capable of acquiring images of a predetermined space as an example.

[0051] The security camera (100) is a CCTV capable of rotation, zooming, and resolution change, and captures an image of a road section (or space) corresponding to the shooting area and transmits it to a moving object location extraction system (200). The administrator I / F (InterFace, 300) displays the video image captured by the security camera (100) and also performs the role of transmitting operation commands by the administrator to the moving object location extraction system (200). This administrator I / F (300) may include a display section and an input section, or it may be a touch screen in which the display section and the input section are integrated.

[0052] The configuration of the moving object location extraction system (200) will be explained in more detail below.

[0053] FIG. 2 illustrates a configuration diagram of the moving object location extraction system (200) in FIG. 1. As shown in FIG. 2, the moving object location extraction system (200) includes a reference information setting unit (210) that sets a road section image at an initial setting time, received and displayed from a security camera (100), as a reference image and stores it in a DB (250) as reference information (along with a reference image ID).

[0054] The reference information setting unit (210) sets points selected by the manager in the reference image (e.g., yellow points shown in FIG. 4) as reference points, generates a Triangulated Irregular Network (TIN) based on the image coordinates (which can be called an image coordinate system) of each reference point, and stores information related to the generated Irregular Network (TIN) (vertex index and location information, triangle ID, etc.) together as part of the reference information.

[0055] In addition, the reference information setting unit (210) stores information that matches the image coordinates of the reference points to actual location coordinates (map coordinate system) on a digital map as part of the reference information.

[0056] For reference, the reference information setting unit (210) includes a TIN generation unit that generates the TIN by triangulating the road section within the reference image when the number of reference points selected by the manager starts to be three or more.

[0057] Meanwhile, the moving object position extraction system (200) includes a TIN position estimation unit (220) that estimates position information of vertices constituting the TIN to be projected onto the current input image by comparing the current input image of the security camera (100) with the reference image. That is, the TIN position estimation unit (220) estimates the camera's rotation angle, scale change, translation, and other geometric deformations by comparing the current input image with the reference image.

[0058] More specifically, the TIN position estimation unit (220) for each vertex constituting the TIN,

[0059] The position information of each vertex constituting the TIN is estimated by extracting an image of a rectangular region centered on one vertex of the TIN from the reference image, extracting common feature points between the image of the rectangular region and the current input image through a feature point extraction algorithm, extracting the pairs of feature points most similar to each other as matched points through a feature point matching algorithm, calculating a position coordinate transformation matrix based on the information of the matched points, and calculating the corresponding coordinates in the current image by performing operations on the coordinates of one vertex of the TIN with the position coordinate transformation matrix.

[0060] Additionally, the moving object position extraction system (200) includes a reference information correction unit (230) that projects each vertex point of the estimated TIN onto the current input image to correct the reference information into reference information corresponding to the current input image, and

[0061] The system further includes a moving body position coordinate extraction unit (240) that identifies a triangle (corresponding to a segmented section) containing a representative pixel of the moving body in the current input image projected with the estimated TIN above in the DB (250), and calculates and outputs or stores the actual position coordinates of the moving body using the image coordinates of the representative pixel, the image coordinates of each vertex constituting the identified triangle, and actual position coordinate information matched to the image coordinates of each vertex.

[0062] For reference, if the segment generation unit is implemented as a TIN generation unit, the moving object position coordinate extraction unit (240) can extract (detect) the moving object from the displayed road segment image, extract the MBR (minimum bounding rectangle) of the moving object, and then set and use a single pixel that can represent the position of the MBR as a representative pixel.

[0063] The unexplained information storage DB (250) includes information storage areas assigned to each security camera (100), and each information storage area stores reference video images set by each security camera (100) and reference information before and after correction.

[0064] The operation of a moving object position extraction system (200) including the configurations described above will be further explained with reference to FIGS. 3 to 8.

[0065] For reference, FIG. 3 is a flowchart illustrating a method for extracting the position of a moving object according to an embodiment of the present invention, and FIGS. 4 to 8 are images illustrating a method for extracting the position of a moving object according to an embodiment of the present invention.

[0066] A moving object position extraction system (200) according to an embodiment of the present invention requires a task of setting a camera image at an initial setting time as a reference image, and a task of matching actual position coordinates for reference points selected on the reference image as a preliminary task by an administrator, and it is necessary for such matching information to be set and stored together as part of the reference information.

[0067] Referring to FIG. 3, first, the reference information setting unit (210) of the moving object location extraction system (200) receives a road section image from the security camera (100) and displays it on the user I / F (300).

[0068] Next, the reference information setting unit (210) sets a pixel of a point selected by the manager in the displayed road section image (e.g., a point located on the edge of the road) as a reference point (step S110), and stores information matching the actual location coordinates of the reference points selected by the manager in the digital map with the image coordinates of the reference point as reference information (step S110). In addition, the reference information setting unit (210) stores the road section image where the reference point was selected (corresponding to the camera image at the time of initial setting) together as a reference image in the DB (250).

[0069] In addition, the TIN generation unit (or segment generation unit) included in the reference information setting unit (210) generates an irregular triangular network (TIN) by triangulating the road segment within the reference image based on the reference points as shown in FIG. 4 when the number of reference points starts to be three or more, and sets the information related to the irregular triangular network (TIN) (vertex index and location information, triangle ID, etc., hereinafter referred to as 'related information') as part of the reference information and stores it together in the DB (250) (step S100).

[0070] For reference, the yellow dots in the image of Fig. 4 are reference points selected by the manager, and the manager simply needs to select the points necessary to divide the road section within the road section image into multiple sections.

[0071] After the setting of reference information for the reference image is completed, if the security camera (100) undergoes changes in rotation angle, zoom, scale (resolution change), aspect ratio, translation, or other geometric deformations due to pan, tilt, zoom movements, or image rotation by an operator, the TIN position estimation unit (220) must estimate the positions of each vertex of the TIN that must be located in the current image input of the security camera (100). That is, by comparing the current image input of the camera (100) with the reference image, the position information of each vertex constituting the TIN that must be projected onto the current image input must be estimated.

[0072] To this end, the TIN position estimation unit (220) sequentially performs the following steps to match the same point between the current input image and the reference image.

[0073] First, the TIN location estimation unit (220) for each vertex constituting the TIN,

[0074] 1) As shown in FIG. 5a, an image of a rectangular area centered on one vertex of the TIN is extracted from the reference image (see FIG. 5b).

[0075] 2) Subsequently, the TIN position estimation unit (220) finds feature points that are invariant to changes in image size, rotation, and brightness (e.g., corner edges, etc.) through a feature point extraction algorithm, e.g., SIFT (Scale-Invariant Feature Transform). That is, it extracts common feature points (such as door frame corners, signboard letters, floor patterns, etc. as shown in FIG. 5b) between the image of the TIN point rectangle area extracted from the reference image and the current input image.

[0076] 3) Next, the TIN position estimation unit (220) extracts the pairs of feature points extracted from the two images that are most similar to each other as matched points through a Brute-Force Matcher, which is one of the feature point matching algorithms (see FIG. 6).

[0077] 4) Next, the TIN position estimation unit (220) calculates a position coordinate transformation matrix (Homography Matrix) that transforms the plane of Image 1 (rectangular area image) into the plane of Image 2 based on the information of the matched points. At this time, an accurate transformation relationship can be found by removing outliers using the RANSAC algorithm.

[0078] 5) Afterwards, the TIN position estimation unit (220) calculates the corresponding coordinates (x', y') in the current image corresponding to Image 2 by performing operations with the position coordinate transformation matrix on the coordinates of the yellow point found in Image 1, that is, the coordinates (x,y) of one vertex of the TIN.

[0079] In this way, the TIN position estimation unit (220) estimates the position of the TIN point (vertex) in the current input image corresponding to the entire reference image by repeating steps 1) through 5) described above for all vertices constituting the TIN, as shown in FIG. 7. In FIG. 7, the yellow dots are reference points in the reference image, and the blue dots connected to each yellow dot represent the estimated TIN point.

[0080] When the estimation of TIN configuration points for the current input image is completed, the reference information correction unit (230) projects each vertex point of the estimated TIN onto the current input image, and corrects the reference information to the reference information corresponding to the current input image (step S120).

[0081] That is, the reference information correction unit (230) applies each vertex point of the estimated TIN (corresponding to a reference point in the image coordinate system) to the image coordinate system, which is the current input image, and corrects the reference image ID, reference point image coordinates, actual location coordinates on the map that match the reference point image coordinates, and triangle information (vertex index, triangle ID, etc.) constituting the TIN into reference information corresponding to the current input image based on the reference point coordinates in the current image coordinate system, and generates the TIN based on the corrected reference information. FIG. 8 is an example of an image in which the estimated TIN is projected onto the current input image, and when compared with FIG. 4, it can be seen that the positions of the vertices constituting the TIN projected onto the current input image are also corrected according to camera rotation.

[0082] Meanwhile, by introducing error evaluation and threshold criteria in the aforementioned reference information correction process, when the image transformation is small, it is processed solely through automatic correction, and when the transformation is large or the error exceeds the threshold, positional accuracy can be secured through minimal auxiliary operations such as reassigning some reference points.

[0083] To this end, the reference information correction unit (230) calculates the error between the actual position coordinates before and after the reference information correction for some of the reference points and determines the estimated TIN as valid only when it is below a set threshold, thereby activating the moving body position coordinate extraction unit (240) to be described later. If the error between the actual position coordinates before and after the reference information correction exceeds the set threshold, the reference information is re-corrected through re-estimation of the transformation matrix, additional feature point matching, or re-designation of some reference points, and if necessary, an auxiliary operation can be requested through the administrator I / F unit (300).

[0084] As explained above, once the reference information set for the initial reference image is corrected to reference information corresponding to the current input image, the actual location coordinates of a moving object (vehicle, pedestrian, etc.) can be automatically calculated in a real-time CCTV video system, etc., based on the corrected reference information.

[0085] More specifically, the moving object position coordinate extraction unit (240) extracts (detects) a moving object that moves first in the current input image projected with the estimated TIN, extracts the MBR (minimum bounding rectangle) area of ​​the extracted moving object, and then selects one pixel that can represent the position of the MBR as a representative pixel.

[0086] Next, the moving body position coordinate extraction unit (240) identifies a triangle (i.e., a segmented section) containing the representative pixel in the DB (250), and then calculates the actual coordinate position of the moving body located within the identified triangle based on the following matrix formula. Since the position information for the reference point corresponding to each vertex constituting the triangle and the actual position coordinates matched thereto are already known, the actual coordinate position (x',y') of the moving body can be calculated based on this according to the following matrix formula. In the following, M can be named as the actual position coordinate transformation matrix.

[0087]

[0088]

[0089]

[0090] x: x-coordinate component of the representative pixel.

[0091] y: y-coordinate component of the representative pixel.

[0092] x1, x2, x3: x-coordinate components of the reference points (reference pixels) corresponding to each vertex in the triangle containing the representative pixel.

[0093] y1, y2, y3: y-coordinate components of the reference point corresponding to each vertex in the triangle containing the representative pixel.

[0094] x1', x2', x3': x-coordinate components of the actual position coordinates matched to each of the above x1, x2, and x3.

[0095] y1',y2',y3': The y-coordinate components of the actual position coordinates matched to each of the above y1, y2, and y3.

[0096] When the actual position coordinates of the moving body are calculated and obtained based on the above-described matrix, the moving body position coordinate extraction unit (240) can display them when the moving body is selected in the current input image, or store them as the movement trajectory of the moving body along with moving body identification information (which may be feature information of the moving body).

[0097] Thus, the moving body position extraction system (200) according to the embodiment of the present invention can automatically calculate and display or output the actual position coordinates of the moving body even if the camera image is changed.

[0098] According to the above-described embodiment, the mobile body position extraction system (200) and the mobile body position extraction method of the present invention are,

[0099] Despite changes in the video coordinate system, such as video rotation, zooming, and resolution changes of security cameras like CCTV, it has the advantage of continuously and accurately extracting the actual location coordinates of moving objects without additional manual work by automatically correcting the reference information generated during initial setup (TIN-related information, video coordinates of reference points, and the actual location coordinates on the map matched thereto).

[0100] In addition, the present invention does not require separate measurement of camera attitude information (internal / external parameters) and is compatible with existing TIN-based position extraction technology, thus having the advantage of being applicable to an extended range without significantly modifying existing moving object position extraction systems.

[0101] Furthermore, through error evaluation and recalibration control functions, the quality of automatic calibration can be quantitatively managed, and accuracy can be maintained with minimal manual intervention only when errors are large, which has the effect of significantly improving maintenance efficiency in actual operating environments.

[0102] In addition, by reducing the time required to reset TINs and reference points in situations of environmental change, such as CCTV replacement, reinstallation, or changes in pan / tilt / zoom patterns, it offers the advantage of providing stable location-based services in various application services, including smart cities, traffic monitoring, public safety, and population flow analysis.

[0103] The present invention has been described above with reference to embodiments illustrated in the drawings, but this is merely illustrative, and those skilled in the art will understand that various modifications and equivalent alternative embodiments are possible therefrom. For example, in the embodiments of the present invention, an irregular triangular mesh (TIN) has been exemplified as an example of a method of dividing a reference image into a plurality of segments (which may be referred to as segments, and each segment is formed by a collection of a plurality of segment points (corresponding to reference points)), but is not limited thereto. Accordingly, the true technical scope of protection of the present invention should be determined by the technical concept of the appended claims.

Claims

Claim 1 A reference information setting unit that stores a camera input image as a reference image, and together stores as reference information TIN-related information generated by creating an irregular triangle network (TIN) based on the image coordinates of reference points selected in the reference image, and information matching the image coordinates of the reference points to actual location coordinates on a map; a TIN location estimation unit that estimates location information of vertices constituting the TIN to be projected onto the current input image by comparing the camera's current input image with the reference image; a reference information correction unit that corrects the reference information to reference information corresponding to the current input image by projecting each vertex point of the estimated TIN onto the current input image; and a moving object location coordinate extraction unit that identifies a triangle containing a representative pixel of the moving object in the current input image onto which the estimated TIN is projected, and calculates and outputs the actual location coordinates of the moving object using the image coordinates of the representative pixel, the image coordinates of each vertex constituting the identified triangle, and actual location coordinate information matched to the image coordinates of each vertex; wherein the reference information correction unit includes the reference points A moving object position extraction system characterized by activating the moving object position coordinate extraction unit by determining the estimated TIN to be valid only when the error between the actual position coordinates before and after reference information correction is calculated for some parts and is below a set threshold. Claim 2 delete Claim 3 A moving object position extraction system according to claim 1, wherein the TIN position estimation unit estimates the position information of each vertex constituting the TIN by, for each vertex constituting the TIN, extracting an image of a rectangular area centered on one vertex of the TIN from the reference image, extracting common feature points between the image of the rectangular area and the current input image through a feature point extraction algorithm, extracting the pairs most similar to each other among the extracted feature points as matched points through a feature point matching algorithm, calculating a position coordinate transformation matrix based on the information of the matched points, and calculating the corresponding coordinates in the current image by performing an operation on the coordinates of one vertex of the TIN with the position coordinate transformation matrix. Claim 4 A moving object position extraction system according to claim 3, wherein the reference information correction unit calculates the error between the actual position coordinates before and after reference information correction for some of the reference points, and if the error exceeds a set threshold, re-corrects the reference information through re-estimation of the transformation matrix, additional feature point matching, or re-designation of some reference points. Claim 5 A moving object location extraction system according to claim 1, wherein the reference information setting unit further comprises a TIN generation unit that generates the TIN by triangulating the road section within the reference image when the number of reference points selected by the manager begins to be three or more. Claim 6 A method for extracting the position of a moving object in a computer system processing camera input images comprises: a step of storing a camera input image at an arbitrary point in time as a reference image, wherein TIN-related information generated by creating an irregular triangle network (TIN) based on the image coordinates of reference points selected by an administrator in the reference image, and information matching the image coordinates of the reference points to actual position coordinates on a map are stored together as reference information; a step of estimating position information of vertices constituting the TIN to be projected onto the current input image by comparing the current input image of the camera with the reference image; a step of correcting the reference information to reference information corresponding to the current input image by projecting each vertex point of the estimated TIN onto the current input image; and a step of identifying a triangle containing a representative pixel of the moving object in the current input image onto which the estimated TIN is projected, and calculating and outputting the actual position coordinates of the moving object using the image coordinates of the representative pixel, the image coordinates of each vertex constituting the identified triangle, and actual position coordinate information matched to the image coordinates of each vertex. A method for extracting the position of a moving body, characterized by performing a step of calculating and outputting the actual position coordinates of the moving body by determining the estimated TIN as valid only when the error between the actual position coordinates before and after correction of reference information is calculated for some of the reference points and the error is below a set threshold. Claim 7 A method for extracting the position of a moving object according to claim 6, wherein the step of estimating the position information of the vertices constituting the TIN comprises: for each vertex constituting the TIN, a step of extracting an image of a rectangular area centered on one vertex of the TIN from the reference image; a step of extracting common feature points between the image of the rectangular area and the current input image through a feature point extraction algorithm; a step of extracting the pairs most similar to each other among the extracted feature points as matched points through a feature point matching algorithm; a step of calculating a position coordinate transformation matrix based on the information of the matched points; and a step of calculating corresponding coordinates in the current image by performing operations on the coordinates of one vertex of the TIN with the position coordinate transformation matrix. Claim 8 delete

Images