Information processing device, control method, and program

Abstract

To easily obtain a dataset to evaluate a shape estimation accuracy.SOLUTION: An information processor includes: first acquisition means for acquiring region information showing a spatial region of a three-dimensional space; second acquisition means for acquiring attribute information on an object; and generation means for generating plural pieces of object data showing the shape and the arrangement of the object on the basis of the region information and the attribute information. For the object data generated by the generation means, a union of ratios of a shape model corresponding to the object is at least a specific ratio in a region corresponding to the spatial region.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a technique for generating three-dimensional shape data.

Background Art

[0002] Conventionally, a shape estimation technique for estimating the three-dimensional shape of a subject based on a two-dimensional image obtained by imaging the subject is known. Further, as a technique for evaluating the accuracy of shape estimation, in Patent Document 1, an evaluation technique using a three-dimensional shape model of a subject is disclosed.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] Generally, the accuracy of shape estimation depends on the positional relationship between the imaging device and the subject that constitute the imaging system. For example, when estimating the shape of a subject such as a person using an imaging system composed of a plurality of digital cameras with different positions and orientations, the number of digital cameras in which the subject is imaged within the angle of view changes depending on the position of the subject. And, the more the number of digital cameras that image the subject, the higher the accuracy tends to be, and the fewer the number, the lower the accuracy tends to be. When the position of the subject changes within such an imaging system, it is convenient to be able to evaluate the accuracy of the shape estimation of the subject according to the position of the subject. As an evaluation method, there is a method of evaluating using a data set in which a three-dimensional shape model of a subject is arranged at various positions. However, there is a problem that a great deal of labor is required to manually create a large amount of data sets.

[0005] Therefore, an object of the present invention is to easily obtain a plurality of data sets for evaluating the accuracy of shape estimation.

Means for Solving the Problem

[0006] One of the information processing apparatuses of the present invention includes a first acquisition means for acquiring region information representing a spatial region in a three-dimensional space, a second acquisition means for acquiring attribute information regarding a subject, and a generation means for generating a plurality of subject data representing the shape and arrangement of the subject based on the region information and the attribute information. The sum of the ratios occupied by the shape models corresponding to the subject in the regions corresponding to the spatial regions of the plurality of subject data generated by the generation means is equal to or greater than a predetermined ratio.

Effect of the Invention

[0007] According to the present invention, a data set for evaluating the shape estimation accuracy can be easily obtained.

Brief Description of the Drawings

[0008] [Figure 1] Block diagram showing the hardware configuration of the information processing apparatus. [Figure 2] Block diagram showing the logical configuration of the information processing apparatus in the first embodiment. [Figure 3] Schematic diagram for explaining voxels and the capture space region. [Figure 4] Schematic diagram showing an example of a subject. [Figure 5] Flowchart showing the overall flow of the processing executed by the information processing apparatus in the first embodiment. [Figure 6] Diagram showing an example of the GUI in the first embodiment. [Figure 7] Flowchart showing the flow of the subject data generation process in the first embodiment. [Figure 8] Schematic diagram for explaining the arrangement of the subject. [Figure 9] Schematic diagram showing an example of the arrangement combination. [Figure 10] Diagram showing an example of another GUI in the first embodiment. [Figure 11]A block diagram showing the logical configuration of the information processing device in the second embodiment. [Figure 12] A flowchart showing the overall flow of processing performed by the information processing device in the second embodiment. [Figure 13] A figure showing an example of a GUI in the second embodiment. [Figure 14] A figure showing an example of a GUI in the third embodiment. [Figure 15] A flowchart showing the flow of the subject data generation process in the third embodiment. [Figure 16] A schematic diagram showing an example of subject data in the third embodiment. [Figure 17] A flowchart showing the overall flow of processing performed by the information processing device in the fourth embodiment. [Figure 18] A schematic diagram illustrating an example of sorting subject data. [Modes for carrying out the invention]

[0009] Embodiments of the present invention will be described below with reference to the drawings. Note that the following embodiments are not limiting to the present invention, and not all combinations of features described in these embodiments are essential to the solution of the present invention. Identical components will be denoted by the same reference numerals.

[0010] (First Embodiment) <Hardware Configuration> Figure 1 shows an example of the hardware configuration of the information processing device in this embodiment. In this figure, the information processing device 100 includes a CPU 101, RAM 102, ROM 103, SATA (Serial ATA) I / F (interface) 104, VC (video card) 105, and general-purpose I / F 106.

[0011] The CPU 101 is a processor that uses RAM 102 as work memory to execute the OS (operating system) and various programs stored in ROM 103, external storage device 111, etc. The OS and various programs may also be stored in internal storage devices. The CPU 101 also controls each component via the system bus 107. The processing described later using a flowchart involves the program code stored in ROM 103 or external storage device 111 being loaded into RAM 102 and executed by the CPU 101.

[0012] An external storage device 111 is connected to SATAI / F104 via serial bus 108. The external storage device 111 is an HDD (hard disk drive) or SSD (solid state drive). A display 112 is connected to VC105 via serial bus 109. An input device 113, such as a mouse or keyboard, is connected to general-purpose I / F106 via serial bus 110.

[0013] The CPU 101 displays a GUI (Graphical User Interface) provided by the program on the display 112 and receives input information representing user instructions obtained via the input device 113. The information processing device 100 is implemented, for example, by a desktop PC (personal computer). Alternatively, it may be implemented by a notebook PC or tablet PC integrated with the display 112. The external storage device 111 can also be implemented by a media (recording medium) and an external storage drive for accessing the media. The media can include FD (flexible disk), CD-ROM, DVD, USB memory, MO, flash memory, etc.

[0014] <Logical Structure> Figure 2 is a block diagram showing the logical configuration of the information processing device 100 in this embodiment. The information processing device 100 functions with the logical configuration shown in Figure 2 by having the CPU 101 execute a program stored in the ROM 103, with the RAM 102 as the work memory. Note that not all of the following processes must necessarily be executed by the CPU 101; the information processing device 100 may be configured so that some or all of the processes are performed by one or more processing circuits other than the CPU 101.

[0015] The information processing device 100 includes a spatial area acquisition unit 201, a subject attribute acquisition unit 202, and a generation unit 203.

[0016] The spatial domain acquisition unit 201 acquires the size of a voxel, which is a unit volume element constituting a three-dimensional space, and the extent of the capture spatial domain as domain information, based on user instructions input via the input device 113. The capture spatial domain is the area in which the object whose shape is to be estimated (i.e., the subject) may exist. An example of a voxel and a capture spatial domain is shown in Figure 3. In this example, the voxel 301 has a side length of d unit It is a cube. Furthermore, the captured space region 302 is the region in xyz space where x0 ≤ x ≤ x1, y0 ≤ y ≤ y1, and z0 ≤ z ≤ z1. The acquired region information is sent to the generation unit 203.

[0017] The subject attribute acquisition unit 202 acquires attribute information about the subject based on user instructions input via the input device 113. Attribute information refers to information that indicates the number and shape characteristics of the subject. In this embodiment, a person is assumed as the subject, and the number of people, height, and width are acquired as attribute information. In addition to height and width, information that indirectly indicates the shape characteristics of the subject, such as gender and age, may also be included in the attribute information. Furthermore, the assumed subject may be an object other than a person, in which case the number and size of the object may be used as attribute information. The acquired attribute information is sent to the generation unit 203.

[0018] The generation unit 203 generates multiple patterns of subject data representing the shape and arrangement of subjects in three-dimensional space based on region information and attribute information. In this embodiment, the subject data consists of a shape model in which the shape of each subject person is represented by a polygon mesh, and the position coordinates and orientation (rotation angle) of each subject person. The shape model consists of a list of three-dimensional coordinates of the vertices that make up the polygon mesh. An example of the shape model and subject data is shown in Figure 4.

[0019] Figure 4(a) shows the shape of an object represented by a shape model. In this figure, shape model 401 is a polygon mesh representing the surface shape of the object. Point 402 and direction 403 are the reference point and front direction of this shape model, respectively. In this example, the foot of the perpendicular from the centroid 404 of shape model 401 to the xz plane is the reference point 402, and the x-axis direction is the front direction 403.

[0020] Figure 4(b) shows an object placed in a three-dimensional space indicated by the object data. In this example, two shape models 405 and 408 are placed, with point 406 and direction 407 representing the reference point and orientation of shape model 405, respectively. Point 409 and direction 410 represent the reference point and orientation of shape model 408, respectively. In this figure, coordinates (px1, py1, pz1) and (px2, py2, pz2) are the position coordinates of shape model 405 and shape model 408, respectively. θ1 and φ1 are rotation angles representing the orientation of shape model 405. θ2 and φ2 are rotation angles representing the orientation of shape model 408.

[0021] The shape model may also include uv coordinates (so-called texture coordinates) corresponding to the xyz coordinates of the vertices that make up the polygon mesh, as well as color information for each vertex. Details of subject data generation will be described later. The generated subject data for multiple patterns is output as a subject dataset to an external storage device 111 or the like and stored.

[0022] <Process to be executed> Figure 5 is a flowchart showing the overall flow of processing performed by the information processing device 100.

[0023] In step S501, the spatial area acquisition unit 201 acquires the size of the voxels and the range of the captured spatial area based on the user's instructions. In step S502, the subject attribute acquisition unit 202 acquires the number of subjects, their height, and their body width based on the user's instructions. The subject attribute acquisition unit 202 receives the user's instructions via a GUI 601 displayed on the display 112, for example, as shown in Figure 6.

[0024] In Figure 6, setting field 602 is a field for inputting the length of one side of a voxel as the size of the voxel. Setting fields 603, 604, 605, 606, 607, and 608 are fields for inputting the start point x0 and end point x1 in the x-axis direction, the start point y0 and end point y1 in the y-axis direction, and the start point z0 and end point z1 in the z-axis direction, respectively, of the capture space area. Subject attribute setting lists 609, 610, and 611 are pull-down lists for selecting the number of subjects, height, and body width, respectively. Button 612 is a button for the user to instruct the execution of the subject data generation process, which will be described later. Display area 613 is an area for displaying images related to the generated subject data. Slider 614 is a slider for specifying the identification number of the subject data to be displayed in display area 613. Button 615 is a button for the user to instruct the user to save the generated subject data. When button 612 is pressed by the user, steps S501 and S502 are executed.

[0025] The information acquired in steps S501 and S502 corresponds to area information and subject attribute information.

[0026] In step S503, the generation unit 203 generates multiple subject data based on the region information acquired in steps S501 and S502 and the subject attribute information from step S5. Here, the generation unit 203 continues to generate subject data by changing the arrangement of the subjects until the union of the proportions that the subjects occupy in the capture space region (hereinafter referred to as the spatial coverage rate) exceeds a predetermined proportion. Details of the processing in this step will be described later with reference to Figure 7, including an example of the spatial coverage rate. In this embodiment, the proportion that the shape surface of the subject occupies in the capture space region is determined in voxel units and used as the spatial coverage rate. In this embodiment, a predetermined proportion is set as the target spatial coverage rate Th C The following is written and explained: Target spatial coverage Th C This can be predetermined or set separately based on user instructions.

[0027] In step S504, the generation unit 203 saves the multiple subject data generated in step S503 as a dataset (subject dataset) to an external storage device 111 or the like, and terminates the process. Alternatively, as shown in the display area 613 of Figure 6, a rendering image based on the generated subject data may be displayed on the GUI. Known computer graphics techniques can be used to generate the rendering image. In the example in Figure 6, a rendering image based on the subject data with the number specified by the slider 614 is displayed. In Figure 6, the subject data indicated by the subject data specified by the slider 614 is two people placed at an interval. The user can switch between the subject data images displayed in the display area 613 by operating the slider 614. The user can then visually confirm what kind of subject data has been generated by referring to the subject data images displayed in the display area 613.

[0028] <Details of the subject data generation process in step S503> The details of the subject data generation process performed by the generation unit 203 will be described using the flowchart shown in FIG. 7. Hereinafter, the number of subjects obtained in step S502 will be described as M (M is an integer of 1 or more).

[0029] In step S701, the generation unit 203 acquires a shape model of the subject based on the height and width of the subject acquired in step S502. For example, the generation unit 203 acquires a shape model corresponding to the height and width acquired in step S502 from a database in which various shape models created for the subject are associated with attributes. Alternatively, the generation unit 203 may acquire a shape model of the subject by deforming the shape model acquired from the external storage device 111 based on the height and width acquired in step S502. In the present embodiment, the generation unit 203 acquires one common shape model for M subjects (persons).

[0030] In step S702, the generation unit 203 sets an index m indicating one of the M subjects (persons) who are the subjects to 1.

[0031] In step S703, the generation unit 203 initializes the value of the counter f COUNT (ix, iy, iz) corresponding to each voxel v(ix, iy, iz) constituting the capture space region to 0. Here, ix, iy, and iz are indices indicating the positions of the voxels.

[0032] In step S704, the generation unit 203 sets an index k indicating which arrangement of the subject with index m to 1.​​​​​​​​​​​​​​​The orientation of the shape model (i.e., the rotation angle around the reference point) may be determined randomly or according to any predetermined rule. For example, considering the typical movement of a person, the coordinate y in the y-axis direction (vertical direction) may be determined P(m,k) We may keep the y-axis constant and limit the change in orientation to rotation around the y-axis only. Hereafter, the position coordinates and orientation of the shape model will be collectively referred to as "arrangement," and the k-th arrangement of the shape model of the object at index m will be represented by P(m,k).

[0034] In step S706, the generation unit 203 determines which voxels constituting the capture space region intersect with the surfaces of the shape model placed in step S705. The generation unit 203 adds +1 to the value of the counter corresponding to the voxels whose surfaces intersect with the surfaces of the shape model placed in step S705.

[0035] In step S707, the generation unit 203 calculates the spatial coverage ratio F based on the value of the counter. COVER The spatial coverage ratio F is calculated. COVER It can be expressed by the following formula. F COVER =VNUM + / VNUM ALL ...Equation (1)

[0036] Here, VNUM + VNUM is the number of voxels whose counter value is 1 or greater. ALL f is the total number of voxels that make up the capture space region. COUNT The value of (ix,iy,iz) is 1 or greater if voxel v(ix,iy,iz) is the surface of the shape model in any of the arrangements P(m,1), P(m,2), ..., P(m,k) of the shape model of the index m subject. Therefore, the spatial coverage ratio F is expressed by equation (1). COVER This represents the union of the proportions of the capture space region occupied by voxels that become the subject surface in one or more of the k possible configurations.

[0037] In step S708, the generation unit 203 generates the spatial coverage ratio F obtained in step S707. COVER Target spatial coverage Th COVER Determine whether the above is true or not. Spatial coverage F COVER Target spatial coverage Th COVER In this case, the arrangement of the shape model of the index m subject with the k arrangements determined by the processing so far is considered to sufficiently cover the capture space region. Therefore, the spatial coverage rate F COVER Target spatial coverage Th COVER If the above conditions are met, the process in step S710 is executed. Spatial coverage F COVER Target spatial coverage Th COVER If it is less than, in step S709, the generation unit 203 adds +1 to index k and returns to the process in step S705.

[0038] Here, using Figure 8, we will explain an example of the arrangement, counter value, and spatial coverage rate determined in steps S705 to S707. In this figure, for the sake of simplicity, we will assume that the capture spatial region 801 is a spatial region composed of voxels arranged on a two-dimensional plane. Figure 8(a) is an example of the arrangement P(m,k) (m=1,k=1,2,···,5) determined in step S705. In this figure, the rectangle 802 is a voxel, the shape 803 is the shape shown by the shape model, the point 804 is the reference point of the shape model, and the direction 805 is the front direction of the shape model. The shaded area indicates voxels that intersect with the surface of the arranged shape model. Figure 8(b) shows the counter value updated in step S706 and the spatial coverage rate F calculated in step S707 when the arrangement is determined as in Figure 8(a). COVER This shows that, in this figure, the numbers in rectangle 806 are counters corresponding to the voxels of rectangle 802, and the shaded areas correspond to the voxels in the shaded areas of Figure 8(a) (i.e., voxels that intersect with the surface of the placed shape model). Target spatial coverage Th COVER If, for example, k=5, then the spatial coverage ratio F is first found when k=5. COVER Target spatial coverage Th COVER That concludes the explanation. In this case, the spatial coverage ratio F COVERWhen the value reaches 100%, the process in step S710 is executed.

[0039] In step S710, the generation unit 203 adds +1 to the index m of the subject person. In step S711, the generation unit 203 determines whether the index m is greater than or equal to M. If the index m is greater than M, the process in step S712 is executed. If the index m is less than or equal to M, the process returns to step S703.

[0040] In step S712, the generation unit 203 generates subject data including M subjects (people) by combining the arrangements determined in steps S702 to S711, and terminates the subject data generation process. Specifically, the generation unit 203 generates and stores subject data for each arrangement at index m (=1, 2, ..., M). In this embodiment, the generation unit 203 performs the above subject data generation for all arrangement combinations. An example for the case where M=2 is shown in Figure 9.

[0041] Figure 9(a) shows an example of the arrangement P(1,k1) (k1=1,2,···,5) of subject 901 with index m = 1. Figure 9(b) shows an example of the arrangement P(2,k2) (k2=1,2,3,4) of subject 902 with index m = 2. Figure 9(c) shows an example of all combinations of arrangement P(1,k1) and arrangement P(2,k2). Through the processing of steps S702 to S711 described above, the spatial coverage rate F corresponding to arrangements P(1,k1) and P(2,k2) is obtained. COVER These represent the target spatial coverage rate Th. COVERThe above has been determined. Therefore, by combining placements P(1,k1) and P(2,k2), a comprehensive subject dataset containing various positional relationship patterns of two people can be generated. Note that, considering that it is unrealistic for multiple people to overlap, subject data where the shape models of the subjects overlap does not need to be generated. For example, P(1,5)+P(2,1), P(1,2)+P(2,2), P(1,1)+P(2,3), and P(1,3)+P(2,4) in Figure 9(c) do not need to be generated as subject data.

[0042] By performing the processing control described above, a dataset of subject data for evaluating the accuracy of shape estimation in the imaging system can be easily obtained.

[0043] In this embodiment, an example was described in which the range entered by the user via the GUI shown in Figure 6 is obtained as the range of the capture space area, but the method of obtaining the range of the capture space area is not limited to this. For example, the range corresponding to the name entered by the user via the GUI may be obtained by referring to a LUT that associates the spatial range of the capture space with the name corresponding to the capture space. Figure 10(a) shows an example of the GUI in this case. In this case, the LUT is created in advance and stored in an external storage device 111 or the like.

[0044] Furthermore, if there are multiple subjects, attribute information such as height and body width may be specified individually, as shown in Figure 10(b). In that case, the generation unit 203 acquires a shape model for each subject in step S701.

[0045] Furthermore, while we have explained an example in which the spatial coverage rate is calculated according to equation (1) based on the number of voxels, the spatial coverage rate only needs to represent the proportion of the subject to the captured spatial region, and may be calculated by other methods. For example, the spatial coverage rate may be the ratio of the union of the volumes of the subject's shape model to the volume of the captured spatial region. In that case, in step S706, in addition to the voxels where the surfaces of the shape model intersect, the voxels contained inside the shape model should also be found, and the corresponding counter values ​​should be incremented by 1.

[0046] (Second Embodiment) In the first embodiment, a method for generating a dataset of subject data (subject dataset) was described. In this embodiment, an example of evaluating the estimation accuracy of shape estimation with respect to an imaging system using the generated dataset will be described.

[0047] The hardware configuration of the information processing device 100 in this embodiment is the same as that of the first embodiment, so its description will be omitted. The following will mainly describe the differences between this embodiment and the first embodiment. Components identical to those in the first embodiment will be denoted by the same reference numerals.

[0048] Figure 11 is a block diagram showing the logical configuration of the information processing device 100 in this embodiment. The information processing device 100 includes a spatial area acquisition unit 201, a subject attribute acquisition unit 202, a generation unit 203, a camera parameter setting unit 1101, a rendering unit 1102, a shape estimation unit 1103, and an evaluation unit 1104. The spatial area acquisition unit 201, the subject attribute acquisition unit 202, and the generation unit 203 are the same as in the first embodiment, so their description is omitted. However, in this embodiment, the subject dataset consisting of multiple patterns of subject data generated by the generation unit 203 is sent to the rendering unit 1102 and the evaluation unit 1104.

[0049] The camera parameter setting unit 1101 sets the camera parameters of the imaging system to be evaluated based on user instructions input via the input device 113. In this embodiment, the camera parameters consist of internal parameters, external parameters, and distortion parameters for each digital camera included in the imaging system. Internal parameters are parameters that represent the position of the principal point of the digital camera and the focal length of the lens. External parameters are parameters that represent the position and orientation of the digital camera. Distortion parameters are parameters that represent the distortion of the lens of the digital camera. The set camera parameters are sent to the rendering unit 1102.

[0050] The rendering unit 1102 generates a simulated image of the subject being imaged by the imaging system based on the camera parameters and subject dataset. In this embodiment, this generated simulated image is also called the imaging simulation image. Details will be described later. The generated imaging simulation image is sent to the shape estimation unit 1103.

[0051] The shape estimation unit 1103 applies a predetermined shape estimation algorithm to the captured simulation image to estimate the shape of the subject. The predetermined shape estimation algorithm can be any method that estimates the three-dimensional shape of the subject based on a two-dimensional image of the subject, such as the known view volume cross-eyed method or a method using stereo matching. Hereinafter, the data representing the three-dimensional shape obtained as a result of the estimation will be called estimated shape data. The obtained estimated shape data is sent to the evaluation unit 1104.

[0052] The evaluation unit 1104 evaluates the estimation accuracy (shape estimation accuracy) of the shape estimation based on the subject data and estimated shape data, and displays the results on the GUI. Further details will be described later.

[0053] Figure 12 is a flowchart showing the overall flow of processing performed by the information processing device 100 in this embodiment.

[0054] In step S1201, the spatial area acquisition unit 201 acquires area information based on the user's instructions. In step S1202, the subject attribute acquisition unit 202 acquires subject attribute information based on the user's instructions.

[0055] In step S1203, the camera parameter setting unit 1101 retrieves the camera parameter file from an external storage device 111 or the like based on the user's instructions and sets the camera parameters. The user's instructions are received via a GUI 1301 as shown in Figure 13. In Figure 13, the camera parameter setting field 1302 is a field for entering the path to the camera parameter file. Button 1303 is a button that is pressed (selected) when instructing the execution of the evaluation. Display areas 1304 and 1305 are areas for displaying the evaluation results. When button 1303 is pressed by the user, steps S1201 to S1203 are executed, and the process proceeds to step S1204. Hereinafter, the number of subjects acquired in step S1202 will be described as M (where M is an integer of 1 or more), as in the first embodiment.

[0056] In step S1204, the generation unit 203 generates subject data based on the region information acquired in step S1201 and the subject attribute information acquired in step S1202. Here, in this embodiment, the generation unit 203 generates N DATA Individual subject data Data(i)(i=1,2,···,N DATA This generates ). The processing in this step is the same as step S503 in Figure 5 of the first embodiment, so the explanation is omitted.

[0057] In step S1205, the rendering unit 1102 renders the N generated in step S1204. DATAThe subject (i.e., a shape model placed in 3D space) represented by individual subject data is rendered to generate an imaging simulation image. Here, the rendering unit 1102 renders using the camera parameters acquired in step S1203. In this embodiment, the rendering algorithm by the rendering unit 1102 uses known computer graphics techniques. The imaging simulation image in this embodiment is the same as the imaging simulation image in the first embodiment.

[0058] Here, N is the number of digital cameras included in the imaging system. CAM Therefore, the number of image simulation images is N per subject data. CAM The following is a simulated image of the subject shown by subject data Data(i) when it is captured by the c-th digital camera included in the imaging system, as shown in Img. i (c)(c=1,2,···,N CAM ) is represented as. Imaging simulation image Img i (c) is an image showing M subjects (people) arranged according to the subject data Data(i).

[0059] In step S1206, the shape estimation unit 1103 is N DATA For each individual subject data, the imaging simulation image Img generated in step S1205 is shown. i (c) is subjected to a predetermined shape estimation algorithm to obtain estimated shape data EData(i).

[0060] In step S1207, the evaluation unit 1104 evaluates the accuracy of the estimated shape obtained in step S1206 using the subject data generated in step S1204, displays the result on the GUI, and terminates the process. Specifically, the evaluation unit 1104 evaluates i=1,2,···,N DATA In each case, the Hausdorff distance d is between the 3D shape shown by the subject data Data(i) and the 3D shape shown by the estimated shape data EData(i). H(i) is calculated as the evaluation value. Furthermore, the evaluation unit 1104 calculates the Hausdorff distance d H (i) The average and maximum values ​​are calculated and displayed on the GUI. For example, as shown in Figure 13, the evaluation value calculated from the subject data corresponding to the identification number specified by the slider 614 (i.e., the subject data displayed in the display area 613) is displayed in the display area 1304. In addition, the average and maximum values ​​of the evaluation value are displayed in the display area 1305 as an evaluation result for the imaging system. Note that the evaluation value is not limited to the Hausdorff distance, but can be any index or value that represents the difference or similarity between two shapes.

[0061] By performing the processing control described above, the second embodiment allows for a comprehensive evaluation of the accuracy of estimating the shape of the assumed subject with respect to the imaging system.

[0062] Alternatively, constraints on the shooting system, such as the number of digital cameras and the locations where they can be installed, may be obtained separately, and multiple camera parameters for shooting systems that satisfy these constraints may be created to perform the evaluation described above.

[0063] (Third embodiment) In this embodiment, we will describe an example of generating a subject dataset in which the subject moves at a specified speed, considering the subject dataset as a series of time-series data representing the movement of the subject.

[0064] The hardware and logical configurations of the information processing device 100 in this embodiment are equivalent to those of the first embodiment, so a description will be omitted.

[0065] The flowchart showing the overall flow of processing performed by the information processing device 100 in this embodiment is largely the same as the flowchart in Figure 5 described in the first embodiment. The differences in processing in this embodiment are as follows: In step S502, the subject attribute acquisition unit 202 acquires the subject's movement speed as attribute information about the subject. Also, the content of the subject data generation process by the generation unit 203 in step S503 is different. Note that the processing in steps S501 and S504 is the same as in the first embodiment, so the explanation is omitted.

[0066] In step S502, the subject attribute acquisition unit 202 acquires the number of subjects, their height, body width, and movement speed based on user instructions via the GUI.

[0067] In step S503, the generation unit 203 generates subject data based on the region information and attribute information acquired in steps S501 and S502, so that the subject moves at the movement speed acquired in step S502. Here, similar to the first embodiment, the generation unit 203 continues to generate subject data by changing the arrangement of the subject until the union of the proportions that the subject occupies in the capture space region is equal to or greater than a predetermined target space coverage rate.

[0068] The details of the subject data generation process in this embodiment will be explained using the flowchart shown in Figure 15. Note that the processes in S1501 and S1502 are the same as steps S701 and S703 in Figure 7 of the first embodiment, respectively, so their explanation will be omitted. Hereafter, the movement speed of the subjects acquired in step S502 will be α [cm / sec], and the number of subjects will be M (M is an integer of 1 or more), as in the first embodiment.

[0069] In S1503, the generation unit 203 initializes time t to 0.

[0070] In S1504, the generation unit 203 generates subject data Data(t) corresponding to time t.

[0071] At time t=0, the generation unit 203 selects M points from within the capture space region and assigns them to the position coordinates p of the M subjects at time t=0. m (0)(m=1,2,···,M) The shape model obtained in S1501 is placed. At this time, the position coordinates may be chosen randomly, or they may be chosen so that they are equally spaced on the xz plane. Also, the orientation of the subjects may be any orientation, may be determined randomly for each subject person, or they may all be uniformly in the same orientation. Hereafter, the position coordinate of the m-th subject at time t is p m Let's represent it as (t).

[0072] If time t > 0, the generation unit 203 generates position coordinates p for each subject person. m A point is randomly selected from a circle with radius r = α*Δt centered at (t-Δt), and this point is assigned to position coordinates p. m A shape model is placed as (t). Here, Δt is a predetermined elapsed time between temporally adjacent object data, for example, Δt = 1 / 30 [seconds]. Also, the radius r corresponds to the distance traveled (change) when moving over Δt at velocity α. The orientation of the placed object can be arbitrary. For example, a 3D vector p m (t)-p m If we roughly match the direction indicated by (t-Δt), the arrangement will appear as if the objects are moving in the direction of travel.

[0073] The generation unit 203 stores the shape models for M people arranged as described above as subject data Data(t) at time t. An example of subject data generated when M=2 is shown in Figure 16.

[0074] In S1505, the generation unit 203 determines the voxels where the surfaces intersect for the shape model arranged in S1504 in the same manner as in step S706 in Figure 7, and increments the counter value by 1.

[0075] In S1506, the generation unit 203 generates a spatial coverage ratio F according to formula (1) of the first embodiment. COVER Calculate.

[0076] In S1507, the generation unit 203 generates the spatial coverage ratio F obtained in S1506. COVER Target spatial coverage Th COVER If the above conditions are met, the subject data generation process will be terminated. Spatial coverage F COVER Target spatial coverage Th COVER If it is less than , the generation unit 203 proceeds to S1508, adds Δt to time t, and returns to S1504.

[0077] Figure 14 shows an example of the GUI 1401 in this embodiment. In this figure, the speed setting field 1402 is a field for inputting the movement speed of the subject. Button 1403 is pressed when the generated subject dataset is played back as a series of time-series data. When button 1403 is pressed by the user, the subject data displayed in the display area 613 switches at the Δt [second] intervals described above.

[0078] By performing the processing control described above, according to the third embodiment, a subject dataset for evaluating the shape estimation accuracy for animal bodies can be obtained.

[0079] (Fourth Embodiment) In this embodiment, we will describe an example of rearranging subject data within a subject dataset according to the distribution of the positions of the placed subjects, in order to make it easier for users to use when performing evaluations.

[0080] The hardware and logical configurations of the information processing device 100 in this embodiment are equivalent to those of the first embodiment, so a description will be omitted.

[0081] Figure 17 is a flowchart showing the overall flow of processing performed by the information processing device 100 in this embodiment. The processing in S1701 to S1703 is the same as steps S501 to S503 in Figure 5 of the first embodiment, so its explanation is omitted.

[0082] In S1704, the generation unit 203 generates NDATA Individual subject data Data(i)(i=1,2,···,N DATA Priority is assigned to the following. In this embodiment, the generation unit 203 processes n=1 to n=N DATA The variance σ of the subject position is calculated from the unassigned subject data using the following formula: 2 Find the subject data Data(i') that maximizes (i'), and assign it a priority rank of n.

[0083]

number

[0084] Here, I order is a set of n-1 indices representing subject data that have already been assigned a priority, and o(i,m) is the position coordinates of the m-th subject person included in the subject data Data(i). In this case, the subject data that maximizes the variance of the position coordinates of the subject person when combined with the n-1 subject data with higher priority is selected as the subject data with priority rank n.

[0085] In S1705, the generation unit 203 saves the subject data, which has been sorted according to the priority assigned in S1704, as a subject dataset in the external storage device 111 or the like, and terminates processing. Figure 18(a) shows an example of the dataset before sorting, and Figure 18(b) shows an example of the dataset after sorting. In Figure 18(b), the subject data is arranged from left to right in order of priority. By applying the processing of steps S1205 to S1207 described in the second embodiment to the subject data with the highest priority, the evaluation of subjects with similar arrangements can be postponed. Furthermore, even without completing the evaluation of all subject data, it is possible to grasp the approximate shape estimation accuracy of the imaging system.

[0086] By performing the processing control described above, according to the fourth embodiment, a subject dataset can be obtained in which the subject data is arranged in an order corresponding to the distribution of the subject's position.

[0087] Alternatively, the subject data may be sorted in descending order of the dispersion of the subject's position. In that case, as shown in Figure 18(c), subject data with a high density of subjects (i.e., high difficulty in shape estimation) will be given a higher priority.

Claims

1. A first acquisition means for acquiring region information representing a spatial region in three-dimensional space, A second means of acquiring attribute information about the subject, The system includes a generation means for generating multiple subject data representing the shape and arrangement of a subject based on the aforementioned region information and attribute information, The plurality of subject data generated by the generation means such that the union of the proportions occupied by the shape models corresponding to the subjects in the region corresponding to the spatial region is greater than or equal to a predetermined proportion. An information processing device characterized by the following:

2. The information processing apparatus according to claim 1, characterized in that the region information consists of information indicating the range of the three-dimensional space and information indicating the size of a voxel, which is a unit volume element constituting the region corresponding to the three-dimensional space.

3. The information processing apparatus according to claim 2, characterized in that the generation means calculates the proportion of the spatial region based on the number of voxels contained within the shape of the subject.

4. The information processing apparatus according to claim 2, characterized in that the generation means calculates the proportion of the spatial region based on the number of voxels that intersect with the surface of the shape of the subject.

5. The information processing device according to claim 1, characterized in that the subject is a person, and the attribute information includes at least information regarding the number of people, height, and body width.

6. The information processing apparatus according to claim 1, characterized in that the generation means acquires shape data representing the shape of the subject based on the attribute information, and determines the arrangement of multiple patterns of the subject based on the area information and the shape data.

7. The information processing apparatus according to claim 1, characterized in that the generation means assigns priority to the subject data based on the distribution of the positions of the placed subjects.

8. The second acquisition means acquires the movement speed of the subject, The generation means associates the subject data with time and generates the subject data such that the amount of change in the position of the subject between temporally adjacent subject data is substantially equal to the amount of change corresponding to the elapsed time in the movement speed and the subject data. The information processing apparatus according to feature 1.

9. A camera parameter setting means for setting camera parameters used for imaging a subject placed in the aforementioned spatial region, A rendering means for generating an imaging simulation image using the subject data and the camera parameters, A shape estimation means for estimating the three-dimensional shape of the subject using the aforementioned imaging simulation image, An evaluation means for evaluating the shape estimation accuracy based on the shape shown in the subject data and the estimated shape estimated by the shape estimation means. The information processing apparatus according to claim 1, further comprising the above.

10. The information processing apparatus according to claim 9, characterized in that the evaluation means calculates the Hausdorff distance between the shape indicated by the subject data and the estimated shape as an evaluation value.

11. The first acquisition step involves obtaining region information representing the spatial region of a three-dimensional space, The second acquisition step involves acquiring attribute information about the subject, The process includes a generation step of generating multiple subject data representing the shape and arrangement of the subject based on the area information and attribute information, The plurality of subject data generated by the generation process are such that the union of the proportions occupied by the shape models corresponding to the subjects in the region corresponding to the spatial region is greater than or equal to a predetermined proportion. A control method for an information processing device characterized by the following features.

12. The first acquisition step involves obtaining region information representing the spatial region of a three-dimensional space, The second acquisition step involves acquiring attribute information about the subject, A program for causing an information processing device to perform a generation step of generating multiple subject data representing the shape and arrangement of a subject based on the aforementioned area information and attribute information, The plurality of subject data generated by the generation process are such that the union of the proportions occupied by the shape models corresponding to the subjects in the region corresponding to the spatial region is greater than or equal to a predetermined proportion. A program characterized by the following features.

Images