Information processing device, computer-readable recording medium, and information processing method

Abstract

The information processing device comprises: a three-dimensional information acquisition unit (110) which acquires three-dimensional information including a plurality of images; a foreground image generation unit (120) which selects an identification object from the three-dimensional information, cuts out images of a range corresponding to the selected identification object, i.e., the identification object range, from each of the plurality of images, thereby generating a plurality of foreground images based on the plurality of images; a background image DB (130) which stores a plurality of background images; and an image synthesis unit (140) which synthesizes each of the plurality of foreground images into each of the plurality of background images, thereby generating a plurality of synthesized images.

Description

Technical Field

[0001] The present invention relates to an information processing device, a computer-readable recording medium, and an information processing method. Background Art

[0002] In the field of image processing, a technique for learning a large number of images to recognize objects has been proposed. If open source is used, a large number of images showing people or cars can be prepared. However, it is difficult to collect a large number of images of a specific object such as a specific car, equipment or product.

[0003] For example, non-patent document 1 proposes the following method: in order to prepare a large number of images of a specific object, after creating a foreground image by rendering a three-dimensional model of the object from multiple positions and angles, it is superimposed on multiple background images prepared in advance, thereby creating a large number of images.

[0004] Prior art literature

[0005] Non-patent literature

[0006] Non-patent document 1: Manik Goyal, Param Rajpura, Hristo Bojinov, and Ravi Hegde, "Dataset Augmentation with Synthetic Images Improves Semantic Segmentation", arXiv: 1709.00849v3, June 26, 2018 Summary of the invention

[0007] Problems to be solved by the invention

[0008] The conventional technology is effective when the shape and texture of the three-dimensional model are highly refined.

[0009] There are two methods described below as a three-dimensional model creation method.

[0010] The first method is a method of manually creating a three-dimensional model. This method has the problem of being very time-consuming.

[0011] The second method is a method of preparing a special environment such as a turntable to create a three-dimensional model. This method has a problem that a three-dimensional model cannot be created unless the object has a weight and size that the special environment such as the turntable allows.

[0012] As described above, it is unrealistic to create a high-definition three-dimensional model of a specific object, and therefore the existing technology lacks practicality.

[0013] Therefore, an object of one or more aspects of the present invention is to enable easy generation of a large number of images of specific objects.

[0014] Means for solving problems

[0015] An information processing device according to one embodiment of the present invention is characterized in that the information processing device comprises: a three-dimensional information acquisition unit, which acquires three-dimensional information including a plurality of images; an identification object selection unit, which selects an identification object from the three-dimensional information; an identification object cutting unit, which cuts out images of a range corresponding to the selected identification object, i.e., an identification object range, from the plurality of images, thereby generating a plurality of foreground images based on the plurality of images; a background image storage unit, which stores a plurality of background images; and an image synthesis unit, which synthesizes each of the plurality of foreground images into each of the plurality of background images, thereby generating a plurality of synthesized images.

[0016] A program of one embodiment of the present invention is characterized in that the program causes a computer to function as: an identification object selection unit, which selects an identification object from three-dimensional information containing multiple images; an identification object cutting unit, which cuts out images of a range corresponding to the selected identification object, i.e., the identification object range, from the multiple images, thereby generating multiple foreground images based on the multiple images; a background image storage unit, which stores multiple background images; and an image synthesis unit, which synthesizes each of the multiple foreground images in each of the multiple background images, thereby generating multiple synthesized images.

[0017] An information processing method according to one embodiment of the present invention is characterized in that three-dimensional information including a plurality of images is obtained, an identification object is selected from the three-dimensional information, images of a range corresponding to the selected identification object, i.e., the range of the identification object, are cut out from the plurality of images respectively, thereby generating a plurality of foreground images based on the plurality of images, and each of the plurality of foreground images is synthesized in each of the plurality of background images, thereby generating a plurality of synthesized images.

[0018] Effects of the Invention

[0019] According to one or more aspects of the present invention, it is possible to easily generate a large number of images of specific objects. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a block diagram schematically showing the configuration of an information processing device according to an embodiment.

[0021] Figure 2 This is a schematic diagram for explaining a three-dimensional coordinate system.

[0022] Figure 3This is a block diagram schematically showing the structure of a foreground image generating unit.

[0023] Figure 4 This is a first schematic diagram for explaining the selection of the recognition target.

[0024] Figure 5 This is a second schematic diagram for explaining the selection of the recognition target.

[0025] Figure 6 This is a third schematic diagram for explaining the selection of the recognition target.

[0026] Figure 7 This is a fourth schematic diagram for explaining the selection of the recognition target.

[0027] Figure 8 This is a block diagram showing a hardware configuration example of an information processing device.

[0028] Fig. 9 This is a flowchart showing the operation of the information processing device.

[0029] Fig.10 : is a flowchart showing the operation of generating a foreground image. DETAILED DESCRIPTION

[0030] Figure 1 It is a block diagram schematically showing the configuration of the information processing device 100 according to the embodiment.

[0031] The information processing device 100 has a three-dimensional information acquisition unit 110, a foreground image generation unit 120, a background image database (hereinafter referred to as background image DB) 130 as a background image storage unit, an image synthesis unit 140 and a learning image database (hereinafter referred to as learning image DB) 150 as a learning image storage unit.

[0032] The three-dimensional information acquisition unit 110 acquires three-dimensional information including a plurality of images and position information indicating the three-dimensional positions of a plurality of points respectively included in the plurality of images. For example, the three-dimensional information acquisition unit 110 acquires the three-dimensional information using SLAM (Simultaneous Localization And Mapping).

[0033] In this embodiment, the three-dimensional information includes, for example, a plurality of images captured from a plurality of shooting positions or angles such as a plurality of viewpoints or angles, and position information indicating the three-dimensional positions (e.g., three-dimensional coordinates) of a plurality of points respectively included in the plurality of images, i.e., three-dimensional point group information. Furthermore, the three-dimensional information may also include the plurality of images and sensor information indicating the positions and angles at which the plurality of images are captured. In addition, the three-dimensional positions of the plurality of points represented by the position information are also referred to as a three-dimensional point group.

[0034] Here, an example in which the three-dimensional information acquisition unit 110 is realized by an RGB-D (Red Green Blue-Depth) camera will be described.

[0035] Assume that a point l (l = 1, 2, ..., L) is obtained when an image f (f = 1, 2, ..., F) is captured at a certain viewpoint. f ) is located at q fl When fl It is obtained by the following formula (1).

[0036]

Mathematical formula 1

[0037] q fl =d fl K -1 q′ l (1)

[0038] F is the number of images captured during SLAM processing, and is an integer greater than or equal to 1.

[0039] L f is the number of pixels for obtaining the distance in each image, and is an integer greater than or equal to 1.

[0040] K is an internal parameter of the RGB-D camera, and is a parameter indicating the focus or lens center, etc.

[0041] q' l is the coordinate of the pixel of the image where the distance is obtained. For example, as shown in the following formula (2), q' l The number of horizontal pixels u when the upper left corner of the image is set as the origin l and the number of pixels in the vertical direction v l The coordinate representation of .

[0042]

Mathematical formula 2

[0043]

[0044] d fl is the distance from the shooting position to the object corresponding to point l in image f. fl Obtained from the Depth channel of the RGB-D camera.

[0045] Click q fl The camera position is set as the origin, the optical axis direction is set as the z-axis, the horizontal direction is set as the x-axis, and the vertical direction is set as the y-axis. Figure 2 The coordinate system shown.

[0046] Here, click q flIt is represented by a camera-referenced coordinate system with the shooting position as the origin. When integrating the point groups when each image is shot to create a large-scale point group, it is necessary to use a unified coordinate system to represent the point group. Generally speaking, the coordinate system of the first shot image (f=1) is used to represent the point group obtained when other images are shot.

[0047] The flow of a typical method for converting into the coordinate system of the image captured for the first time is as follows.

[0048] First, local features are extracted from the initial image and other images.

[0049] Next, the local feature quantities extracted from the two images are matched.

[0050] Next, the relationship between the two images (for example, the amount of parallel translation and the amount of rotation) is determined using the matched local feature quantity pairs.

[0051] The above method is a representative method. In addition, there is a method of integrating the relationship between images adjacent in time to obtain the positional relationship with the initial image.

[0052] The relationship between the first image obtained by this method and other images f is set as T 1→f When each point is expressed using the coordinate system when the image is first captured, the following equation (3) is obtained.

[0053]

Mathematical formula 3

[0054]

[0055] Then, the final three-dimensional point group becomes point r fl A collection of .

[0056] Then, the three-dimensional information acquisition unit 110 generates three-dimensional information including three-dimensional point group information indicating the three-dimensional point group detected as described above, sensor information indicating the position and angle of the image sensor of the RGB-D camera, and the image captured by the RGB-D camera. The three-dimensional information acquisition unit 110 provides the generated three-dimensional information to the foreground image generation unit 120.

[0057] The position and angle of the image sensor are the position and angle when each image is captured. n , set the angle to r n The image index is n=1, 2, ..., N, and the number of captured images is N (an integer greater than or equal to 2). Angle r n It can be expressed in the form of Euler angles, Quaternion or Euler-Rodrigues, etc. As long as the three-dimensional rotation angle is determined, it can be in any form.

[0058] The foreground image generating unit 120 generates a plurality of foreground images representing the recognition target based on the three-dimensional information acquired by the three-dimensional information acquiring unit 110 , and generates foreground image information representing the plurality of foreground images.

[0059] Figure 3 2 is a block diagram schematically showing the configuration of the foreground image generating unit 120 .

[0060] The foreground image generating unit 120 includes an identification target selecting unit 121 and an identification target cutting unit 122 .

[0061] The recognition target selection unit 121 selects a recognition target from the three-dimensional information acquired by the three-dimensional information acquisition unit 110 .

[0062] The three-dimensional point group represented by the three-dimensional point group information contained in the three-dimensional information or the image contained in the three-dimensional information may contain objects other than the recognition object. For example, when the recognition object is placed on the ground for shooting, the three-dimensional point group of the ground is also shot at the same time. In order to exclude information other than the recognition object, the recognition object selection unit 121 selects the recognition object from the three-dimensional point group or the image. As a selection method, there are the following two modes.

[0063] The first mode is manual selection.

[0064] The recognition target selection unit 121 displays a three-dimensional point group or an image on a display unit (not shown) based on the three-dimensional information, and receives selection of a recognition target range from the user via an input unit (not shown).

[0065] For example, the image contained in the three-dimensional point group information is Figure 4 The case of the image 170 shown will be described as an example. The image 170 includes an identification object 171 .

[0066] In this case, if Figure 5 As shown, the user specifies four points S in image 170. A,1 , S A,2 , S A,3 , S A,4 , thereby selecting the identification object 171. At this time, Figure 5 As shown, not all recognition objects 171 may be selected at once. Figure 5 This is an example of selecting only the face A that is one face constituting the recognition target 171. After selecting the face A, the user can select the faces B and C respectively, thereby finally selecting all the recognition targets 171.

[0067] Here, the recognition object selection unit 121 specifies the point S A,1 , S A,2 , S A,3 , SA,4 In the case of the surface A, the above equations (1) and (3) are used to find the relationship between these points S. A,1 , S A,2 , S A,3 , S A,4 The corresponding three-dimensional position is the three-dimensional coordinate. Assume that the three-dimensional coordinate obtained here is expressed using a coordinate system based on the time when the shooting starts. As a result, the three-dimensional coordinate S' is obtained. A,1 , S' A,2 , S' A,3 , S' A,4 .

[0068] Similarly, select faces B and C. In addition, Figure 4 In the case where it is not easy to select face B or face C in the image 170 shown, for example Figure 6 As shown, the recognition target selection unit 121 may receive a selection by displaying an image 172 captured from another viewpoint on a display unit (not shown).

[0069] By selecting surface B and surface C, find the three-dimensional coordinate S' of surface B B,1 , S' B,2 , S' B,3 , S' B,4 The three-dimensional coordinates S' of surface C C,1 , S' C,2 , S' C,3 , S' C,4 .

[0070] When accepting selection using a three-dimensional point group, the user directly specifies the three-dimensional coordinates S′ of the surface A from the three-dimensional point group displayed on the display unit (not shown). A,1 , S' A,2 , S' A,3 , S' A,4 , the three-dimensional coordinates S' of surface B B,1 , S' B,2 , S' B,3 , S' B,4 The three-dimensional coordinates S' of surface C C,1 , S' C,2 , S' C,3 , S' C,4 .

[0071] The recognition target selection unit 121 generates the three-dimensional coordinates S′ representing the surface A determined as described above. A,1 , S' A,2 , S' A,3 , S' A,4 , the three-dimensional coordinates S' of surface B B,1 , S' B,2 , S' B,3 , S'B,4 The three-dimensional coordinates S' of surface C C,1 , S' C,2 , S' C,3 , S' C,4 The identification target range information is provided to the identification target cutting unit 122. The identification target range information is information indicating the selected identification target range.

[0072] The second mode for selecting the recognition object is semi-automatic selection.

[0073] The recognition object selection unit 121 may also use a region segmentation technique such as GrabCut, a well-known algorithm, to find the boundary between the recognition object and other parts to find the recognition object range. Figure 7 As shown in the image 173, point groups S1 to S2 representing the boundaries between the recognition target and other parts are obtained. M (M is an integer greater than or equal to 2). Then, the recognition target selection unit 121 sets these point groups S1 to S M The point group S'1~S' obtained by converting into three-dimensional coordinates M is the range of the recognition object 171. As described above, it is converted into three-dimensional coordinates using equations (1) and (3).

[0074] The recognition target selection unit 121 generates the point groups S'1 to S' determined as described above. M The three-dimensional position of the recognition target range information represents the recognition target range, and the recognition target range information is provided to the recognition target cutting unit 122.

[0075] As described above, the recognition object selection unit 121 determines the recognition object range based on at least one image included in the three-dimensional information, but the recognition object can also be determined using the position information included in the three-dimensional information. In this case, the recognition object selection unit 121 generates plane information representing multiple planes based on the position information included in the three-dimensional information. The plane information represents a plane corresponding to the surface of the object. Then, the recognition object selection unit 121 displays the generated plane on the display unit, accepts the selection of the recognition object, or finds the boundary based on the generated plane.

[0076] As a representative method for detecting a plane, there is RANSAC (RANdom Sample Consensus). The method of detecting a plane using RANSAC is as follows.

[0077] First, select several points from the three-dimensional point group to create a plane.

[0078] Next, if a certain number of points exist on the plane, the produced plane is adopted, and if a certain number of points do not exist on the plane, the produced plane is discarded.

[0079] Next, the above-mentioned process is repeated to detect a plurality of planes.

[0080] The identification target cutting unit 122 cuts out images of the identification target range corresponding to the identification target selected by the identification target selecting unit 121 from the plurality of images included in the three-dimensional information, thereby generating a plurality of foreground images from the plurality of images.

[0081] For example, the identification object cutting unit 122 projects the identification object range represented by the identification object range information onto the image contained in the three-dimensional information, and cuts out the image of the projected range, thereby cutting out the image of the range corresponding to the selected identification object, i.e., the identification object image, from the multiple images contained in the three-dimensional information.

[0082] The following formula (4) shows a method of projecting the point S' included in the recognition target range information onto the image n.

[0083]

Mathematical formula 4

[0084] s n =K n P[r n |p n ]s′ (4)

[0085] Here, [r n |p n ] is a matrix generated based on the shooting position or angle of image n. P is a matrix of a normal coordinate system that converts homogeneous coordinates into non-homogeneous coordinates, and is a matrix represented by the following formula (5).

[0086]

Mathematical formula 5

[0087]

[0088] In addition, Kn represents the internal parameters of the camera corresponding to image n, s n Represents the position when point S' is projected onto image n.

[0089] After the recognition object cutting unit 122 obtains the recognition object range in each image using the above method, it cuts out the range. Figure 5 As shown in FIG. 1 , when face A, face B, and face C are selected, the entire face is cut out as the recognition target range. Figure 6 The corresponding range is cut out from the image 172 captured from different viewpoints. Figure 6 In the calculation, Figure 5The range corresponding to the selected range is cut out, and the range corresponding to the recognition object is cut out.

[0090] As described above, the recognition object cutting unit 122 cuts out an image of a selected range from all images included in the three-dimensional information and sets the cut image as a foreground image, thereby generating a large number of foreground images with a low processing load. Then, the recognition object cutting unit 122 provides image data representing the foreground image, that is, foreground image data, to the image synthesis unit 140.

[0091] The background image DB 130 stores background image data which is image data of a background image.

[0092] The image synthesis unit 140 synthesizes the foreground image represented by the foreground image data provided by the foreground image generation unit 120 with the background image represented by the background image data stored in the background image DB 130, thereby setting the synthesized image as a learning image, and generates learning image data representing the learning image. In other words, the image synthesis unit 140 synthesizes each of the background images in the plurality of background images with each of the foreground images in the plurality of foreground images, thereby generating a plurality of synthesized images.

[0093] Specifically, the image synthesis unit 140 overlaps the foreground image on the background image. Both the background image and the foreground image can be color images composed of three channels, namely red, blue and green, or grayscale images with only one channel. In addition, the background image and the foreground image can also be images with distance and also have a depth channel.

[0094] In addition, the image synthesis unit 140 may also overlap the foreground image at various positions of the background image to further expand the data.

[0095] The learning image DB 150 stores the learning image data generated by the image synthesis unit 140 .

[0096] Figure 8 2 is a block diagram showing a hardware configuration example of the information processing device 100 .

[0097] The information processing apparatus 100 includes a computer 1 , an output device 2 , an input device 3 , and a sensor 4 .

[0098] The computer 1 is a computer that can be composed of a memory and a processor such as a CPU (Central Processing Unit) that executes a program stored in the memory. Such a program can be provided via a network or recorded in a recording medium. That is, such a program can be provided as a program product, for example.

[0099] Furthermore, a part of the computer 1 may be constituted by a processing circuit such as a single circuit, a complex circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuits) or an FPGA (Field Programmable Gate Array).

[0100] here, Figure 1 The foreground image generating unit 120 , the background image DB 130 , the image synthesizing unit 140 , and the learning image DB 150 shown can be realized by the computer 1 .

[0101] The output device 2 is a device that functions as an output unit that displays various screen images, such as a display.

[0102] The input device 3 is a device that functions as an input unit such as a mouse, a keyboard, or a touch panel.

[0103] The sensor 4 is a device that functions as a three-dimensional information acquisition unit 110 that acquires three-dimensional information.

[0104] In Embodiment 1, the sensor 4 is implemented by an RGB-D camera. Alternatively, the sensor 4 can be implemented by a combination of an IMU (Inertial Measurement Unit) or a sensor that can obtain distance using infrared, LiDAR (Light Detection and Ranging) or ultrasonic waves and a camera that can capture a two-dimensional image.

[0105] Fig. 9 It is a flowchart showing the operation of the information processing device 100 .

[0106] The information processing device 100 repeatedly performs the following process as many times as the number of background images represented by the background image data stored in the background image DB 130 ( S10 ).

[0107] The three-dimensional information acquisition unit 110 acquires three-dimensional information ( S11 ).

[0108] The foreground image generating unit 120 generates a plurality of foreground images based on the three-dimensional information acquired by the three-dimensional information acquiring unit 110 ( S12 ).

[0109] The image synthesis unit 140 repeatedly performs the following processing as many times as the number of foreground images generated by the foreground image generation unit 120 ( S13 ).

[0110] The image synthesis unit 140 synthesizes the foreground image represented by the foreground image data generated by the foreground image generation unit 120 with the background image represented by the background image data stored in the background image DB 130 to generate a learning image ( S14 ). The learning image data representing the learning image generated in this way is stored in the learning image DB 150 .

[0111] Fig.10 It is shown Fig. 9 Flowchart of the action of generating a foreground image in step S12.

[0112] First, the recognition target selection unit 121 repeatedly performs the following processing until all recognition target ranges are acquired from the three-dimensional information acquired by the three-dimensional information acquisition unit 110 ( S20 ).

[0113] The recognition target selection unit 121 determines a recognition target range based on the three-dimensional point group or image included in the three-dimensional information ( S21 ).

[0114] Next, the recognition target cutting unit 122 repeatedly performs the following processing as many times as the number of images included in the three-dimensional information ( S22 ).

[0115] The recognition target cutting unit 122 projects the recognition target range determined by the recognition target selecting unit 121 onto the image ( S23 ).

[0116] Then, the recognition target cutting unit 122 cuts out the projected range from the image, thereby setting the cut-out image as a foreground image ( S24 ).

[0117] As described above, according to the present embodiment, a large amount of learning data can be produced using three-dimensional information obtained in a general environment. Therefore, there is no need to prepare a special environment as in the prior art. In addition, the recognition object range is determined based on one image, and thus the same range can be cropped from other images, so a large number of images can be produced with less labor and time. There is no need for the labor and time of manually producing a high-precision three-dimensional model or the labor and time of selecting the recognition object range for all images as in the past.

[0118] Description of symbols

[0119] 100: information processing device; 110: three-dimensional information acquisition unit; 120: foreground image generation unit; 130: background image DB; 140: image synthesis unit; 150: learning image DB; 121: recognition object selection unit; 122: recognition object cutting unit.

Claims

1. An information processing device, characterized in that: The information processing device comprises: a three-dimensional information acquisition unit that acquires three-dimensional information including a plurality of images; an identification object selection unit, which selects an identification object from the three-dimensional information; an identification object cutting unit for cutting out images of a range corresponding to the selected identification object, that is, an identification object range, from the plurality of images, thereby generating a plurality of foreground images from the plurality of images; A background image storage unit storing a plurality of background images; as well as an image synthesis unit that synthesizes each of the plurality of foreground images with each of the plurality of background images to generate a plurality of synthesized images, The recognition target selection unit generates recognition target range information representing the recognition target range by three-dimensional positions of a plurality of points included in the at least one image based on the at least one image included in the plurality of images. The recognition target cutting unit projects the recognition target range onto each of the plurality of images, thereby determining the recognition target range in each of the plurality of images.

2. The information processing device according to claim 1, characterized in that The plurality of images are images obtained by photographing the recognition object from different photographing positions or different angles.

3. The information processing device according to claim 1 or 2, characterized in that: The image synthesis unit synthesizes one foreground image included in the multiple foreground images at different positions in one background image included in the multiple background images, thereby generating multiple synthesized images based on the combination of the one background image and the one foreground image.

4. An information processing device, characterized in that: The information processing device comprises: a three-dimensional information acquisition unit that acquires three-dimensional information including a plurality of images and position information indicating the three-dimensional positions of a plurality of points respectively included in the plurality of images; an identification object selection unit, which selects an identification object from the three-dimensional information; an identification object cutting unit for cutting out images of a range corresponding to the selected identification object, that is, an identification object range, from the plurality of images, thereby generating a plurality of foreground images from the plurality of images; A background image storage unit storing a plurality of background images; as well as an image synthesis unit that synthesizes each of the plurality of foreground images with each of the plurality of background images to generate a plurality of synthesized images, The recognition object selection unit selects three-dimensional positions of a plurality of points included in at least one of the plurality of images based on the position information, and generates recognition object range information indicating the recognition object range by the selected positions. The recognition target cutting unit projects the recognition target range onto each of the plurality of images, thereby determining the recognition target range in each of the plurality of images.

5. The information processing device according to claim 4, characterized in that The plurality of images are images obtained by photographing the recognition object from different photographing positions or different angles.

6. The information processing device according to claim 4 or 5, characterized in that: The image synthesis unit synthesizes one foreground image included in the multiple foreground images at different positions in one background image included in the multiple background images, thereby generating multiple synthesized images based on the combination of the one background image and the one foreground image.

7. A computer-readable recording medium storing a computer program, which, when executed by a processor, Selecting objects to be identified from 3D information containing multiple images, The images of the range corresponding to the selected recognition object, that is, the recognition object range, are cut out from the plurality of images, thereby generating a plurality of foreground images from the plurality of images. The plurality of composite images are generated by synthesizing each of the plurality of foreground images in each of the plurality of background images, wherein: When selecting the recognition object from the three-dimensional information, based on at least one image included in the plurality of images, recognition object range information representing the recognition object range by three-dimensional positions of a plurality of points included in the at least one image is generated, When the plurality of foreground images are generated from the plurality of images, the recognition target range is projected onto the plurality of images respectively, thereby determining the recognition target range in the plurality of images respectively.

8. A computer-readable recording medium storing a computer program, which, when executed by a processor, selecting a recognition object from three-dimensional information including a plurality of images and position information indicating the three-dimensional positions of a plurality of points respectively included in the plurality of images, The images of the range corresponding to the selected recognition object, that is, the recognition object range, are cut out from the plurality of images, thereby generating a plurality of foreground images from the plurality of images. The plurality of composite images are generated by synthesizing each of the plurality of foreground images in each of the plurality of background images, wherein: When selecting the recognition object from the three-dimensional information, the three-dimensional positions of a plurality of points included in at least one of the plurality of images are selected based on the position information, and recognition object range information representing the recognition object range by the selected positions is generated. When the plurality of foreground images are generated from the plurality of images, the recognition target range is projected onto the plurality of images respectively, thereby determining the recognition target range in the plurality of images respectively.

9. An information processing method, wherein: Get three-dimensional information containing multiple images, selecting an identification object from the three-dimensional information, The images of the range corresponding to the selected recognition object, that is, the recognition object range, are cut out from the plurality of images, thereby generating a plurality of foreground images from the plurality of images. The plurality of composite images are generated by synthesizing each of the plurality of foreground images in each of the plurality of background images, wherein: When selecting the recognition object from the three-dimensional information, based on at least one image included in the plurality of images, recognition object range information representing the recognition object range by three-dimensional positions of a plurality of points included in the at least one image is generated, When the plurality of foreground images are generated from the plurality of images, the recognition target range is projected onto the plurality of images respectively, thereby determining the recognition target range in the plurality of images respectively.

10. An information processing method, wherein: acquiring three-dimensional information including a plurality of images and position information indicating the three-dimensional positions of a plurality of points respectively included in the plurality of images, selecting an identification object from the three-dimensional information, The images of the range corresponding to the selected recognition object, that is, the recognition object range, are cut out from the plurality of images, thereby generating a plurality of foreground images from the plurality of images. The plurality of composite images are generated by synthesizing each of the plurality of foreground images in each of the plurality of background images, wherein: When selecting the recognition object from the three-dimensional information, the three-dimensional positions of a plurality of points included in at least one of the plurality of images are selected based on the position information, and recognition object range information representing the recognition object range by the selected positions is generated. When the plurality of foreground images are generated from the plurality of images, the recognition target range is projected onto the plurality of images respectively, thereby determining the recognition target range in the plurality of images respectively.

Images