JP7799478B2 - Information processing device, information processing method, and program - Google Patents

Description

This disclosure relates to time-series processing of shape data representing the three-dimensional shape of an object.

The shape data of objects used in fields such as virtual viewpoint video and 3D printing is placed in a space in which the objects exist, with the same number of data as the objects themselves. In this case, the larger the space in which the objects exist, the more difficult it becomes to allocate enough memory to place all of the basic unit elements of the shape data (for example, voxels when the shape data is volume data) within that space. Therefore, sparse data management is used, in which unit elements are assigned only to areas in the space where objects exist, and no unit elements are assigned to areas where no objects exist. Patent Document 1 discloses a method of generating low-resolution voxel models for each of multiple objects existing in a space, and then 3D modeling the bounding boxes of these objects using smaller voxels to obtain high-resolution volume data.

Japanese Patent Application Laid-Open No. 2021-71749

The technology in Patent Document 1 aims to speed up the overall processing by managing volume data sparsely and adjusting the spatial resolution in the process of generating a voxel model for each object. However, the volume data handled by Patent Document 1 is discontinuous in the time direction, making it difficult to speed up time-series processing of volume data collected over time. This is because time-series processing requires determining, one by one, to which voxel model a voxel contained in a voxel model in one frame belongs in another frame. Furthermore, the greater the number of objects present in a space, the more frequently the process of searching between frames for the voxel model to which each voxel belongs, and the longer the processing time required.

The information processing device according to the present disclosure comprises an acquisition means for acquiring time-series shape data consisting of a group of temporally consecutive frames, each frame containing shape data indicating the three-dimensional shape of an object; a setting means for setting a frame range from the group of frames constituting the acquired time-series shape data in accordance with the content of a predetermined time-series processing ; and a processing means for performing the predetermined time-series processing on the set frame range, wherein the processing means performs the predetermined time-series processing by excluding, from the shape data belonging to frames other than a frame of interest in the set frame range, shape data that does not satisfy a predetermined condition between the shape data of interest belonging to the frame of interest .

The technology disclosed herein can reduce the load of time-series processing of shape data collected over time for multiple objects in a space.

FIG. 1 is a diagram showing an example of a hardware configuration of an information processing apparatus. FIG. 2 is a block diagram showing the software configuration (functional configuration) of the information processing apparatus. FIG. 1A is a conceptual diagram of time-series volume data, and FIG. 1B is a diagram illustrating position information held by a bounding box. 10 is a flowchart showing the flow of time-series processing according to the first embodiment. 5A and 5B are schematic diagrams illustrating a case where the time-series processing according to the first embodiment is applied to time-series volume data. FIG. 10 is an explanatory diagram of a correction to the result of the smoothing process. 10 is a flowchart showing the flow of time-series processing according to the second embodiment. 10A and 10B are diagrams illustrating the reuse of collision detection results; 10A and 10B are diagrams showing an example of a determination result list. 11 is a flowchart showing the flow of time-series processing according to the third embodiment. FIG. 11 is a schematic diagram illustrating a case where the time-series processing according to the third embodiment is applied to time-series volume data.

The following describes embodiments of the present disclosure with reference to the accompanying drawings. Note that the following embodiments do not limit the present disclosure, and not all of the combinations of features described in the embodiments are necessarily essential to the solution of the invention.

Before proceeding with the description of each embodiment, let's review the concept of object (subject) shape data. This embodiment will be described using volume data as an example, which uses voxels as the basic unit elements representing shape. In the case of volume data, the object's shape is expressed using multiple unit elements (voxels) that divide the object's circumscribing rectangle (commonly called a "bounding box") set in three-dimensional space into a grid. Each voxel has information (hereinafter referred to as a "voxel value") that indicates whether it exists within the object's physical entity in three-dimensional space. In this embodiment, voxel values are described as having binary information of "0" or "1." A voxel with a voxel value of "1" is referred to as an "ON voxel," as it is partially or completely inside the object. A voxel with a voxel value of "0" is referred to as an "OFF voxel," as it is completely outside the object (i.e., completely outside the object).

In the description of each embodiment, time-series shape data consisting of multiple frames consecutive in the time direction, with one or more bounding boxes set for each frame, and using voxels as unit elements, will be referred to as "time-series volume data." Note that some frames may not contain shape data. Note that the data formats to which this embodiment can be applied are not limited to volume data. In other words, it can also be applied to point cloud data, which represents the three-dimensional shape of an object as a set of points, mesh data, which represents a set of polygons, and so on.

[Embodiment 1]
<Hardware configuration>
FIG. 1 is a diagram showing an example of the hardware configuration of an information processing device 100 according to this embodiment, which performs time-series processing on time-series volume data. In FIG. 1 , a CPU 101 uses a RAM 102 as a work memory, executes programs stored in a ROM 103 and a hard disk drive (HDD) 105, and controls the operation of each block (described later) via a system bus 112. An HDD I/F 104 is an interface, such as a serial ATA (SATA), that connects a secondary storage device such as the HDD 105 or an optical disk drive. The CPU 101 can read data from and write data to the HDD 105 via the HDD I/F 104. Furthermore, the CPU 101 can load data stored in the HDD 105 onto the RAM 102, and conversely, can save data loaded onto the RAM 102 onto the HDD 105. The CPU 101 can then execute the data loaded onto the RAM 102 as a program. The input I/F 106 is a serial bus interface, such as USB or IEEE 1394, that connects input devices 107, such as a keyboard, mouse, digital camera, and scanner. The CPU 101 can read data from the input device 107 via the input I/F 106. The output I/F 108 is a video output interface, such as DVI or HDMI (registered trademark), that connects the information processing apparatus 100 to a display, which is an output device. The CPU 101 can send data to the display via the output I/F 108 and display a predetermined video on the display. The network I/F 110 is an interface that connects to a network, such as the Internet or a LAN. The CPU 101 can exchange various data with an external server, etc., connected to the LAN or the Internet via the network I/F 110.

<Software configuration>
2 is a block diagram showing the software configuration of the information processing device 100, more specifically, the functional configuration for executing predetermined time-series processing on input time-series volume data. The various functions shown in FIG. 2 are realized by the CPU 101 executing dedicated programs.

The information processing device 100 of this embodiment includes a data acquisition unit 201, a frame range setting unit 202, a reference model determination unit 203, and a time series processing unit 204.

The data acquisition unit 201 acquires time-series volume data consisting of multiple frames collected over time, for example, from another information processing device (not shown) on the LAN via the network I/F 110. Figure 3(a) is a conceptual diagram of time-series volume data. A frame at a certain time T contains bounding boxes for each object, the number of which is equal to the number of objects present in the target space at that time T. Here, as shown in Figure 3(b), the three-dimensional position of each bounding box 301 is expressed by relative coordinates 302 with respect to the origin coordinates in a common space. Note that if multiple objects are in contact with each other, one bounding box will correspond to multiple objects.

The frame range setting unit 202 sets a frame range consisting of two or more consecutive frames to be processed when performing time-series processing on the time-series volume data acquired by the data acquisition unit 201, according to the content and intent of the time-series processing. Here, time-series processing means comparing the voxel sets (hereinafter referred to as "voxel models") for each object contained in each frame of the time-series volume data between multiple frames, and modifying the voxel model in a specific frame or changing the number of frames. Modification of the voxel model includes changes to the shape, position, voxel values, and assigned labels of the voxel model, and changing the number of frames includes adding (interpolating) or deleting frames.

The reference model determination unit 203 determines a voxel model of another frame to be referenced when changing the voxel model (target voxel model) that is the target of the smoothing process in the time direction, based on the frame range set by the frame range setting unit 202. A specific method for this determination will be described later.
The time series processing unit 204 performs predetermined time series processing on a voxel model of interest in a frame of interest within a set frame range, using voxel models of other frames determined by the reference model determination unit 203. In this embodiment, an example will be described in which smoothing processing in the time direction is performed as the predetermined time series processing.

<Details of time series processing>
4 is a flowchart showing the flow of executing smoothing processing in the time direction on input time-series volume data according to this embodiment. Note that in the following description, the symbol "S" represents a step.
First, in S401, the data acquisition unit 201 acquires time-series volume data. At this time, the data may be received and acquired from another information processing device (not shown), or may be previously received and saved and read from the HDD 105. Here, it is assumed that time-series volume data 500 consisting of multiple frames, created based on video captured of a space in which three dynamic objects exist, as shown in FIG. 5, has been acquired. In FIG. 5, white squares 501 represent off-voxels, and black squares 502 represent on-voxels. Although FIG. 5 schematically represents voxels as a two-dimensional grid, the actual entity being handled is a three-dimensional volume.

In S402, the frame range setting unit 202 sets a frame range to be subjected to the time-direction smoothing process for the time-series volume data acquired in S401. In the time-direction smoothing process, the frame to which the voxel model to be smoothed belongs is set as the frame of interest, and voxel models belonging to its adjacent frames (the previous frame and the next frame) are referenced. Therefore, in this embodiment, three consecutive frames 503 as shown in Figure 5 are set as the frame range to be processed. However, since there is no frame previous to the frame of interest immediately after the start of processing and no frame subsequent to the frame of interest immediately before the end of processing, two consecutive frames are set as the frame range. Note that the frame range to be set can be associated with the processing content and stored in advance in the HDD 105, etc.

In S403, the reference model determination unit 203 sets the frame to which the voxel model to be smoothed belongs, from the frame range set in S402, as the frame of interest. In the example of Figure 5, the second (t=n) frame of the three frames 503 is set as the frame of interest.

In S404, the reference model determination unit 203 sets a target voxel model to be smoothed from one or more voxel models included in the target frame. Here, in the example of Figure 5, we will continue the explanation assuming that voxel model 504a is set as the target voxel model from the three voxel models 504a to 504c included in the frame t=n, which is the target frame.

In S405, the reference model determination unit 203 performs a reference determination between a voxel model included in a frame other than the frame of interest within the set frame range and the voxel model of interest. For this reference determination, for example, collision determination using the bounding box method is used. In the bounding box method, when the following equation (1) is satisfied for two voxel models v1 and v2, it is determined that the two voxel models v1 and v2 collide.

(v1x_max>v2x_min)∧(v2x_max>v1x_min)∧(v1y_max>v2y_min) ∧(v2y_max>v1y_min)∧(v1z_max>v2z_min)∧(v2z_max>v1z_min)
...Formula (1)
In the above formula (1), v1x_max, v1x_min, v1y_max, v1y_min, v1z_max, and v1z_min represent the maximum and minimum values of the x, y, and z axes for the voxel model v1. Furthermore, v2x_max, v2x_min, v2y_max, v2y_min, v2z_max, and v2z_min represent the maximum and minimum values of the x, y, and z axes for the voxel model v2. In the example of FIG. 5 , assume that the voxel model 504 is the target voxel model. In this case, for example, the voxel model 504 may be designated as the voxel model V1, and the voxel models 505a to 505c included in the frame at t=n−1 and the voxel models 506a to 506c included in the frame at t=n+1 may be designated as the voxel model V2, and sequential determination may be performed. Note that collision detection may be performed using other methods, such as the bounding sphere method, which uses the radii of two voxel models to perform collision detection. Voxel models that are determined not to collide by this collision detection are excluded from the processing target in S407, which will be described later. In an example where voxel model 504 is the target voxel model, voxel model 505a is determined to collide with voxel model 504 in frame t=n-1, and voxel models 506a and 506b are determined to collide with voxel model 504 in frame t=n-1. As a result, voxel models 505b, 505c, and 506c are excluded from the processing target in S407.

In S406, it is determined whether collision detection has been completed between the voxel model of interest and each voxel model included in frames other than the frame of interest within the set frame range. If collision detection has been completed with all voxel models included in other frames, the process proceeds to S407; if not, the process returns to S405 to continue collision detection with the remaining voxel models.

In S407, the time series processing unit 204 performs smoothing processing in the time direction on the voxel model of interest using the voxel model determined to collide in the collision determination in S405. Specifically, it references voxel models determined to collide in other frames, and uses a majority filter to change the on voxel to an on voxel if there are on voxels in two or more frames, or to an off voxel if there are not. As a result, in the example of Figure 5, voxel model 504 in the frame at t=n is changed to voxel model 507.

In S408, it is determined whether smoothing processing has been completed for all voxel models included in the frame of interest. If smoothing processing has been completed for all voxel models, proceed to S409; if not, return to S404 to determine the next voxel model of interest and continue processing.

In S409, it is determined whether the smoothing process for the input time-series volume data is complete. If it is complete, this process ends. If it is not complete, the process returns to S402, where the next frame range is determined and processing continues.

The above is the flow of the time-direction smoothing process for time-series volume data according to this embodiment. In this way, this embodiment makes it possible to efficiently perform time-direction smoothing process for time-series volume data.

<Modification>
In one frame, multiple objects represented as separate voxel models 601 and 602 (as shown in FIG. 6( a) ) may be represented by a single voxel model 603 (as shown in FIG. 6( b) ) due to their proximity in another frame. Performing the above-described time-direction smoothing process on time-series volume data including such frames may result in an On voxel group 604 corresponding to part of another object being mixed into a voxel model 601′ after the smoothing process on the voxel model 601, as shown in FIG. 6( c) . If the smoothing process results in two or more discontinuous On voxel groups being included within a single bounding box, a correction process may be additionally performed to delete the extra On voxel groups, thereby correcting the shape represented by the voxel model. Specifically, in the case of FIG. 6( c) , the On voxel groups highly similar in shape to the On voxel groups in the voxel model 601 (as shown in FIG. 6( a) ) before the smoothing process are retained, and all On voxel groups 604 with low similarity in shape are replaced with Off voxels. The similarity between voxel models may be determined using, for example, the maximum value of a three-dimensional cross-correlation function or the distance between centers of gravity.

In addition, while this embodiment has been described using an example in which voxel values are binary information, voxel values may also be multi-valued or continuous. When voxel values are multi-valued or continuous, the time series processing unit 204 can use a mean filter or Gaussian filter instead of a majority filter to perform time series processing using halftone information for the voxel values of each voxel that makes up the voxel model.

As described above, according to this embodiment, the voxel models to be referenced in time-direction smoothing are narrowed down in advance. This allows for efficient time-direction smoothing processing of time-series volume data, thereby achieving faster processing.

[Embodiment 2]
Next, an aspect in which the result of collision determination (S405) is stored and the number of times that subsequent collision determinations are executed is reduced to perform smoothing processing in the time direction more efficiently will be described as embodiment 2. Note that description of the content common to embodiment 1, such as the hardware configuration of the information processing device 100, will be omitted or simplified, and the following description will focus on the differences.

<Details of time series processing>
7 is a flowchart showing the flow of executing smoothing processing in the time direction on input time-series volume data according to this embodiment. Note that in the following description, the symbol "S" represents a step.

In S701, similar to S401 in the first embodiment, the data acquisition unit 201 acquires time-series volume data. Subsequently, in S702, the reference model determination unit 203 initializes a list for storing the collision detection results (hereinafter referred to as the "determination result list"). Here, the determination result list is array information for reusing the collision detection results the next time (the smoothing step targeting the newly set frame range).

Steps S703 to S705 correspond to steps S402 to S404 in embodiment 1, respectively. That is, the frame range setting unit 202 first sets a frame range to be processed for the time-series volume data (S703). Then, a frame of interest is set from the set frame range (S704), and then a voxel model of interest within the frame of interest is set (S705).

In S706, the reference model determination unit 203 refers to the determination result list and performs reference determination (here, collision determination) between each voxel model included in other undetermined frames within the set frame range and the target voxel model. Assume that the frame range 801 shown in FIG. 8 was set in S703 for the time-series volume data 500 shown in FIG. 5, as described in the first embodiment. At this point, the determination result list shown in FIG. 9(a) has been obtained, representing the results of the collision determination in the previous process (the results of the collision determination when processing was performed on the frame range 503 in FIG. 5). Here, "Vol1," "Vol2," and "Vol3" in the determination result list in FIG. 9(a) are labels assigned to each voxel model within the frame. Note that the same voxel model does not necessarily have the same label across frames. In the determination result list in FIG. 9(a), "1" indicates that a collision has occurred, and "0" indicates that a collision has not occurred. Currently, since the processing target is frame range 801, the frame of interest is frame t=n+1. In this case, collision detection is required between the voxel model of interest in t=n+1 and each voxel model included in the previous frame t=n and each voxel model included in the subsequent frame t=n+2. In this case, in this embodiment, the detection result list shown in Figure 9(a) is referenced, and collision detection is not performed with each voxel model included in frame t=n for which detection processing has been completed. In other words, collision detection is only performed with each voxel model included in frame t=n+2 for which detection processing has not been completed.

In S707, it is determined whether collision detection has been completed between the voxel model of interest and each voxel model included in frames other than the frame of interest within the set frame range. If collision detection has been completed with all voxel models included in other frames, the process proceeds to S708; if not, the process returns to S706 to continue collision detection with the remaining voxel models.

In S708, the reference model determination unit 203 writes the result of the collision detection made in S706 into the detection result list, updating the list. When processing is performed for the frame range 801 described above, a new detection result list is created, showing the results of collision detection when t = n + 1 is the frame of interest, as shown in Figure 9(b). The shaded areas in the detection result list in Figure 9(b) indicate that collision detection has already been made for the frame at t = n.

In S709, the time series processing unit 204 references the determination result list updated in S708 and performs smoothing on the target voxel model using the voxel model determined to be in collision.

In S710, it is determined whether smoothing processing has been completed for all voxel models included in the frame of interest. If smoothing processing has been completed for all voxel models, proceed to S710; if not, return to S705 to determine the next voxel model of interest and continue processing.

In S711, it is determined whether the time-direction smoothing process for the time-series volume data has been completed. If it has been completed, this process ends. If it has not been completed, the process returns to S703, where the next frame range is determined and processing continues.

This completes the flow of the time-direction smoothing process for time-series volume data according to this embodiment.

As described above, according to this embodiment, the most recent collision detection results are saved and reused, reducing the number of times collision detection is performed and enabling faster processing.

[Embodiment 3]
Next, a time-series process in which frame interpolation processing is performed to insert another frame between two adjacent frames in a frame group will be described as embodiment 3. Note that description of the content common to embodiment 1, such as the hardware configuration of the information processing device 100, will be omitted or simplified, and the following description will focus on the differences.

<Details of time series processing>
10 is a flowchart showing the flow of frame interpolation processing performed on input time-series volume data according to this embodiment. Note that in the following description, the symbol "S" denotes a step.

S1001 to S1004 correspond to S401 to S404 in embodiment 1, respectively. First, in S1001, the data acquisition unit acquires time-series volume data to be processed.

In the next step S1002, the frame range setting unit 202 sets the frame range to be processed for the input time-series volume data. In the case of this embodiment, which performs frame interpolation processing, two adjacent frames (frame t=n-2 and frame t=n-1) 1101 in the frame group are set, as shown in FIG. 11. Note that the frame range to be set (i.e., the position where the frames are to be interpolated) can be stored in advance in the HDD 105 or the like in association with the processing content.

In S1003, the reference model determination unit 203 sets one of the two frames that make up the frame range set in S1002 as the frame of interest. Here, the frame that is earlier in time is set as the frame of interest. In the example of Figure 11, of the two frames 1101, the frame at t = n-2 is set as the frame of interest.

In S1004, the reference model determination unit 203 sets a target voxel model to be used in the next determination process from among one or more voxel models included in the target frame. In the example of Figure 11, we will continue the explanation assuming that voxel model 1102 is set as the target voxel model from among the three voxel models included in the frame t=n-2, which is the target frame.

In S1005, the reference model determination unit 203 performs a reference determination between each voxel model included in a non-target frame (a frame later in time) within the set frame range and the target voxel model. In this embodiment, the reference determination is performed by determining the spatial distance and the degree of shape similarity between the target voxel model in the target frame and each voxel model in the next frame. Then, among the voxel models included in the non-target frame, only voxel models with a close spatial distance and a high degree of shape similarity are used to create the interpolated voxel model described below (other voxel models are excluded from the creation process). Here, for example, the distance determination may be performed by calculating the distance between the centers of gravity of both voxel models and determining that the spatial distance is close if the result is below a threshold. Furthermore, for example, the shape similarity determination may be performed by calculating the three-dimensional cross-correlation between both voxel models and determining that the degree of shape similarity is high if the maximum value is above a threshold. In the example of Figure 11, the above two judgment processes are performed between voxel model 1102 (voxel model of interest) and three voxel models 1103 to 1105 in the frame at t = n-1, and it is determined that voxel model 1103 satisfies both conditions.

In S1006, it is determined whether a complementary frame has already been inserted within the set frame range. If a complementary frame has not been inserted, the process proceeds to S1008; if a complementary frame has already been inserted, the process proceeds to S1009. Then, in S1008, the time series processing unit 204 creates a new frame and inserts it as an interpolated frame between the two frames that make up the set frame range. In the example of Figure 11, a complementary frame corresponding to t=n-1.5 is inserted between the frame at t=n-2 and the frame at t=n-1.

In S1009, the time series processing unit 204 creates a voxel model that interpolates between the voxel model of interest in the previous frame and the voxel model in the subsequent frame that was determined in S1005 to satisfy both the distance determination and shape matching conditions. Specifically, calculations are performed on the two voxel models v1 and v2 using the following equations (2) and (3).

Here, the timestamp of the frame to which voxel model v1 belongs is t1, the timestamp of the frame to which voxel model v2 belongs is t2, and the timestamp of the frame to which the interpolated voxel model belongs is t_int. In this case, the translational movement amount shift_int from the target voxel model in the interpolated voxel model can be calculated using the following equation (2):

shift_int = (x_shift*r, y_shift*r, z_shift*r) ... (Formula 2)
In the above formula (2), x_shift, y_shift, and z_shift represent the translational movement amounts from the two voxel models v1 to v2, and are calculated from the coordinates that provide the maximum value of the three-dimensional cross-correlation between the voxel models v1 and v2. The voxel value v_int of each voxel that constitutes the interpolated voxel model can be calculated using the following formula (3).

v_int (x, y, z) = v1(x, y, z) * r + v2(x + x_shift, y + y_shift, z + z_shift) * (1 - r)
...(Formula 3)
In the above formula (3), v1(x, y, z) and v2(x, y, z) represent the voxel values at the coordinates (x, y, z) of the voxel models v1 and v2, respectively. Note that r in the above formulas (2) and (3) is a variable expressed by the following formula (4).

r = (t2 - t_int) / (t2 - t1) ... Formula (4)
11, a voxel model 1106 is obtained as an interpolated voxel model that fills the gap between a voxel model 1102 in the frame at t=n-2 and a voxel model 1103 in the frame at t=n-1. The voxel model 1106 is then placed in the inserted interpolated frame at t=n-1.5.

In S1010, it is determined whether all voxel models included in the frame of interest have been processed. If all voxel models have been processed, proceed to S1011; if there are any unprocessed voxel models, return to S1004 to determine the next voxel model of interest and continue processing.

In S1011, it is determined whether frame interpolation processing for the time-series volume data is complete. If it is complete, this processing ends. If it is not complete, the processing returns to S1002, where the next frame range is determined and processing continues.

This completes the flow of frame interpolation processing for time-series volume data according to this embodiment.

According to this embodiment, the voxel models to be referenced during frame interpolation are narrowed down in advance. This allows frame interpolation processing of time-series volume data to be performed efficiently, resulting in faster processing.

[Other embodiments]
The present disclosure can also be realized by a process in which a program that realizes one or more functions of the above-described embodiments is supplied to a system or device via a network or a storage medium, and one or more processors in a computer of the system or device read and execute the program. The present disclosure can also be realized by a circuit (e.g., an ASIC) that realizes one or more functions.

100 Information processing device 201 Data acquisition unit 202 Frame range setting unit 203 Reference model determination unit 204 Time series processing unit

Claims (13)

an acquisition means for acquiring time-series shape data consisting of a group of temporally consecutive frames, each frame containing shape data indicating a three-dimensional shape of an object;
a setting means for setting a frame range from among a group of frames constituting the acquired time-series shape data in accordance with the content of a predetermined time-series process;
a processing means for performing the predetermined time series processing on the set frame range;
and
the processing means performs the predetermined time series processing by excluding shape data that does not satisfy a predetermined condition between the shape data of interest belonging to the frame of interest and the shape data of interest belonging to the frame of interest from the shape data of other frames other than the frame of interest within the set frame range.
1. An information processing device comprising: 2. The information processing device according to claim 1, further comprising a determination means for determining whether or not the shape data belonging to the other frame and the shape data of interest satisfy the predetermined condition according to the content of the predetermined time-series processing. the predetermined time series processing is a smoothing process in the time direction,
the setting means sets two or more consecutive frames including the frame of interest from among a group of frames constituting the time-series shape data as the frame range;
the determining means determines whether or not to refer to the shape data included in the other frames other than the frame of interest within the set frame range, and the shape data of interest; and the processing means performs the smoothing process in the time direction, excluding the shape data included in the other frames determined not to be referenced by the determining means.
3. The information processing apparatus according to claim 2 , wherein: Further, a storage means for storing information on the result of the reference determination is provided.
the determination means performs the reference determination on frames within the set frame range for which the reference determination result does not exist in the information stored by the storage means.
4. The information processing apparatus according to claim 3 , The information processing device according to claim 3, characterized in that the determination means performs collision determination using the bounding box method or the bounding sphere method as the reference determination. The information processing device of claim 5 further comprises a correction means for retaining only one piece of shape data that has a high degree of shape similarity to the shape data before smoothing, if the result of the time-direction smoothing process contains two or more pieces of discontinuous shape data within a single bounding box. the predetermined time-series processing is a frame interpolation processing,
the setting means sets two consecutive frames including the frame of interest from the frame group as the frame range;
the determining means performs a distance determination and a shape coincidence determination between the shape data of interest and shape data included in the other frames other than the frame of interest within the set frame range,
the processing means performs the frame interpolation process by excluding shape data included in the other frames that are determined by the determination means not to satisfy the conditions regarding distance and shape coincidence.
3. The information processing apparatus according to claim 2 , wherein: The determination means
In the distance determination, it is determined that the condition is satisfied when the distance between the centers of gravity is equal to or less than a threshold value;
In the shape matching determination, it is determined that the condition is satisfied when the maximum value of the three-dimensional cross-correlation is equal to or greater than a threshold value.
8. The information processing apparatus according to claim 7 , 9. The information processing apparatus according to claim 1, wherein the shape data is volume data that uses voxels as unit elements representing a shape. 9. The information processing apparatus according to claim 1, wherein the shape data is point cloud data that uses points as unit elements representing a shape. 9. The information processing apparatus according to claim 1, wherein the shape data is mesh data that uses polygons as unit elements representing a shape. an acquisition step of acquiring time-series shape data consisting of a group of temporally consecutive frames, each frame containing shape data indicating a three-dimensional shape of an object;
a setting step of setting a frame range from among a group of frames constituting the acquired time-series shape data in accordance with the content of a predetermined time-series process;
a processing step of performing the predetermined time series processing on the set frame range;
and
In the processing step, the predetermined time series processing is performed by excluding shape data that does not satisfy a predetermined condition between the shape data of interest belonging to the frame of interest and the shape data of interest belonging to the frame of interest from the shape data of other frames in the set frame range other than the frame of interest.
An information processing method comprising: A program for causing a computer to function as the information processing device according to any one of claims 1 to 11 .