Device, method and computer program for reconstructing a motion of an object

• US 9,754,400 B2
• Filed: 11/12/2013
• Issued: 09/05/2017
• Est. Priority Date: 11/22/2012
• Status: Expired due to Fees

• Current Assignee: Fraunhofer Gesellschaft Zur Foerderung Der Angewandten Forsching E.V.
• Original Assignee: Fraunhofer Gesellschaft Zur Foerderung Der Angewandten Forsching E.V.
• Inventors: Otto, Stephan, Bretz, Ingmar, Franke, Norbert, Von Der Gruen, Thomas, Mutschler, Christopher
• Primary Examiner(s): Nguyen, Anh-Tuan V
• Application Number:
  US14/646,663
• Publication Number:
  US 20150310656A1
• Time in Patent Office: 1,393 Days
• Field of Search: None
• CPC Class Codes: G06T 13/40 of characters, e.g. humans, animals or virtual beings
• CPC Class Codes: G06T 17/00 Three dimensional [3D] modelling, e.g. data description of 3D objects
• CPC Class Codes: G06V 40/23 Recognition of whole body movements, e.g. for sport training

Abstract

In a method for reconstructing a motion of an object from a sequence of motion pattern segments of a computer model of the object, a motion transition between an initial motion state and a final motion state of the object in a time interval of the motion is captured based on position data of the at least one sampling point which is received from the position marker. Further, at least one motion pattern segment corresponding to the motion transition is selected from a plurality of motion patterns of the computer model which are stored in a database such that the selected motion pattern segment leads with sufficient probability from the initial motion state to the final motion state for the time interval. Furthermore, an image of the motion of the object for the time interval is reconstructed using the initial motion state and the selected motion pattern segment.

Claims

1. A method for reconstructing a motion of an object from a sequence of motion pattern segments of a 3D computer model of the object, wherein each motion pattern segment corresponds to a different time interval of the motion and includes at least one frame, and wherein the object comprises at least one sampling point coupled to a position marker, the method comprising:

a) capturing a sequence of motion transitions between an initial motion state and a final motion state of the object in a time interval of the motion based on position data of the at least one sampling point which is received from the position marker;

b) selecting, based on possible motion states of a Markov model, at least one digital motion pattern segment corresponding to the sequence of motion transitions from a plurality of predefined digital motion patterns of the 3D computer model which are stored in a database and correspond to possible real motions of the object such that the at least one digital motion pattern segment leads with sufficient probability from the initial motion state to the final motion state for the time interval; and

c) reconstructing an image of the motion of the object for the time interval by displaying the 3D computer model on a display using the initial motion state and the at least one digital motion pattern segment, wherein

by selecting the at least one digital motion pattern segment, the most probable sequence of the possible motion states of the Markov model and the sequence of motion transitions is determined in order to reconstruct the motion of the object.

2. The method of claim 1, wherein acts a) to c) form a current iteration of an iterative method for reconstructing the motion, and wherein the initial motion state for the current iteration was reconstructed in a previous iteration of the iterative method so that the reconstructed image of the motion adds up from a plurality of digital motion pattern segments corresponding to selected and consecutive time intervals or iterations.

3. The method of claim 1, wherein, when selecting, captured motion transitions of the object are iteratively compared to the plurality of predefined digital motion patterns of the 3D computer model such that a hypothesis is determined for a current motion state of the object based on at least one past motion state of the object and a current transition probability which corresponds to a possible motion transition between at least two temporarily consecutive, discrete motion states of the object.

4. The method of claim 3, wherein the current motion state and the at least one past motion state of the object form a sequence of the Markov model.

5. The method of claim 1, wherein the at least one digital motion pattern segment is initially selected based on a subset of a set of possible motion states, wherein the subset is adjusted in order to be able to at least coarsely reconstruct the image of the motion, and wherein, following a coarse reconstruction, the image of the motion is finely reconstructed based on further states taken from the set of possible motion states outside the subset.

6. The method of claim 1, wherein, when selecting the at least one digital motion pattern segment, after a predefined period of time after capturing the sequence of motion transitions, a digital motion pattern segment is selected which, after this period of time, leads with the highest probability from the initial motion state to the final motion state for the time interval.

7. The method of claim 1, wherein the plurality of predefined digital motion patterns stored in the database are normalized, and wherein, to select the at least one digital motion pattern segment, a scaling factor corresponding to an object size is applied to the position data in order to obtain a normalized motion of the object.

8. The method of claim 1, wherein the at least one digital motion pattern segment is normalized and wherein a scaling factor corresponding to an object size is applied to the same for reconstructing the image in order to reconstruct a denormalized motion course of the object with the at least one sampling point.

9. The method of claim 1, wherein, up to a first instant after receiving the position data, the at least one digital motion pattern segment is selected from the plurality of predefined digital motion patterns which, at this first instant, corresponds with the highest probability to the motion of the object, and wherein, after the first instant up to a second instant, search for a better-matching digital motion pattern segment for the time interval is continued.

10. The method of claim 1, wherein different priorities are assigned to different sampling points of the object, and position data of higher-priority sampling points is evaluated earlier in time than position data of lower-priority sampling points.

11. The method of claim 1, wherein a higher priority is assigned to a larger object than to a comparatively smaller object, and wherein position data of higher-priority objects is evaluated earlier in time than position data of lower-priority objects.

12. The method of claim 1, further comprising:

transmitting, by a radio transmitter coupled to the position marker, the position data to an evaluation device configured to perform the method for reconstructing the motion of the object.

13. The method of claim 1, wherein the position data is obtained from position markers of a real-time locating system and the method is performed under real-time conditions.

14. A computer program product comprising a non-transitory computer-readable medium having a computer readable program code embodied therein, the computer readable program code adapted to be executed to implement a method for reconstructing a motion of an object from a sequence of motion pattern segments of a 3D computer model of the object, wherein each motion pattern segment corresponds to a different time interval of the motion and includes at least one frame, and wherein the object comprises at least one sampling point coupled to a position marker, the method comprising:

a) capturing a sequence of motion transitions between an initial motion state and a final motion state of the object in a time interval of the motion based on position data of the at least one sampling point which is received from the position marker;

b) selecting, based on possible motion states of a Markov model, at least one digital motion pattern segment corresponding to the sequence of motion transitions from a plurality of predefined digital motion patterns of the 3D computer model which are stored in a database and correspond to possible real motions of the object such that the at least one digital motion pattern segment leads with sufficient probability from the initial motion state to the final motion state for the time interval; and

c) reconstructing an image of the motion of the object for the time interval by displaying the 3D computer model on a display using the initial motion state and the at least one digital motion pattern segment, wherein

by selecting the at least one digital motion pattern segment, the most probable sequence of the possible motion states of the Markov model and the sequence of motion transitions is determined in order to reconstruct the motion of the object.

15. A device for reconstructing a motion of an object from a sequence of motion pattern segments of a 3D computer model of the object, wherein each motion pattern segment corresponds to a different time interval of the motion and includes at least one frame, and wherein the object comprises at least one sampling point coupled to a position marker, the device comprising:

one or more processors configured to execute computer-readable instructions to

capture a sequence of motion transitions between an initial motion state and a final motion state of the object in a time interval of the motion based on position data of the at least one sampling point which is received by the position marker;

select, based on possible motion states of a Markov model, at least one digital motion pattern segment corresponding to the sequence of motion transitions from a plurality of predefined digital motion patterns of the 3D computer model which are stored in a database and correspond to possible real motions of the object, wherein the at least one digital motion pattern segment leads with sufficient probability from the initial motion state to the final motion state for the time interval; and

reconstruct an image of the motion of the object for the time interval by displaying the 3D computer model on a display using the initial motion state and the at least one digital motion pattern segment, wherein

by selecting the at least one digital motion pattern segment, the most probable sequence of the possible motion states of the Markov model and the sequence of motion transitions is determined in order to reconstruct the motion of the object.