Granted Patent B2

US 12,573,057 · App. 18/206,216 · Granted Mar 10, 2026

Inventor: Yoshihisa Asayama (Kawasaki, JP)

Assignee: FUJITSU LIMITED

G06T7/246G06T7/521G06T2207/10028G06T2207/30196G06T2207/30221

Abstract

Claims (54)

1 . A skeleton estimation device comprising:

memory; and

processor circuitry coupled to the memory, the processor circuitry being configured to perform processing including:

acquiring estimation information in which a change in a skeleton in a predetermined period is estimated in units of frames;

calculating a feature amount of the change in the skeleton in the predetermined period;

calculating an approximate line based on the calculated feature amount;

calculating a correlation state between the calculated feature amount and the calculated approximate line;

calculating a likelihood based on the feature amount in the predetermined period; and

detecting, as a defective skeleton, the frame portion in which estimation is abnormal in the estimation information in the predetermined period based on the correlation state, and detecting the defective skeleton based on the likelihood.

2 . The skeleton estimation device according to claim 1 , the processing further comprising:

calculating an approximate curve by a multidimensional function from the feature amount in the predetermined period in a case where the feature amount exceeds a preset range;

calculating the likelihood of the feature amount in the predetermined period from a correlation coefficient between the calculated feature amount and the calculated approximate curve; and

detecting, as the defective skeleton, the frame with a maximum difference between the feature amount and the approximate curve in a case where the likelihood is smaller than a preset threshold.

3 . The skeleton estimation device according to claim 1 , the processing further comprising:

calculating each of the approximate curves by a multidimensional function from the feature amount for each divided section obtained by dividing the predetermined period into a plurality of divided sections in a case where the feature amount falls within the preset range;

calculating the likelihood from a correlation coefficient between the feature amount calculated in each of the divided sections and the calculated approximate curve; and

detecting, as the defective skeleton, the frame with a maximum difference between the feature amount and the approximate curve in a case where the likelihood is smaller than the preset threshold.

4 . The skeleton estimation device according to claim 3 , the processing further comprising:

performing the division such that some of the plurality of divided sections temporarily overlaps each other.

5 . The skeleton estimation device according to claim 1 , the processing further comprising:

calculating a maximum value and a minimum value of the feature amount in the predetermined period in a case where the feature amount is less than the preset range;

calculating the likelihood based on a cumulative amount of the feature amounts that includes the calculated maximum value and the calculated minimum value; and

detecting the defective skeleton based on the likelihood.

6 . The skeleton estimation device according to claim 1 , the processing further comprising:

estimates a skeleton again for the detected frame portion of the defective skeleton.

7 . The skeleton estimation device according to claim 1 , wherein the feature amount is a motion feature amount that corresponds to a predetermined motion of a human body assumed in advance in the predetermined period.

8 . A skeleton estimation method implemented by a computer, the skeleton estimation method comprising:

acquiring estimation information in which a change in a skeleton in a predetermined period is estimated in units of frames;

calculating a feature amount of the change in the skeleton in the predetermined period;

calculating an approximate line based on the calculated feature amount;

calculating a correlation state between the calculated feature amount and the calculated approximate line;

calculating a likelihood based on the feature amount in the predetermined period; and

detecting, as a defective skeleton, the frame portion in which estimation is abnormal in the estimation information in the predetermined period based on the correlation state, and detecting the defective skeleton based on the likelihood.

9 . A gymnastics scoring support system of a gymnastics competition in which a gymnast executes a technique, the gymnastics scoring support system comprising first processor circuitry configured to perform processing comprising:

acquiring estimation information in which a change in a skeleton of the gymnast over a predetermined period is estimated in units of frames;

calculating a feature amount of the change in the skeleton in the predetermined period;

calculating an approximate line based on the calculated feature amount;

calculating a correlation state between the calculated feature amount and the calculated approximate line;

calculating a likelihood based on the feature amount in the predetermined period; and

detecting, as a defective skeleton, the frame portion in which estimation is abnormal in the estimation information in the predetermined period based on the correlation state, and detecting the defective skeleton based on the likelihood.

10 . The gymnastics scoring support system according to claim 9 , the system further comprising:

a 3D laser sensor configured to capture the gymnast and output an image of a three-dimensional point cloud that includes distance information in units of frames; and

a skeleton estimation device, wherein

the skeleton estimation device includes second processor circuitry configured to perform processing including:

generating a depth image from the image of the three-dimensional point cloud and recognizes a three-dimensional skeleton of the gymnast based on a training model trained in advance from the depth image; and

outputting a three-dimensional skeleton adapted to the skeleton of the gymnast as the estimation information based on the three-dimensional point cloud and the three-dimensional skeleton, and

the first processor circuitry is further configured to:

obtain the feature amount set by the technique in advance; and

detect the defective skeleton in the estimation information output by the fitting unit, corresponding to execution of a predetermined technique by the gymnast.

11 . The gymnastics scoring support system according to claim 10 , wherein the first processor circuitry is configured to:

present the estimation information to a scoring person of the gymnastics competition; and

present the estimation information again after the skeleton is estimated for the frame portion of the defective skeleton at the time when the defective skeleton is detected.

12 . The gymnastics scoring support system according to claim 11 , wherein the feature amount is a motion feature amount different by technique executed by the gymnast.

13 . The gymnastics scoring support system according to claim 12 , wherein the technique includes a twist and a salto.

Assignments (1)

Continuity (2)

References Cited (19)