DEVICE METHOD AND PROGRAM FOR HUMAN HAND POSTURE ESTIMATION

• US 20110142353A1
• Filed: 04/20/2009
• Published: 06/16/2011
• Est. Priority Date: 06/04/2008
• Status: Active Grant

• Current Assignee: Kiyoshi Hoshino
• Original Assignee: University of Tsukuba National University Corporation
• Inventors: Hoshino, Kiyoshi, Tomida, Motomasa
• Granted Patent: US 9,002,119 B2
• Time in Patent Office: Days
• US Class Current: 382/203
• CPC Class Codes: G06F 18/2431 Multiple classes
• CPC Class Codes: G06V 10/764 using classification, e.g. of video objects
• CPC Class Codes: G06V 40/107 Static hand or arm

Abstract

The objective is to provide a finger shape estimating device that can estimate the image most similar to a finger image quickly and with high precision, and that can facilitate construction of a database. Provided is a finger shape estimating device provided with a matching part that reads second shape data in a specific data set from a database that has multiple data sets in which finger angle data, second shape data relating to dimensions in the vertical direction and the horizontal direction of a second finger image of said finger, and second image feature quantities in the second finger image form a set, to match the second shape data and first shape data related to dimensions in the vertical direction and the horizontal direction in a separately acquired first finger image, and an estimating part that matches the second image feature quantities in the data set comprising the compatible second shape data, from matching by the matching part, with first image feature quantities in the first finger image, and estimates the finger shape in the first finger image.

Claims

1-16. -16. (canceled)

17. A human hand posture estimating device comprising:

a hand image acquiring section that acquires a first hand image;

a shape data calculating section that calculates first shape data relating to vertical and horizontal sizes of the first hand image;

an image characteristic amount extracting section that extracts a first image characteristic amount of the first hand image;

a collation section that reads, from a database holding a plurality of data sets each having a set of angle data of a finger, second shape data relating to vertical and horizontal sizes of a second hand image of the hand, and a second image characteristic amount of the second hand image, the second shape data in a predetermined data set, and collates the second shape data with the first shape data; and

an estimation section that collates the second image characteristic amount in that data set which contains second shape data matched in collation performed by the collation section with the first image characteristic amount to estimate a human hand posture of the first hand image,

wherein the collation section collates the first hand image with the second hand image based on a difference between the first shape data and the second shape data.

18. The human hand posture estimating device according to claim 17, further incorporating the database.

19. The human hand posture estimating device according to claim 17, wherein the image characteristic amount is an image characteristic amount of a segment image of the hand image.

20. The human hand posture estimating device according to claim 17, wherein the image characteristic amount is an image characteristic amount of an image of a predetermined size of pixels converted from the hand image.

21. The human hand posture estimating device according to claim 17, wherein the shape data includes a first shape ratio obtained from a total number of vertical pixels of the hand image and a total number of horizontal pixels thereof.

22. The human hand posture estimating device according to claim 17, wherein the shape data includes a second shape ratio obtained based on a positional relation between a predetermined base pixel in the hand image and a pixel of an outer end of the hand image.

23. The human hand posture estimating device according to claim 17, wherein the angle data of the finger includes joint angle data of the finger and rotational angle data of a wrist.

24. The human hand posture estimating device according to claim 17, wherein the angle data of the finger includes flexion and extension, and adduction and abduction of a wrist.

25. The human hand posture estimating device according to claim 17, wherein the estimation section outputs angle data of a finger in that data set which contains a second image characteristic amount most similar to the first image characteristic amount.

26. The human hand posture estimating device according to claim 17, further comprising a threshold value calculating device to be used at a time the collation section performs collation based on the difference between the first shape data and the second shape data.

27. The human hand posture estimating device according to claim 17, wherein the hand image acquiring section includes:

a tilt calculating section that calculates tilt of a forearm portion in an original image of the first hand image; and

an image correcting section that rotates the first hand image based on the tilt of the forearm portion calculated by the tilt calculating section in such a way that a direction of extension of the forearm portion faces a predetermined direction.

28. The human hand posture estimating device according claim 17, wherein the hand image acquiring section includes:

an outermost contour extracting section that obtains an outermost contour pixel of a finger portion in the hand image in an original image of the hand image;

a base extracting section that obtains a base pixel from an outermost contour pixel of the finger portion with a labeling process; and

a hand image cutout section that determines a range of cutting out the hand image from the original image based on the outermost contour pixel and the base pixel.

31. The human hand posture estimating device according to claim 17, wherein the image characteristic amount is an image characteristic amount of a segment image of an image of a predetermined size of pixels converted from the hand image.

29. A human hand posture estimating method comprising:

a step of acquiring a first hand image;

a step of calculating first shape data relating to vertical and horizontal sizes of the first hand image and a first image characteristic amount of the first hand image;

a step of reading, from a database holding a plurality of data sets each having a set of angle data of a finger, second shape data relating to vertical and horizontal sizes of a second hand image of the hand, and a second image characteristic amount of the second hand image, the second shape data in a predetermined data set;

a step of collating the first hand image with the second hand image based on a difference between the first shape data and the second shape data;

a step of reading the second image characteristic amount in that data set which contains second shape data matched in the collation step; and

a step of collating the first image characteristic amount with the second image characteristic amount to estimate a human hand posture of the first hand image.

30. A program installed in a computer apparatus to allow the computer apparatus to execute predetermined processes, the program allowing the computer apparatus to execute:

a process of acquiring a first hand image;

a process of calculating first shape data relating to vertical and horizontal sizes of the first hand image and a first image characteristic amount of the first hand image;

a process of reading, from a database holding a plurality of data sets each having a set of angle data of a finger, second shape data relating to vertical and horizontal sizes of a second hand image of the hand, and a second image characteristic amount of the second hand image, the second shape data in a predetermined data set;

a process of collating the first hand image with the second hand image based on a difference between the first shape data and the second shape data;

a process of reading the second image characteristic amount in that data set which contains second shape data matched in the collation process; and

a process of collating the first image characteristic amount with the second image characteristic amount to estimate a human hand posture of the first hand image.