Optical motion capture data processing method based on module piecewise linear model
Optical motion capture data processing method based on module piecewise linear model
 CN 101,533,528 B
 Filed: 04/18/2009
 Issued: 11/26/2014
 Est. Priority Date: 04/18/2009
 Status: Active Grant
First Claim
1. the optical motion capture data processing method based on module piecewise linear model, is characterized in that comprising the steps:
 (1) set up human geometry'"'"'s model and establish human body gauge point disposing way, importing manikin and parameter;
(2) according to close on topological structure and on physiological structure relevant principle, combine contacting bone closely, according to human physiological structure, human skeleton is split as to seven large modules, described module comprises head, waist, huckle, chest, foot, shoulder and hand module, to set up human physiological structure'"'"'s module, import human physiological structure'"'"'s module parameter;
(3) constructing module denoise algorithm;
scattered data being is carried out to noise preservice, rationally pick out noise data, ensure that remainder strong point number is 3034;
The scheme that described denoise algorithm is concrete is;
in the time there is noise spot, be gauge point to be calibrated by selecting the point that can more mate with template, and other point deletion;
For fear of the gauge point of rejecting other modules, in the time deleting other noise spot, see whether this module intersects with hand module, as intersected to see whether those noise spots are ingredients of hand module, if not just deleting;
(4) set up module piecewise linear model;
utilize the statistical information of noise data, dynamically exercise data is described, obtain Statistics table, following the tracks of to process for the Data Matching in later stage provides effective constraint condition;
Specific implementation step is;
(4.1) taking hand module as example, hand module is substantially fixing quadrilateral, does not become constraint condition carry out form fit according to the tetragonal length of side and diagonal line length in motion process, and module such in each frame is all found, as do not disregarded, explanation is due to shortcoming;
Its four point coordinate is averaged, carry out record with a twodimensional array, transverse axis represents frame t, the longitudinal axis represent respectively present frame with respect to former frame the sideplay amount in x, y, z direction;
(4.2) taking 5 continuous frames as a time period, first check in these 5 frame data and whether have and obviously depart from larger frame, if any the phenomenon that mismatches of thinking in coupling, remove it;
Then obtain the mean change displacement of residue frame, try to achieve successively and be stored in array a, transverse axis represents the numbering of which section, and the longitudinal axis represents that the vector sum of pace of change, first frame and last frame changed angle with respect to the direction of a upper time period;
(4.3) from first section, compare successively the record of adjacent segment, merge as similar, otherwise bidding will bit representation is a trend section, similar condition is;
1. pace of change differs and is less than certain value, and 2. direction of motion is level and smooth, and angle is less than 90 degree;
(4.4) as two sections can merge, amendment array a, the transverse axis of currentitem is constant, and the speed that the longitudinal axis changes changes the mean value of two sections into, and first frame is constant, and last frame changes the last frame value of latter a section into, unites two into one due to current two, and therefore array rear portion moves forward successively;
Until inquire array end, finish Fusion query;
Step (4.5) is carried out the processing of above (4.14.4) step to each module, and final statistics is described with a bivariate table, for use in prediction and the tracking of exercise data;
(5) prediction of the exercise data based on module piecewise linear model and track algorithm;
Specific implementation step comprises;
(5.1) calling module denoise algorithm, carries out the noise reduction process before exercise data tracking;
(5.2) the segmentation statistical information of the each module of foundation, sort from low to high by the amplitude of variation of each module, for each frame data, start to carry out module coupling from the low module of amplitude of variation, matching process is to utilize geometric properties, as the match is successful, exercise data is carried out to mark, and from raw data, the exercise data of mark is deleted, as it fails to match, forward step (5.3) to;
(5.3) as not having, the match is successful, utilize the information recording data of current feature array segment to carry out Newton interpolation, the result that interpolating function obtains is the centre coordinate of module, be designated as A, being reduced into each gauge point seat calibration method from centre coordinate is;
the centre coordinate that previous frame is obtained to this module of mating is by moving to A, translation matrix is B, the point coordinate obtaining after the phorogenesis of each gauge point in this module of previous frame by matrix B is the required threedimensional coordinate when the each gauge point of front module, if the information recording data in this array segment are very few, the future position that utilizes Newton'"'"'s interpolation to obtain is larger by error, should assist constraint condition to carry out predicting tracing,(5.4) after each module is predicted and is followed the tracks of, carry out structure verification;
structure verification is that the result of coupling or tracking is tested, divide the overall situation and partial check, when overall situation inspection, template counter structure and joint are carried out distance and are detected, and meet thinking of distance error scope and follow the tracks of correctly, otherwise think trailanderror, when partial check, carry out the inspection of health left and right sides;
(6) exercise data after every frame tracking and matching is carried out to structure verification, obtain final nominal data;
(7) exercise data of success being demarcated is according to the form data writing file reading in.
Chinese PRB Reexamination
Abstract
The invention discloses an optical motion capture data processing method based on a module piecewise linear model. Based on overall information of optical human motion capture scattered data, a data processing algorithm based on the module piecewise linear model is provided. By using the module piecewise linear model to generalize the change characteristics of different modules, the method determines a matching priority and an intrasegment fitting function of module data, effectively performs overall layered prediction and tracking for each threedimensional motion data module, performs denoising processing for the noise data based on the module, and provides an interpolation fitting algorithm for the missing motion data based on segmenting Newton to perform reasonable supply. The optimized method does not need manual intervention during processing, and can meet realtime requirement.
2 Claims

1. the optical motion capture data processing method based on module piecewise linear model, is characterized in that comprising the steps:

(1) set up human geometry'"'"'s model and establish human body gauge point disposing way, importing manikin and parameter; (2) according to close on topological structure and on physiological structure relevant principle, combine contacting bone closely, according to human physiological structure, human skeleton is split as to seven large modules, described module comprises head, waist, huckle, chest, foot, shoulder and hand module, to set up human physiological structure'"'"'s module, import human physiological structure'"'"'s module parameter; (3) constructing module denoise algorithm;
scattered data being is carried out to noise preservice, rationally pick out noise data, ensure that remainder strong point number is 3034;The scheme that described denoise algorithm is concrete is;
in the time there is noise spot, be gauge point to be calibrated by selecting the point that can more mate with template, and other point deletion;
For fear of the gauge point of rejecting other modules, in the time deleting other noise spot, see whether this module intersects with hand module, as intersected to see whether those noise spots are ingredients of hand module, if not just deleting;(4) set up module piecewise linear model;
utilize the statistical information of noise data, dynamically exercise data is described, obtain Statistics table, following the tracks of to process for the Data Matching in later stage provides effective constraint condition;
Specific implementation step is;(4.1) taking hand module as example, hand module is substantially fixing quadrilateral, does not become constraint condition carry out form fit according to the tetragonal length of side and diagonal line length in motion process, and module such in each frame is all found, as do not disregarded, explanation is due to shortcoming;
Its four point coordinate is averaged, carry out record with a twodimensional array, transverse axis represents frame t, the longitudinal axis represent respectively present frame with respect to former frame the sideplay amount in x, y, z direction;(4.2) taking 5 continuous frames as a time period, first check in these 5 frame data and whether have and obviously depart from larger frame, if any the phenomenon that mismatches of thinking in coupling, remove it;
Then obtain the mean change displacement of residue frame, try to achieve successively and be stored in array a, transverse axis represents the numbering of which section, and the longitudinal axis represents that the vector sum of pace of change, first frame and last frame changed angle with respect to the direction of a upper time period;(4.3) from first section, compare successively the record of adjacent segment, merge as similar, otherwise bidding will bit representation is a trend section, similar condition is;
1. pace of change differs and is less than certain value, and 2. direction of motion is level and smooth, and angle is less than 90 degree;(4.4) as two sections can merge, amendment array a, the transverse axis of currentitem is constant, and the speed that the longitudinal axis changes changes the mean value of two sections into, and first frame is constant, and last frame changes the last frame value of latter a section into, unites two into one due to current two, and therefore array rear portion moves forward successively;
Until inquire array end, finish Fusion query;Step (4.5) is carried out the processing of above (4.14.4) step to each module, and final statistics is described with a bivariate table, for use in prediction and the tracking of exercise data; (5) prediction of the exercise data based on module piecewise linear model and track algorithm;
Specific implementation step comprises;(5.1) calling module denoise algorithm, carries out the noise reduction process before exercise data tracking; (5.2) the segmentation statistical information of the each module of foundation, sort from low to high by the amplitude of variation of each module, for each frame data, start to carry out module coupling from the low module of amplitude of variation, matching process is to utilize geometric properties, as the match is successful, exercise data is carried out to mark, and from raw data, the exercise data of mark is deleted, as it fails to match, forward step (5.3) to; (5.3) as not having, the match is successful, utilize the information recording data of current feature array segment to carry out Newton interpolation, the result that interpolating function obtains is the centre coordinate of module, be designated as A, being reduced into each gauge point seat calibration method from centre coordinate is;
the centre coordinate that previous frame is obtained to this module of mating is by moving to A, translation matrix is B, the point coordinate obtaining after the phorogenesis of each gauge point in this module of previous frame by matrix B is the required threedimensional coordinate when the each gauge point of front module, if the information recording data in this array segment are very few, the future position that utilizes Newton'"'"'s interpolation to obtain is larger by error, should assist constraint condition to carry out predicting tracing,(5.4) after each module is predicted and is followed the tracks of, carry out structure verification;
structure verification is that the result of coupling or tracking is tested, divide the overall situation and partial check, when overall situation inspection, template counter structure and joint are carried out distance and are detected, and meet thinking of distance error scope and follow the tracks of correctly, otherwise think trailanderror, when partial check, carry out the inspection of health left and right sides;(6) exercise data after every frame tracking and matching is carried out to structure verification, obtain final nominal data; (7) exercise data of success being demarcated is according to the form data writing file reading in.


2. according to the optical motion capture data processing method based on module piecewise linear model described in claims 1, it is characterized in that the method that in described step (), gauge point is put is:

(1.1) all bones that need seizure of the essential covering of gauge point; (1.2) gauge point need reflect the degree of freedom of bone; (1.3) gauge point is consistent with skeleton motion; (1.4) be placed in the place that is difficult for being blocked; (1.5) gauge point spacing is not of uniform size; (1.6) geometric figure of formation rule between mark of correlation point.

Specification(s)