CN101339604B - Novel mark point graph and its recognition, tracking and positioning algorithm based on visual sense invariance - Google Patents

Abstract

The invention relates to a novel marked point pattern based on vision invariants and identification and tracking positioning algorithm thereof. The invention comprises a design of novel marked point pattern based on vision invariants and identification and tracking positioning algorithm based on the novel marked point. The novel marked point pattern based on vision invariants is a black pentagon with a round mark. The marked point can be a convex pentagon or a concave pentagon. The round mark is used for identifying a first vertex of the pentagon. The pattern is coded using the cross-ratio invariant of the pentagon and combining the concavity and convexity of the vertex. The marked point pattern based on vision invariants can conduct attitude calculation using self-characteristics and realize the coding of the marked point, thus saving the area needed by coding patterns, reducing the pattern size of the marked point and improving the registration precision in unit area. The tracking registration algorithm can be used for identifying the marked point, tracking the head posture of the user and calculating the accurate position and posture of virtual scenes in true environmental space.

Description

A kind of novel monumented point figure and identification, Tracking and Orientation Arithmetic based on the vision invariant

Technical field

The present invention relates to a kind of monumented point design, especially relate to a kind of augmented reality of this distinctive mark point that utilizes and follow the tracks of registration algorithm, belong to the virtual reality field based on the vision invariant.

Background technology

Based on artificial target's tracking registration technology is the most ripe and practical application registration technology the most widely in the current augmented reality system.So-called " monumented point " is meant the special graph pattern of being printed by artificial design.Because the singularity of its pictorial pattern, be difficult to various scenes in environmental background similar, therefore be convenient to from environment, distinguish.Monumented point can obtain the corresponding point set of its 3 d space coordinate and its projected image through the identification location algorithm, calculates the six degree of freedom attitude information of video camera with respect to real scene by the attitude algorithm for estimating.

Important achievement has all been made aspect the registration technology following the tracks of based on the artificial target in external many laboratories.HTL lab design as Washington, DC university has been developed the ARToolKit software package.The square monumented point that uses in this software package is the most frequent adopted monumented point pattern in the current AR system.Because the ARToolKit processing procedure adopts bianry image, ratio of precision is lower, owing to adopted the figure relevant matches aspect codes match, matching efficiency and group/cording quantity all are restricted simultaneously.Japan's Sony computer laboratory was delivered a kind of and monumented point system like the ARToolKit comparing class in 1998, and its encryption algorithm has been carried out certain improvement, made it can design coding more easily.Graduate X.Zhang of Siemens and N.Navab have carried out important improvement to the monumented point system of ARToolKit pattern, have significantly improved its tracking accuracy.The algorithm of this monumented point system has very outstanding character, and is very big for the AR registration technology contribution based on sign.And the people such as Charles B. of U.S. Michigan State University have provided the graphical design method that can obtain maximum group/cording quantity in theory after having analyzed inner pattern coding method.After this, M.Fiala delivered the ARTag monumented point and has followed the tracks of Accreditation System in 2005.All there is very important contribution in this monumented point system for identification, feature location and the two-dimensional bar coding and decoding algorithm of monumented point registration.Except the above square monumented point that is similar to the ARToolKit pattern, also there are some circular monumented point systems to publish at present.Circle marker dot system and Diego L ó pez de that the people such as Leonid Naimark of the wherein more representational InterSense of being company delivered in 2002

Deng the TRIP system that delivered in 2002.

Yet in these monumented point systems, the monumented point design of patterns often is divided into two independent parts, and a part is specifically designed to the coding of contour pattern, and another part is used for the extraction and the Attitude Calculation of unique point.The result of Fen Liing causes the monumented point pattern to need large-area coding region like this, does not utilize the size that reduces the monumented point figure, improves the available precision of unit area.In following the tracks of registration process, need coding is decoded simultaneously, increased the computation burden of computing machine.

Summary of the invention

In order to overcome existing shortcoming of following the tracks of the registration technology aspect based on the augmented reality of monumented point, system of the present invention comprises a kind of novel monumented point figure and identification, Tracking and Orientation Arithmetic based on the vision invariant, can be applicable to augmented reality and follows the tracks of in the Accreditation System.Beneficial effect of the present invention is, novel monumented point based on the vision invariant has made full use of the vision invariant information that the monumented point figure self is had, save the required area of coding pattern, closely reduced the bulk of monumented point figure, improved the available precision of unit area.Accreditation System is followed the tracks of in enhancing based on this vision invariant monumented point, has higher robustness to changing with great visual angle.

The present invention propose a kind of adopt pentagon that black has white circular mark as a token of the augmented reality of dot pattern follow the tracks of Accreditation System.Pentagonal summit is as carrying out the matching characteristic that attitude is estimated; Select to constitute simultaneously this pentagonal five limits calculate its double ratio projective invariant with and concavity and convexity, in order to encode and the distinguishing mark point.For solving the problem that unique point is mated in proper order, in the pentagon figure, add circular mark, be used to discern pentagonal first summit.Discern pentagonal first summit, just determined the order direction on its summit, thereby can determine the matching relationship of unique point.With reference to Fig. 1, be a pentagon that has black or white circle marker based on the novel monumented point figure of vision invariant, the position on this each summit of pentagon can be selected arbitrarily as required as long as guarantee not conllinear of any three summits.And this novel monumented point does not limit pentagonal concavo-convex characteristic, and promptly pentagon can be that protruding pentagon also can be recessed pentagon.Circle marker is used for determining pentagonal first summit.If pentagon be protruding pentagon then circle marker be that white is positioned at pentagon inside, if pentagon be recessed pentagon then circle marker be that black is positioned at the pentagon outside.Determine pentagonal summit order, just can determine the matching relationship of unique point.Utilize pentagonal five limits to calculate its double ratio invariant (I ₁, I ₂) and in conjunction with the concavity and convexity on summit figure is encoded.

$I_1=\frac{\left|M_{431}\right|\left|M_{521}\right|}{\left|M_{421}\right|\left|M_{531}\right|},$ $I_2=\frac{\left|M_{431}\right|\left|M_{532}\right|}{\left|M_{432}\right|\left|M_{521}\right|}---\left(1\right)$

M wherein _Ijk=(l _i, l _j, l _k), (l=(l ¹, l ², l ³) l represents straight-line equation: l ¹X+l ²Y+l ³=0), | M| is a determinant of a matrix

Fig. 2 is the design of circular mark placement location in the monumented point.In order to solve the problem that unique point is mated in proper order, in pentagon, added a circular mark, be used to discern pentagonal first summit.Discern pentagonal first summit, just determined the order (other summit is chosen along clockwise direction) on each summit on it, thereby can determine the matching relationship of unique point.The home position of circular mark is positioned at (x ₀, y ₀).Wherein:

$x_0=\frac{x_1+x_c}{2}---\left(2\right)$

$y_0=\frac{y_1+y_c}{2}$

(x ₁, y ₁) be the coordinate on first summit of pentagon, (x _c, y _c) two the leg-of-mutton barycentric coordinates that summit constituted adjacent with first summit of pentagon of serving as reasons.

Designed identification, the Tracking and Orientation Arithmetic flow process based on the novel monumented point of vision invariant of the present invention is as follows:

(1) all black regions in the extraction institute images acquired;

(2) connected region of the black region that arrives of Detection and Extraction;

(3) pentagon detects;

(4) extract the pentagon edge;

(5) edge thinning obtains single pixel wide edge;

(6) the fitting a straight line algorithm obtains the straight line parameter;

(7) the circular mark in the detection monumented point figure obtains the series arrangement on its summit;

(8) all candidate's pentagonal regions are carried out error-detecting.If the error-detecting result less than preset threshold, will be identified as correct monumented point figure; Do not withdraw from if detect then;

(9) detect pentagonal concavo-convex characteristic;

(10) calculate double ratio invariant by pentagonal limit constituted;

(11) utilize the monumented point characteristic information in convex and concave feature vector and double ratio invariant and the database to compare, determine the coding of monumented point;

(12) repeat (9)-(11) step, all monumented point in detecting present image;

(13) adopt augmented reality registration algorithm to carry out Attitude Calculation based on plane characteristic point;

(14) realize the fusion of true and virtual scene.

According to foregoing description of the present invention, it is big that the present invention can effectively solve certain monumented point figure area occupied, the shortcoming that coding is complicated, and the feature that makes full use of figure self realizes following the tracks of registration based on the augmented reality of this novel monumented point.

Description of drawings

Fig. 1 is the monumented point graphic designs based on the vision invariant among the present invention; The recessed pentagon monumented point of Fig. 1 a.-, the protruding pentagon monumented point of Fig. 1 b.-;

Fig. 2 is the design of circular mark in the monumented point.

Embodiment

Below in conjunction with concrete embodiment the present invention is described in further details.At this, illustrative examples of the present invention and explanation thereof are used to explain the present invention, but not as a limitation of the invention.

In real scene, place the designed pentagon monumented point of the present invention, utilize video camera to take the image information that this scene is obtained scene.The concrete implementing procedure of identification, Tracking and Orientation Arithmetic is as follows:

(1) all black regions in the extraction institute images acquired, this place does not limit and adopts which kind of image binaryzation algorithm;

Adopt the watershed divide adaptive thresholding algorithm to carry out black region in the present embodiment and extract, to guarantee the adaptivity of algorithm illumination variation.

(2) connected region of the black region that arrives of Detection and Extraction, this place does not limit and adopts which kind of connected region detection algorithm;

Adopt label connected region detection algorithm in the present embodiment.

(3) utilize the pentagon detection algorithm to detect pentagon, and extract the pentagon edge, it is refine to single pixel wide;

(4) then adopt the fitting a straight line algorithm to obtain the straight line parameter, this place does not limit and adopts which kind of fitting a straight line algorithm;

Adopted least square method to carry out fitting a straight line in the present embodiment.

(5) utilize white or black circle in the figure to determine pentagonal first summit, and choose other summits on the pentagon in the direction of the clock;

(6) the homography error-detecting is carried out in all candidate regions of choosing.If the error-detecting result less than preset threshold, with being identified as correct match flag point, does not detect then and withdraws from;

(7) detect the concavo-convex characteristic on each summit of monumented point successively, and constitute convex and concave feature vector (a ₁, a ₂, a ₃, a ₄, a ₅), if the concave crown point is expressed as " 1 ", concave vertex is expressed as " 0 ".As the recessed pentagon monumented point of Fig. 1 a pattern, its convex and concave feature vector is (1,0,0,0,0).

(8) calculate double ratio invariant (I by pentagonal limit constituted ₁, I ₂); Utilize double ratio invariant and concavo-convex vector as eigenvector, adopt the monumented point information in mahalanobis distance and the database to compare, as if the monumented point that identifies that praises less than the preset threshold system; If then all monumented points in detecting present image are detected in other candidate region greater than threshold value;

(9) adopt augmented reality registration algorithm to carry out Attitude Calculation, calculate rotation matrix R and the translation matrix T of video camera with respect to the monumented point world coordinate system based on plane characteristic point.

(10) will rotate and translation matrix value assignment is given virtual video camera, and make it, finish virtual and fusion real scene along this orbiting motion.

Above-described specific descriptions; purpose, technical scheme and beneficial effect to invention further describe; institute is understood that; the above only is specific embodiments of the invention; and be not intended to limit the scope of the invention; within the spirit and principles in the present invention all, any modification of being made, be equal to replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (1)

1. one kind based on the novel monumented point figure of vision invariant and identification thereof, Tracking and Orientation Arithmetic, it is characterized in that, novel monumented point figure based on the vision invariant is a pentagon that has black or white circle marker, and the position on this each summit of pentagon can be selected arbitrarily as required as long as guarantee not conllinear of any three summits; And this novel monumented point does not limit pentagonal concavo-convex characteristic, and promptly pentagon can be that protruding pentagon also can be recessed pentagon; Wherein circle marker is used for determining pentagonal first summit; If pentagon be protruding pentagon then circle marker be that white is positioned at pentagon inside, if pentagon be recessed pentagon then circle marker be that black is positioned at the pentagon outside; Determine pentagonal summit order, just can determine the matching relationship of unique point; Utilize pentagonal five limits to calculate its double ratio invariant (I ₁, I ₂) and in conjunction with the concavity and convexity on summit figure is encoded; Identification, Tracking and Orientation Arithmetic flow process based on the novel monumented point of vision invariant are as follows:

(1) all black regions in the extraction institute images acquired;

(2) connected region of the black region that arrives of Detection and Extraction;

(3) pentagon detects;

(4) extract the pentagon edge;

(5) edge thinning obtains single pixel wide edge;

(6) the fitting a straight line algorithm obtains the straight line parameter;

(7) the circular mark in the detection monumented point figure obtains the series arrangement on its summit;

(8) all candidate's pentagonal regions are carried out error-detecting, if the error-detecting result less than preset threshold, will be identified as correct monumented point figure; Do not withdraw from if detect then;

(9) detect pentagonal concavo-convex characteristic;

(10) calculate double ratio invariant by pentagonal limit constituted;

(11) utilize the monumented point characteristic information in convex and concave feature vector and double ratio invariant and the database to compare, determine the coding of monumented point;

(12) repeat (9)-(11) step, all monumented point in detecting present image;

(13) adopt augmented reality registration algorithm to carry out Attitude Calculation based on plane characteristic point;

(14) realize the fusion of true and virtual scene.

Images