CN-122023719-A - Position sensing mark facing deformed surface and reconstruction method thereof

CN122023719ACN 122023719 ACN122023719 ACN 122023719ACN-122023719-A

Abstract

The invention provides a position sensing mark facing a deformed surface and a reconstruction method thereof, wherein the mark is formed by arranging a plurality of positioning round points in a honeycomb shape, a hexagon is used as a self-identification unit, the positioning round points are selected from a stable color set calibrated by multiple brightness levels, unique IDs are generated through the color characteristics of each positioning round point and a preset ID reading sequence, and the reconstruction method comprises the steps of image acquisition, characteristic detection, grid reconstruction, ID identification, three-dimensional reconstruction and experimental verification, and high-precision reconstruction is realized by combining binocular vision and interpolation correction technology. The invention breaks through the dependence of the traditional matrix coding on the rigid surface, generates a marking field through graph theory modeling, improves the brightness interference resistance through color calibration, ensures that the coding value matching process is not influenced by adjacent dots, is suitable for the fields of non-planar object positioning, clothing dynamic capturing, AR and the like, and has the advantages of stable identification, high reconstruction precision and strong robustness.

Inventors

HE BINGWEI
LIN JIAN
ZHU MINGZHU
GU XIANRUI

Assignees

福州大学

Dates

Publication Date: 20260512
Application Date: 20260130

Claims (10)

1. The position sensing mark facing the deformed surface is characterized in that the mark is formed by arranging a plurality of positioning round points in a honeycomb shape, and a hexagon formed by surrounding one central positioning round point by six positioning round points is taken as a self-identification unit; the color of the positioning dot is selected from a stable color set obtained through color calibration, visual differences which can be stably distinguished by a machine vision system are arranged between each reference color and background color in the stable color set, and the reference colors meet the constraint condition of furthest distance in an RGB space; each self-identification unit generates a unique self-identification unit ID through a preset ID reading sequence and the color characteristics of each positioning dot in the unit.
2. The method for identifying the position of the deformed surface according to claim 1, wherein the color features are inherent coding marks of reference colors corresponding to the positioning dots in a stable color set, each reference color in the stable color set corresponds to a unique coding value, after the positioning dots are matched with the target reference color, the coding values of the target reference color are assigned to the positioning dots, and finally, the self-identification unit ID is generated through combination of the coding values of the positioning dots in the self-identification unit and a preset ID reading sequence.
3. The deformed surface-facing position-aware marking of claim 1, wherein the stable color set is obtained by: sampling in each channel of RGB color space to generate an initial color plate which takes background color as base color and contains multiple color samples; Mapping the collected actual RGB values of the color sample to a three-dimensional square reference color grid, screening out the color sample which can be stably mapped to the same reference point under various brightness levels and marking the color sample as the color category represented by the reference point; Calculating the point cloud density of each type of color sample in the RGB space, and extracting the color sample with the maximum density in each type as a color candidate; And simulating the performance of each candidate color combination under various brightness levels by a multi-objective optimization algorithm, calculating a comprehensive stability score according to a predefined illumination weight, and selecting a global optimal color combination meeting the color differentiation and stability requirements from the color candidates to form the stable color set.
4. The deformed surface-facing position sensing mark of claim 1, wherein the predetermined ID reading sequence is to sequentially read an upper left positioning dot, an upper right positioning dot, a lower left positioning dot, a left positioning dot, and finally read the center dot for any one of the self-recognition units, sequentially extract the encoded values of the positioning dots, and combine them to form the self-recognition unit ID.
5. A method of reconstruction of a deformable surface based on position-aware marking according to any of claims 1-4, comprising the steps of: S1, acquiring a real-time image of a deformed surface attached with the position sensing mark through a binocular camera; s2, performing feature detection on the acquired image, identifying all positioning dots and acquiring centroid coordinates of the positioning dots; s3, carrying out gridding treatment based on the centroid coordinates, screening and reconstructing grids formed by a plurality of hexagonal self-identification units; S4, identifying the ID of each hexagonal self-identification unit, and acquiring the three-dimensional coordinates of each positioning dot based on a binocular vision triangulation principle; and S5, based on the three-dimensional coordinates, realizing three-dimensional reconstruction of the deformed surface through surface fitting and interpolation correction.
6. The method for reconstructing a deformable surface according to claim 5, wherein said S2 specifically comprises: s21, performing blurring processing on the image by adopting a low-pass filter so as to inhibit surface texture interference; S22, converting the blurred image into a gray level image, and acquiring a binary image through a self-adaptive threshold segmentation algorithm; s23, performing morphological open operation on the binary image, and calculating the mass center of the connected region to obtain the mass center coordinates of the positioning dots.
7. The method for reconstructing a deformable surface according to claim 5, wherein said S3 comprises: S31, triangulating the centroid coordinates to obtain an initial grid formed by triangles; S32, screening grid cells conforming to a hexagonal structure from the initial grid; And S33, integrating the screened hexagonal grid units to finish grid reconstruction.
8. The method for reconstructing a deformed surface according to claim 5, wherein the specific step of identifying each hexagonal self-identification unit ID in S4 is to match, for each hexagonal self-identification unit, an actual RGB value of a positioning dot with an RGB value of each reference color in a stable color set to determine a coded value of the positioning dot, extract the coded value of each positioning dot based on the preset ID reading sequence, and combine to generate an initial ID candidate result; And if the initial ID candidate result has matching ambiguity, determining the correct code value of the positioning dot by adopting a voting mechanism based on the condition of the code values matched by the 7 self-identification units associated with the positioning dot, so as to correct and determine the optimal ID of the hexagonal self-identification unit.
9. The method of reconstructing a deformable surface according to claim 5, wherein said S1 position-aware markers are prepared from marker fields generated by a graph-theory model, said generating of marker fields comprising the steps of: Constructing a honeycomb topology network of the target deformation surface based on a three-dimensional model of the target deformation surface, and defining a simple embedding diagram F= (V, E, C), wherein a vertex set V comprises theoretical three-dimensional coordinates of positioning dots and label information to be distributed, an edge set E describes a space adjacent relation between the vertexes, a surface set C defines a local geometric structure through the surrounding sequence of the vertexes, and a basic unit of the surface set C is a hexagon and forms a self-identification unit; Drawing subgraphs isomorphic with the hexagonal self-recognition unit from the graph F based on the VF3 graph isomorphic matching principle as candidates, and filtering non-coplanar candidate subgraphs in the candidate subgraphs based on the surface normal direction consistency; and allocating color coding labels to each vertex in the vertex set V through a risk minimization strategy, and combining a backtracking mechanism to solve the conflict, so as to generate the marking field.
10. The method of reconstruction of a deformable surface according to claim 9, wherein each vertex in the set V of vertices is assigned a color coded label by a risk minimization strategy, in particular as follows: And preferentially distributing labels to the vertexes with the largest quantity of the associated candidate subgraphs, and selecting the optimal labels according to a label safety function S (Li) = ΣlogS (Ar), wherein S (Ar) represents the success probability that the (r) th hexagonal unit associated with the current vertex can finally form a global unique ID under the condition that the labels Li are distributed to the current vertex.

Description

Position sensing mark facing deformed surface and reconstruction method thereof Technical Field The invention belongs to the technical field of computer vision and three-dimensional reconstruction, and particularly relates to a position sensing mark facing a deformed surface and a reconstruction method thereof. Background The reconstruction of deformed surface is a core technology in the fields of computer vision and robots, and accurate tracking of surface points is required to be realized when the surface is deformed such as bending, wrinkling, folding, stretching and the like. The technology is widely applied to scenes such as non-planar object positioning, clothing dynamic capturing, augmented reality, visual touch technology and the like. The conventional deformation surface reconstruction technology mainly comprises three types of piezoelectric films, structural lights and visual positioning marks, wherein the three types of the conventional deformation surface reconstruction technology mainly have obvious application limitations, the piezoelectric films depend on sensors deployed at edges, the reconstruction precision of the micro deformation of a central area is insufficient, the sensors are easy to fatigue and damage under frequent or extreme deformation, the structural lights are constrained by illumination conditions and scene textures, non-rigid deformation is difficult to effectively process, the visual positioning mark method has the advantages of low cost and quick response, but the conventional mark has obvious defects that firstly, the deformation resistance is weak, the self-identification mark and the position perception mark are easy to cause boundary loss due to physical deformation in practical application, and the problems are further aggravated due to factors such as low mark density, oversized self-identification unit, and the like, so that identification failure is caused, and secondly, the color recognition accuracy is disturbed due to brightness change, printer characteristics, paper textures, camera sensitivity and the like. In particular, the problem of visual positioning marks at the present stage is concentrated on two major core scenes of boundary loss and brightness sensitivity, namely, the position sensing marks can cover more than several times of characteristic points of the self-identification marks under the same area by virtue of higher positioning characteristic density for boundary loss, the influence of boundary loss is greatly relieved, but flexible surface deformation is easy to cause distortion of mark shapes and displacement of the characteristic points, the complex curved surface also has the problems of shielding, perspective deformation and the like, and partial mark detection failure still can be caused, so that the design and detection of the position sensing marks become research emphasis, and the color distortion caused by the combination of the position sensing marks, hardware characteristics and ambient light can seriously interfere the accurate identification of mark colors for brightness sensitivity, so that the application range of the traditional visual positioning marks is further limited. In view of the above, the invention designs a position sensing mark facing to a deformed surface in order to solve the problems that the existing visual mark has weak deformation resistance and poor brightness sensitivity. The triangular mark is generated based on a stable color set calibrated by multiple brightness levels, has stronger brightness sensitivity resistance and deformation resistance, and provides powerful technical support for deformed surface tracking. Disclosure of Invention Aiming at the problems of boundary loss, brightness sensitivity, poor coding adaptability and the like in the application of the existing visual positioning mark on the deformed surface, the invention provides a position sensing mark facing the deformed surface and a reconstruction method thereof, wherein the coding is generated by matching the color characteristics of the positioning dots with a stable color set, and ID is combined according to a specific sequence, so that the high-precision and high-robustness surface reconstruction under a complex deformed scene is realized. In order to achieve the above purpose, the technical scheme of the invention is as follows: The position sensing mark facing the deformed surface is formed by arranging a plurality of positioning round points in a honeycomb shape, and takes a hexagon formed by surrounding one central positioning round point by six positioning round points as a self-identification unit; the color of the positioning dot is selected from a stable color set obtained through color calibration, visual differences which can be stably distinguished by a machine vision system are arranged between each reference color and background color in the stable color set, and the reference colors meet the constraint condition of furthest