Search

CN-121998949-A - Automatic docking method and system based on monocular vision

CN121998949ACN 121998949 ACN121998949 ACN 121998949ACN-121998949-A

Abstract

The invention discloses an automatic docking method and system based on monocular vision, wherein the method comprises the following steps: the method comprises the steps of constructing an abnormal concentric double-ring oil port model, extracting an effective arc section by adopting an anti-interference elliptical detection scheme combining software and hardware, constructing a pose solving model based on a double-ring geometric structure, and obtaining the oil port pose through multi-constraint optimization. The system comprises a modeling unit, an image acquisition unit, a pose calculation unit and a control unit. The invention does not need to rely on manual control points, has strong anti-interference capability, provides reliable technical support for accurate butt joint of the automatic oil filling port of the automobile, and has extremely high engineering application value. The invention can be widely applied to the field of machine vision.

Inventors

  • ZHANG XIAOHU
  • YANG WENYI
  • WANG WEI
  • ZHANG TONG

Assignees

  • 中山大学

Dates

Publication Date
20260508
Application Date
20260128

Claims (10)

  1. 1. An automatic docking method based on monocular vision is characterized by comprising the following steps: Modeling the oil filler to construct a different-surface concentric double-ring oil filler model; Acquiring an oil filler image and filtering based on brightness information to obtain a preprocessed image; Performing edge chain extraction and arc segment segmentation on the preprocessed image to obtain a candidate arc segment set; recombining the candidate arc segment sets by adopting a multi-constraint arc segment aggregation strategy to obtain recombined arc segment sets; Screening candidate ellipses based on the recombined arc segment set, and calculating to obtain projection ellipse parameters; Solving the initial pose of the space circle based on the projected ellipse parameters, and selecting an optimal initial solution through a geometric structure error function; constructing an objective function and combining the optimal initial solution to perform pose optimization to obtain the final pose of the oil port; And guiding the cooperative arm to drive the butt joint to finish butt joint based on the final pose of the oil port.
  2. 2. The method for automatically docking according to claim 1, wherein the step of obtaining the image of the fuel filler and filtering based on the brightness information to obtain the preprocessed image comprises: Introducing an optical filter and collecting an oil filler image; and calculating the Sobel gradient matrix of the oil filler image and filtering by combining brightness information to obtain a preprocessed image.
  3. 3. The automatic docking method based on monocular vision according to claim 1, wherein the step of performing edge chain extraction and arc segment segmentation on the preprocessed image to obtain a candidate arc segment set specifically comprises the following steps: performing edge chain extraction on the preprocessed image through an edge detection algorithm to obtain an initial edge chain set; and sampling dominant points, calculating and judging inflection points based on vector included angles, and performing arc segment segmentation on the initial edge chain set to obtain a candidate arc segment set.
  4. 4. The automatic docking method based on monocular vision according to claim 1, wherein the multi-constraint arc segment aggregation strategy specifically comprises coarse screening and fine geometry constraints, wherein: The coarse screening comprises distance constraint and direction consistency constraint; the judging condition of the distance constraint is that the Euclidean distance of the midpoints of the two arc sections is not more than 3 times of the sum of the two arc lengths; The fine geometry screening comprises local angle constraint, global direction consistency verification and a secondary correction mechanism; the global direction consistency constraint is determined by cross product results of arc segment key vectors.
  5. 5. The automatic docking method based on monocular vision according to claim 1, wherein the step of screening candidate ellipses based on the recombined arc segment set and calculating to obtain projection ellipse parameters specifically comprises the following steps: screening and optimizing ellipses in the recombined arc segment set based on geometric features and gradient information to obtain candidate ellipses; And calculating the gradient confidence score of the candidate ellipse, and separating the inner ring ellipse and the outer ring ellipse through a two-layer clustering strategy to obtain four projection ellipse parameters corresponding to the double rings.
  6. 6. The method for automatically interfacing based on monocular vision according to claim 5, wherein the gradient confidence score is obtained by calculating a cosine similarity mean of a local gradient direction of the sampling points of the elliptical contour and a tangential direction of the ellipse.
  7. 7. The monocular vision-based automatic docking method of claim 5, wherein the two-layer clustering strategy comprises: the coarse clusters are grouped by the difference constraint of the center distance and the axial length; and the fine clustering is used for screening the optimal ellipse pairs by taking the principle that the center concentricity is highest and the axial ratio difference is smallest.
  8. 8. The monocular vision-based automatic docking method of claim 1, wherein the geometric error function is formulated as follows: Wherein, the Represents the center coordinates of the inner side of the large circle, Represents the center coordinates of the outer edge of the large ring, Represents the center coordinates of the inner side of the small circle, Represents the center coordinates of the outer edge of the small circular ring, 、 、 、 Respectively corresponding normal vectors.
  9. 9. The monocular vision-based automatic docking method of claim 8, wherein the objective function is expressed as follows: Wherein, the Representing concentric circle center consistency constraint terms, Represents a normal vector consistency constraint term, Represents a normal vector unitized constraint term, Represents a constraint item of colinear circle centers, Representing the term of the axial distance constraint, The corresponding weight coefficient is represented by a set of weights, Representing the distance between the center points of two concentric rings in space.
  10. 10. An automatic docking system based on monocular vision, comprising: the modeling unit is used for modeling the oil filler to construct a different-surface concentric double-ring oil port model; the image acquisition unit is used for acquiring the oil filler image and filtering the oil filler image based on the brightness information to obtain a preprocessed image; The system comprises a pose computing unit, a multi-constraint arc segment aggregation strategy, a projection ellipse parameter calculation unit, a target function construction unit, a pose optimization unit, a hydraulic port final pose calculation unit, a geometric structure error function calculation unit and a hydraulic port final pose calculation unit, wherein the pose computing unit is used for carrying out edge chain extraction and arc segment segmentation on the preprocessed image to obtain a candidate arc segment set; and the control unit is used for guiding the cooperative arm to drive the butt joint to finish butt joint based on the final pose of the oil port.

Description

Automatic docking method and system based on monocular vision Technical Field The invention relates to the field of machine vision, in particular to an automatic butt joint method and system based on monocular vision. Background At present, a gas station mainly adopts a worker to manually refuel a fuel vehicle, and because of factors such as time for the worker to receive training, theoretical knowledge mastering conditions and the like, uncertainty exists in operation of the refueler, and certain security threat can be possibly brought. In addition, manual oiling requires a great deal of manpower and material resources. Because of the rapid development of machine vision, the application prospect of automatic oiling of a gas station is very wide, in an automatic oiling system of an automobile, accurate automatic butt joint of an oiling port and a butt joint is a core link, and accurate perception of the spatial pose of the oil port is a precondition for realizing reliable butt joint. Most of the existing automobile oil filling ports are of circular symmetrical structures, high-reflectivity materials are adopted for manufacturing, in an actual oil filling scene, oil port surface reflection and equipment optical component reflection are overlapped, imaging edge breakage and stray edges are increased, and feature extraction accuracy is seriously affected. Therefore, the monocular vision pose estimation method which does not need to depend on manual control points, is high in anti-reflection interference capability and pose estimation accuracy is developed, and has important significance for promoting engineering application of an automobile automatic oiling technology. Disclosure of Invention In view of the above, in order to solve the technical problem that the oil gun docking accuracy is not high due to imaging reflection interference in the existing automatic oiling scheme, in a first aspect, the invention provides an automatic docking method based on monocular vision, which comprises the following steps: The method comprises the steps of establishing a geometric model, firstly establishing a double-circular-ring structure with different concentric surfaces as space geometric expression of a fuel filler, preprocessing an image, acquiring a fuel filler image, carrying out filtering and enhancement according to brightness characteristics to obtain a clear characteristic image, extracting edges and arc segments, extracting edge chains from the preprocessed image, dividing the arc segments to form a candidate arc segment set, recombining the arc segments, screening and polymerizing the candidate arc segments through multiple constraint conditions to generate a structured arc segment set, carrying out ellipse fitting and parameter calculation, carrying out ellipse fitting based on the recombined arc segments, calculating geometric parameters of the ellipse fitting and parameter calculation, carrying out initial pose estimation, carrying out inverse pushing on a possible pose of a space circle according to the ellipse parameters, combining geometric structure error evaluation, selecting an optimal initial solution, carrying out pose optimization, carrying out iterative optimization with the initial solution as a starting point, determining the accurate space pose of the fuel filler, and carrying out butt joint, and guiding a cooperative mechanical arm to adjust the butt joint position according to the final pose, so as to finish accurate butt joint operation. The invention also provides an automatic docking system based on monocular vision, which comprises a modeling unit, an image acquisition unit, a pose calculation unit and a control unit. Based on the scheme, the invention provides an automatic butt joint method and system based on monocular vision, provides a different-surface concentric double-ring oil port model, eliminates pose ambiguity by utilizing natural geometric constraint, does not need to rely on manual control points, can be used for migration adaptation of different types of oil ports, further suppresses interference of ambient light by a near infrared light source and a narrow-band filter, combines arc section self-adaptive segmentation and multi-constraint aggregation strategies, solves the problem of edge fracture caused by reflection, and designs a direct resolving model without control points, solves pose based on the corresponding relation between two-dimensional elliptic projection and three-dimensional geometric structure of double rings, and has high precision and strong physical interpretability. Drawings FIG. 1 is a flow chart of the steps of an automatic docking method based on monocular vision according to the present invention; FIG. 2 is a graph comparing position error broken lines of the method of the present invention with those of an unoptimized bit circular method; FIG. 3 is a position error bin graph based on the data statistics of FIG. 2; FIG. 4 is a graph comparing angular