Ocean CV Calibration Library

The Ocean CV Calibration Library provides a comprehensive framework for camera calibration using marker-based calibration boards. More...

Collaboration diagram for Ocean CV Calibration Library:

Data Structures
class	Ocean::CV::Calibration::CalibrationBoard
	This class implements a basic calibration board. More...
class	Ocean::CV::Calibration::CalibrationBoardDetector
	This class implements the detector for the calibration board. More...
class	Ocean::CV::Calibration::CalibrationBoardObservation
	This class implements an observation of a calibration board. More...
class	Ocean::CV::Calibration::CalibrationDebugElements
	This class implements debug elements for the calibration library. More...
class	Ocean::CV::Calibration::CameraCalibrator
	This class implements a camera calibrator using a couple of images of a calibration board. More...
class	Ocean::CV::Calibration::Marker
	This class implements the base class for a marker in a calibration board. More...
class	Ocean::CV::Calibration::MarkerCandidate
	This class implements a candidate of a marker. More...
class	Ocean::CV::Calibration::MetricCalibrationBoard
	This class extends the calibration board with metric information. More...
class	Ocean::CV::Calibration::MetricSize
	This class implements an object holding a metric size like width, height, length, or thickness. More...
class	Ocean::CV::Calibration::Point
	This class holds the relevant information of a detected marker point. More...
class	Ocean::CV::Calibration::PointDetector
	This class implements a point detector for marker points. More...
class	Ocean::CV::Calibration::SVGImage
	This class implements functions to write SVG images. More...
class	Ocean::CV::Calibration::Utilities
	This class implements utility functions for calibration. More...

Typedefs
using	Ocean::CV::Calibration::CalibrationBoardObservations = std::vector< CalibrationBoardObservation >
	Definition of a vector holding CalibrationBoardObservation objects.
using	Ocean::CV::Calibration::MarkerCandidates = std::vector< MarkerCandidate >
	Definition of a vector holding marker candidates.
using	Ocean::CV::Calibration::Points = std::vector< Point >
	Definition of a vector holding points.

Detailed Description

The Ocean CV Calibration Library provides a comprehensive framework for camera calibration using marker-based calibration boards.

Overview

Camera calibration is the process of determining the intrinsic and extrinsic parameters of a camera. This library uses a pattern-based approach where a calibration board containing uniquely identifiable markers is used to establish 2D-3D correspondences needed for calibration.

Core Components

1. Marker System

Markers are the fundamental building blocks of the calibration system. Each marker is a 5x5 grid of points (25 points total) that can be either black dots on a white background (positive markers) or white dots on a black background (negative markers). The marker system provides several key features:

• Unique Identification: Each marker has a unique layout pattern that distinguishes it from others
• Orientation Detection: Markers can be detected and identified regardless of their orientation (0°, 90°, 180°, 270°)
• Sign Variants: Each marker can exist in positive (black on white) or negative (white on black) form
• Layout Management: The system manages a collection of unique marker layouts that are rotation-invariant

The marker layout is defined as a 5x5 grid:

 ----------------
|  0  1  2  3  4 |
|  5  6  7  8  9 |
| 10 11 12 13 14 |
| 15 16 17 18 19 |
| 20 21 22 23 24 |
 ----------------

The border points (used for initial detection) are arranged as:

 ----------------
|  0  1  2  3  4 |
| 15           5 |
| 14           6 |
| 13           7 |
| 12  11 10 9  8 |
 ----------------

2. Calibration Boards

Calibration Boards are regular grids of markers arranged in a rectangular pattern. Each board:

• Contains multiple markers arranged in rows and columns
• Has a unique board ID that determines the specific marker arrangement
• Ensures that each marker's 4-neighborhood (adjacent markers) is unique within the board
• Supports different physical sizes through the MetricCalibrationBoard class

3. Point Detection Pipeline

The marker point detection follows a multi-stage process:

Stage 1: Point Detection

The PointDetector uses pattern matching to identify potential marker points in the image:

• Pattern Matching: Uses circular or rectangular patterns of varying radii to detect blob-like structures
• Intensity Analysis: Distinguishes between dark points (black dots) and bright points (white dots)
• Strength Calculation: Computes detection strength based on intensity differences between center and surrounding pixels
• Non-Maximum Suppression: Eliminates duplicate detections in overlapping regions
• Sign Detection: Determines whether each point is positive (black) or negative (white)

Stage 2: Marker Candidate Formation

The MarkerCandidate system groups detected points into potential markers:

• Line Detection: Finds continuous lines of 5 points with the same sign
• Rectangle Formation: Extends lines to form closed rectangles of 16 border points
• Validation: Ensures the detected rectangle has the correct geometric properties
• Point Classification: Associates the 16 border points with the full 25-point marker layout

Stage 3: Marker Identification

Once all 25 points of a marker candidate are known:

• Layout Matching: Compares the detected point pattern against known marker layouts
• Orientation Determination: Identifies the marker's orientation (0°, 90°, 180°, 270°)
• ID Assignment: Assigns the unique marker ID based on the matched layout
• Sign Confirmation: Verifies the marker's sign (positive or negative)

Stage 4: Neighborhood Analysis

The CalibrationBoardDetector establishes spatial relationships:

• Neighbor Detection: Identifies adjacent markers based on geometric proximity
• Orientation Consistency: Ensures neighboring markers have consistent relative orientations
• Board Matching: Compares detected marker neighborhoods against known calibration boards
• Coordinate Assignment: Assigns board coordinates to identified markers

4. Camera Calibration Process

The CameraCalibrator orchestrates the complete calibration workflow:

Multi-Image Processing

• Processes multiple images of the calibration board from different viewpoints
• Accumulates 2D-3D correspondences from detected markers
• Handles various camera types (pinhole, fisheye, etc.)

Initial Parameter Estimation

• Estimates initial camera field-of-view based on marker geometry
• Determines initial camera poses using detected markers
• Handles different camera models through the AnyCamera interface

Optimization

• Performs non-linear optimization to refine camera parameters
• Uses robust estimation techniques to handle outliers
• Optimizes both intrinsic parameters (focal length, distortion) and extrinsic parameters (poses)

The library is platform independent and provides comprehensive debugging and visualization capabilities through the CalibrationDebugElements class.

Typedef Documentation

CalibrationBoardObservations

using Ocean::CV::Calibration::CalibrationBoardObservations = typedef std::vector<CalibrationBoardObservation>

Definition of a vector holding CalibrationBoardObservation objects.

MarkerCandidates

using Ocean::CV::Calibration::MarkerCandidates = typedef std::vector<MarkerCandidate>

Definition of a vector holding marker candidates.

Points

using Ocean::CV::Calibration::Points = typedef std::vector<Point>

Definition of a vector holding points.