PTZ Camera Calibration Using Iterative Feature Matching

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for calibrating pan, tilt, zoom (PTZ) cameras with respect to fixed cameras suffer from low precision due to limited features in images and geometric distortions, such as barrel distortion, which affect the accuracy of feature matching and mapping.

Innovation Solution

A method that involves receiving an overview image from a fixed camera, identifying features in the PTZ camera's image, localizing these features in the overview image, logging positional data, and defining a mapping between the two images. If the mapping quality is below a threshold, the PTZ camera is redirected to improve feature density or reduce distortion effects, and the calibration process is repeated until the quality meets the threshold.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If feature-matching calibration is performed using a single image direction, then the calibration process is simple and fast, but the precision and accuracy of the calibration are insufficient due to limited features and geometric distortions

Engineering Contradiction:
Improvecalibration precisionVSAvoidcalibration process complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs preliminary actions by capturing multiple images from different PTZ camera directions before calibration. These pre-captured images with varying feature densities and distortion levels are stored and selected during calibration to ensure optimal matching conditions are met, thereby improving precision without adding complexity to the actual calibration execution

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The calibration process is made dynamic by allowing selection of different PTZ image directions based on real-time assessment of feature density and distortion characteristics. Instead of using a fixed single-direction approach, the system adaptively chooses the most suitable image direction from multiple pre-captured options, enabling precision improvement while maintaining process efficiency

Inventive Principle:
Principle #15Dynamics

2Measurement precision

If the PTZ camera is directed towards portions of the scene with few objects, then the field of view coverage is achieved, but the number of features for matching is insufficient, reducing calibration accuracy

Engineering Contradiction:
Improvefeature matching accuracyVSAvoidscene coverage
Core Design Contradiction:
Measurement precisionVSArea of stationary object

Solution Approach 1:

The calibration process is segmented into multiple independent calibration points across different scene regions. Instead of relying on a single comprehensive image, the system divides the calibration into multiple stages, each using images from specific PTZ directions that capture sufficient features in particular areas, thereby ensuring both accuracy and coverage

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system adds a temporal dimension to the calibration process by capturing and utilizing multiple images from different PTZ directions rather than relying on a single spatial image. This multi-dimensional approach allows feature matching to be performed across multiple views, ensuring sufficient feature availability while maintaining comprehensive scene coverage

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Area of stationary object

If wide field of view lenses are used in fixed cameras, then overview coverage is achieved, but barrel distortion and geometric distortions increase, making feature matching difficult

Engineering Contradiction:
Improveoverview field of viewVSAvoidfeature matching precision
Core Design Contradiction:
Area of stationary objectVSMeasurement precision

Solution Approach 1:

The system changes the parameter of image selection by choosing PTZ images from directions where distortion effects are minimized. Instead of attempting to correct distortion in all images, the methodology selectively uses images where the geometric distortion parameters are most favorable for accurate feature matching, thereby maintaining both wide coverage and high precision

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The system incorporates feedback mechanisms to assess the quality of feature matching based on distortion levels. By evaluating matching results and identifying when distortion degrades precision, the system can select alternative PTZ image directions with lower distortion, ensuring accurate calibration while preserving the benefits of wide field of view lenses

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3451650B1A method of calibrating a direction of a pan, tilt, zoom, camera with respect to a fixed camera, and a system in which such a calibration is carried out
Publication Date: 2020.01.08 AXIS
  • EP3451650B1 patent drawingFigure 1
  • EP3451650B1 patent drawingFigure 2
  • EP3451650B1 patent drawingFigure 3

AI summary

There is provided a method of calibrating a direction of a pan, tilt, zoom (PTZ) camera with respect to a fixed camera. An overview image of a scene captured by the fixed camera, and an image of the scene captured by the PTZ camera when directed in a first direction are received. By matching features in the overview image and in the image captured by the PTZ camera, a fist calibration is carried out by correlating the first direction to a position of the matching features in the overview image. Further, a mapping between the image captured by the PTZ camera and the overview image is defined based on the matching features. The mapping is used to map an object from the image captured by the PTZ camera to the overview image. Based on an appearance of the mapped object, a quality of the mapping is calculated. If the quality is not good enough, the PTZ camera is redirected to a second direction, and a further calibration is carried out by again matching features.