Self-learning image geometrical distortion correction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional methods for distortion correction in images captured by cameras with non-rectilinear lenses, such as fisheye lenses, require manual adjustment of parameters and are time-consuming, especially in large installations, as they necessitate manual segmentation of floor and wall areas within the image.

Innovation Solution

A self-learning method that identifies the movement of bottom portions of objects, like feet, to automatically determine the floor boundary and build a 3D model, allowing cameras to learn and correct distortion without manual input, by assuming areas not involved in object movement are walls.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual adjustment of parameters is used for distortion correction, then the correction accuracy can be improved, but the installation time and labor cost increase significantly

Engineering Contradiction:
Improvedistortion correction accuracyVSAvoidinstallation time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The camera system automatically performs distortion correction by capturing images of a calibration pattern (such as a chessboard or circular grid), detecting the pattern features, and computing correction parameters without human intervention. This self-calibration process eliminates manual parameter adjustment while achieving accurate distortion correction.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system performs distortion correction parameters calculation and application in advance during the installation phase by capturing calibration images and processing them automatically. This preliminary calibration ensures accurate distortion correction is ready before the camera is put into operational use.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If manual segmentation of floor and wall areas is required, then the 3D model accuracy can be improved, but the complexity of operation increases

Engineering Contradiction:
Improve3D model accuracyVSAvoidoperation complexity
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The system replaces manual visual segmentation with automated computer vision algorithms that detect edges, lines, and geometric features in the captured images. These algorithms automatically identify floor and wall boundaries by analyzing pixel intensity gradients and geometric patterns, eliminating the need for manual area segmentation while maintaining accurate 3D model construction.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The camera system automatically performs feature detection and 3D model construction by processing captured images through algorithms that identify spatial relationships between detected features. The system self-calculates room dimensions, wall positions, and floor boundaries without requiring operator intervention for segmentation.

Inventive Principle:
Principle #25Self-service

3Area of stationary object

If multiple cameras are installed, then the coverage area increases, but the total calibration time increases proportionally

Engineering Contradiction:
Improvecoverage areaVSAvoidtotal calibration time
Core Design Contradiction:
Area of stationary objectVSLoss of time

Solution Approach 1:

Each camera in the multi-camera system independently performs automatic distortion correction and 3D model construction by capturing and processing its own calibration images. This parallel self-calibration approach allows multiple cameras to be calibrated simultaneously rather than sequentially, reducing total calibration time while maintaining comprehensive coverage area.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system performs preliminary automatic calibration for all cameras during the installation phase by capturing calibration patterns and processing images in parallel. This batch calibration approach prepares all cameras for operation simultaneously, avoiding the time-consuming sequential manual calibration that would be required for multiple devices.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4300410B1Self-learning image geometrical distortion correction
Publication Date: 2024.05.08 AXIS
  • EP4300410B1 patent drawingFigure 1A~1B
  • EP4300410B1 patent drawingFigure 1C~1G
  • EP4300410B1 patent drawingFigure 1H~2A

AI summary

A method (200) of distortion correction in an image captured by a non-rectilinear camera (210) is provided, including obtaining multiple images (212) of a scene captured by the camera over time, determining (220) where bottom portions (222) of objects having moved over a horizontal surface in the scene are located in the images, determining (230) a boundary (232) of the horizontal surface in the scene based on the determined locations of the bottom portions, generating (240) a three-dimensional model (242) of the scene by defining one or more vertical surfaces around the determined boundary of the horizontal surface of the scene, and correcting (250) a distortion of at least one of the images by projecting the image onto the three-dimensional model of the scene. A corresponding device, computer program and computer program product are also provided.