Depth Image Registration Using Geometric Shape Relationships

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing depth image registration methods face challenges in aligning images captured from different fields of view, requiring significant computational resources and time, and struggle to accurately identify 3D objects in scenes with varying camera positions.

Innovation Solution

A method that registers depth images by identifying geometric relationships between geometric shapes in each image, such as planes, rather than the objects themselves, using techniques like RANSAC for plane detection and geometric transformations to align the images.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If depth images are registered using conventional structure-from-motion or multiview-stereo approaches, then 3D models can be generated, but the methods are extremely sensitive to initial conditions and require manually placed control points which limits automation

Engineering Contradiction:
Improvedepth image registration accuracyVSAvoidmanual control point placement requirement
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The system automatically detects and tracks feature points across multiple depth images without requiring manual control point placement. The automated feature point selection and registration process eliminates the need for user intervention, making the system self-sufficient while maintaining high registration accuracy through iterative optimization algorithms.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system performs preliminary feature point detection and initial registration before final depth map fusion. By pre-processing the depth images to identify key features and establish initial correspondences, the system prepares the data in advance for more accurate and efficient registration, reducing sensitivity to initial conditions.

Inventive Principle:
Principle #10Preliminary action

2Area of stationary object

If multiple depth maps from different viewpoints are captured to improve 3D model completeness, then coverage increases, but registration errors and processing complexity increase

Engineering Contradiction:
Improve3D model coverageVSAvoidregistration processing complexity
Core Design Contradiction:
Area of stationary objectVSDevice complexity

Solution Approach 1:

The system segments the 3D modeling process into distinct stages: feature point detection, correspondence matching, initial registration, and refinement. By dividing the complex registration task into manageable segments, the system can process multiple depth maps from different viewpoints systematically, reducing overall processing complexity while maintaining comprehensive coverage.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The registration process uses dynamic iterative optimization that adapts to the specific characteristics of each depth map pair. The algorithm dynamically adjusts registration parameters and feature weighting based on the quality and quantity of available depth maps, allowing the system to handle varying levels of complexity automatically as more viewpoints are added.

Inventive Principle:
Principle #15Dynamics

3Measurement precision

If feature points are densely sampled across the entire depth image to improve registration accuracy, then precision increases, but computational time and processing load increase significantly

Engineering Contradiction:
Improveregistration precisionVSAvoidcomputational processing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system applies different feature point sampling densities to different regions of the depth image based on local importance. Areas with high geometric complexity or significant depth variations receive denser feature point sampling, while uniform regions use sparser sampling. This local quality approach maintains registration precision in critical areas while reducing overall computational burden.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The system uses a two-stage feature point selection process where an initial coarse set of features is used for rough registration, followed by a refined set of features for precision optimization. This partial action approach achieves sufficient precision without the excessive computational cost of using all possible feature points from the beginning.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3928281B1Method for registering depth images
Publication Date: 2026.05.06 FOGALE NANOTECH SA
  • EP3928281B1 patent drawingFigure 1
  • EP3928281B1 patent drawingFigure 2
  • EP3928281B1 patent drawingFigure 3~4

AI summary

The invention relates to a method (100) for registering two depth images of a real scene, comprising the following steps: - for each of the depth images: - detecting (1081, 1121, 1082, 1122) a plurality of geometric shapes in the depth image, and - determining (1101, 1102, 1141, 1142) at least one geometric relationship between at least two geometric shapes; - identifying (116, 118) geometric shapes common to the two images by collating the detected geometric relationships; - calculating (120), according to the common geometric shapes, a displacement matrix; and - registering (122) one of the images, with respect to the other image, according to the displacement matrix. It also relates to a method for monitoring an area of interest and a method for monitoring the environment of a robot implementing such a method for registering depth images.