Environment Recognition for Depth Estimation Beyond Stereo Overlap

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing depth estimation techniques using stereo cameras face accuracy issues in regions where fields of view do not overlap, as they rely on direct distance measurement from overlapping regions, leading to degraded estimation accuracy.

Innovation Solution

An environment recognition apparatus and method that utilizes a first depth computation unit for overlapping regions and a second depth computation unit for non-overlapping regions, employing relevance ratio computation and feature integration to estimate depth accurately in non-overlapping regions using two-dimensional and three-dimensional information.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If a distance in a field of view overlapped region is directly taken as a distance to an object, then the depth estimation in the overlapped region is simplified, but the estimation accuracy is degraded when the distance measurement is wrong

Engineering Contradiction:
Improvedepth estimation process complexityVSAvoiddepth estimation accuracy
Core Design Contradiction:
Device complexityVSMeasurement precision

Solution Approach 1:

The patent introduces an intermediary process between direct distance measurement and final depth estimation. A relevance ratio computation unit calculates the reliability of depth information from the overlapped region, and this relevance ratio is used to weight and correct the depth estimation in the non-overlapped region, rather than directly adopting the overlapped region depth value.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent implements a feedback mechanism where the depth computation unit uses the relevance ratio (which reflects the quality of depth measurement) to adjust and correct the depth estimation. The system continuously evaluates measurement quality and uses this feedback to improve estimation accuracy in regions where direct measurement is unreliable.

Inventive Principle:
Principle #23Feedback

2Device complexity

If depth measurement is performed only in the field of view overlapped region, then the measurement process is simple, but it cannot be applied to objects in the field of view non-overlapped region

Engineering Contradiction:
Improvemeasurement process complexityVSAvoidapplicability to non-overlapped region objects
Core Design Contradiction:
Device complexityVSAdaptability or versatility

Solution Approach 1:

The patent divides the image space into three distinct regions: first overlapped region, second non-overlapped region, and third region. Different depth computation strategies are applied to each region, allowing the system to handle objects in all regions while maintaining measurement accuracy through region-specific processing.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extends depth estimation from the two-dimensional overlapped region into the third spatial dimension by projecting depth information to adjacent non-overlapped regions. This dimensional extension allows depth estimation for objects that are not directly visible in the overlapped region, significantly improving system versatility.

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

3Ease of operation

If depth estimation in non-overlapping regions is performed using direct distance measurement from overlapping regions, then the process is straightforward, but accuracy is degraded due to error propagation

Engineering Contradiction:
Improvedepth estimation easeVSAvoiddepth estimation accuracy
Core Design Contradiction:
Ease of operationVSMeasurement precision

Solution Approach 1:

The patent changes the parameter used for depth estimation by introducing a relevance ratio that quantifies the reliability of depth information. Instead of directly using distance measurements, the system computes a relevance ratio based on measurement quality, and uses this parameter to weight and correct depth estimates, thereby improving accuracy while maintaining operational simplicity.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4648009A1Environment recognition device and environment recognition method
Publication Date: 2025.11.12 HITACHI LTD
  • EP4648009A1 patent drawingFigure 1~2
  • EP4648009A1 patent drawingFigure 3
  • EP4648009A1 patent drawingFigure 4~5

AI summary

Provided are an environment recognition device and an environment recognition method which are capable of achieving a highly accurate depth estimation in a field of view non-overlap region. This environment recognition device is characterized by comprising: an image acquisition unit for acquiring an image captured by a camera; a first depth calculation unit which calculates a first depth in a first region which partially overlaps or is adjacent to the field of view of the camera; and a second depth calculation unit which uses the first depth in the first region and the image captured by the camera and calculates a second depth in a second region which is not included in the first region in the field of view of the camera.