Surround View Image Merging via Dual Projection Surface Segmentation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for merging images from different camera positions in surround view systems face challenges due to significant perspective differences, leading to sharp edges and unfavorable visual effects, as well as increased computing effort and image artifacts.

Innovation Solution

The method involves creating a first and second virtual projection surface with geometric shapes in a three-dimensional space, reshaping the first surface based on detected position information, assigning texture information, and transferring it to the second surface without geometric changes, to generate a common image with reduced disparities and fluid visualization.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If the projection surface is deformed dynamically to reduce disparity between images, then the visual alignment improves, but sharp edges and unfavorable visual effects occur

Engineering Contradiction:
Improveimage alignment precisionVSAvoidsharp edges and visual artifacts
Core Design Contradiction:
Manufacturing precisionVSObject-generated harmful factors

Solution Approach 1:

The projection surface is divided into two separate surfaces: a first projection surface that is deformed to match the perspective of individual camera images, and a second projection surface that remains geometrically stable for generating the final composite view. This segmentation allows the deformation operation to be isolated to only the first surface, preventing the generation of sharp edges and visual artifacts in the final output while still achieving accurate image alignment through the texture mapping process.

Inventive Principle:
Principle #1Segmentation

2Manufacturing precision

If dynamic deformation of projection surface is applied to merge images, then image disparity is reduced, but computing effort increases

Engineering Contradiction:
Improveimage merging accuracyVSAvoidcomputing effort
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The first projection surface is pre-deformed to match the perspective of individual camera images before texture mapping occurs. By performing this deformation operation in advance on the first surface, the system avoids the need for complex real-time deformation calculations during the final image generation process, thereby reducing overall computing effort while maintaining high merging accuracy.

Inventive Principle:
Principle #10Preliminary action

3Adaptability or versatility

If multiple camera images are merged to create surround view, then comprehensive environmental monitoring is achieved, but perspective differences cause image discrepancies

Engineering Contradiction:
Improvesurround view capabilityVSAvoidimage consistency
Core Design Contradiction:
Adaptability or versatilityVSManufacturing precision

Solution Approach 1:

The first projection surface acts as an intermediary that is specifically deformed to match the perspective of each individual camera image. This intermediary surface serves as a bridge between the different camera perspectives and the final composite view, allowing accurate texture mapping from multiple cameras while maintaining image consistency through the association of corresponding points between the first and second projection surfaces.

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP3117399B1Method for assembling single images that have been taken from different positions by a camera system to form a joint image
Publication Date: 2021.03.10 ROBERT BOSCH GMBH
  • EP3117399B1 patent drawingFigure 1
  • EP3117399B1 patent drawingFigure 2
  • EP3117399B1 patent drawingFigure 3~4

AI summary

The present invention relates to a method for assembling single images that have been taken from different positions by a camera system to form a joint image, comprising provision of a first projection area (G') having a first geometric shape and of a second projection area (G) having a second geometric shape, wherein each point (P'0–P'7) on the first projection area (G') has an associated point (P0–P7) on the second projection area (G), capture of position information that describes an arrangement of objects (30), shown in the single images, in relation to the camera system, reshaping of the first geometric shape of the first projection area (G') on the basis of the captured position information, association of texture information from the single images with area regions of the reshaped first projection area (G'), transfer of the texture information from the points (P'0–P'7) on the first projection area (G') to the corresponding associated points (P0–P7) on the second projection area (G), and production of the joint image from a view of the second projection area (G).