Camera Overlap Region Obstacle Identification via Feature Projection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional driver assistance systems with overlapping camera fields of view struggle to effectively identify obstacles in overlap regions, leading to inadequate safety functions due to poor visibility of obstacles located in or extending into these areas.

Innovation Solution

A system comprising multiple cameras and a processing unit that prioritizes image data from cameras based on the number of projected features in the overlap region, using binary feature extraction and projection onto a ground plane to assemble a representative image of the surroundings for improved obstacle identification.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Area of stationary object

If conventional surroundings imaging systems use overlapping camera fields of view to cover the vehicle's surroundings, then the coverage area is improved, but the obstacle identification accuracy in overlap regions deteriorates

Engineering Contradiction:
Improvecoverage areaVSAvoidobstacle identification accuracy
Core Design Contradiction:
Area of stationary objectVSMeasurement precision

Solution Approach 1:

The patent applies local quality by differentiating the treatment of different image regions. Overlap regions are identified and processed differently from non-overlap regions. The system selectively prioritizes image data from specific cameras based on the location within the image, assigning different weights or selection criteria to overlap versus non-overlap areas, thereby improving obstacle identification in the critical overlap regions while maintaining overall coverage

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent changes the parameter of image selection priority based on the spatial location. By determining whether a region is an overlap region and adjusting the priority of image data accordingly, the system dynamically modifies which camera's data is used in different areas. This parameter change allows the system to optimize for accuracy in overlap regions while preserving the benefits of overlapping coverage

Inventive Principle:
Principle #35Parameter changes

2Area of stationary object

If image data from both cameras is combined in overlap regions, then the coverage is improved, but the complexity of image processing increases

Engineering Contradiction:
Improveimage coverageVSAvoidimage processing complexity
Core Design Contradiction:
Area of stationary objectVSDevice complexity

Solution Approach 1:

The patent extracts and identifies overlap regions from the combined image data of multiple cameras. By separating the overlap region determination as a distinct step, the system can then apply specific processing rules to these regions, simplifying the overall complexity by creating a clear workflow: identify overlap regions, then process them with prioritized selection rather than complex fusion algorithms

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Instead of combining image data from both cameras in overlap regions, the patent inverts the approach by selectively choosing data from one camera based on priority criteria. This inversion simplifies processing by avoiding the need to merge and reconcile data from multiple sources, reducing computational complexity while maintaining coverage through the prioritized selection strategy

Inventive Principle:
Principle #13The other way round (Inversion)

Data Source

PatentUS10824884B2Device for providing improved obstacle identification
Publication Date: 2020.11.03 CONTI TEMIC MICROELECTRONIC GMBH
  • US10824884B2 patent drawing
  • US10824884B2 patent drawing
  • US10824884B2 patent drawing

AI summary

A device provides improved obstacle identification. A first camera acquires first vehicle image data and provides it to a processing unit. A second camera acquires and provides second vehicle image data. An image overlap region has at least a portion of the first vehicle image data and at least a portion of the second vehicle image data. The first and second vehicle image data extend over a ground plane and the image overlap region extends over an overlap region of the ground plane. The processing unit extracts first image features from the first vehicle image data and extracts second image features from the second vehicle image data. The processing unit projects the first and the second image features onto the ground plane. The processing unit produces at least one image of the surroundings, having either at least a portion of the first vehicle image data associated with the overlap region, or at least a portion of the second vehicle image data associated with the overlap region, based in part on the determination of first image features whose projections lie in the overlap region of the ground plane, and on second image features whose projections lie in the overlap region of the ground plane.