Vehicle Camera Stitching for Twin-Effect Artefact Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image stitching methods fail to reliably address twin-effect artefacts, where objects in component images map to separate discrete locations in the stitched image, leading to misleading and distracting visual artifacts.
Innovation Solution
A method using a convolutional neural network (CNN) to detect twin-effect artefacts by analyzing overlapping regions, followed by replacing these artefacts with dynamic imposters generated from component image data, thereby enhancing the reliability of the stitched image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If conventional image stitching methods are used to combine component images, then a stitched image with larger field of view is produced, but twin-effect artefacts appear where objects map to separate discrete locations
Solution Approach 1:
The patent applies preliminary action by detecting potential twin-effect artefacts before they become problematic visual artifacts. The system pre-identifies objects in component images that are likely to map to separate locations, and prepares correction strategies in advance. This is achieved through object detection algorithms that analyze component images before final stitching, allowing the system to anticipate and prevent artefact formation while maintaining the expanded field of view.
Solution Approach 2:
The patent introduces an intermediary correction layer between the raw stitched image and the final output. This intermediary system uses detected object information to generate corrective mappings that resolve twin-effect artefacts. The intermediary process analyzes the spatial relationships of detected objects across component images and applies targeted corrections to ensure objects map to consistent locations, thereby improving reliability without compromising the broad field of view.
2Reliability
If object detection and correction processes are applied to remove twin-effect artefacts, then image reliability is improved, but processing time and computational complexity increase
Solution Approach 1:
The patent applies segmentation by dividing the image processing task into distinct stages: initial object detection in component images, identification of potential twin-effect artefacts, and targeted correction of specific problematic regions. This segmented approach allows the system to focus computational resources only on areas where artefacts are likely to occur, rather than processing the entire stitched image uniformly. The segmentation enables efficient real-time processing while maintaining high reliability through focused correction of critical regions.
3Measurement precision
If dynamic imposters are used to replace detected artefacts, then visual accuracy is improved, but device complexity increases
Solution Approach 1:
The patent uses copying by creating dynamic imposter objects that replicate the visual appearance of correctly mapped objects. When twin-effect artefacts are detected, the system generates imposter copies of the intended object representation and places them at the correct locations in the stitched image. These copied visual elements are derived from the original component images, ensuring visual accuracy while using a relatively simple copying mechanism rather than complex generative models. The imposter copying approach maintains visual fidelity without requiring excessive system complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method, for detecting an artefact in a stitched image, comprises: acquiring component images of an environment from respective vehicle mounted cameras with overlapping fields of view; forming (410) a stitched image from the component images; processing (420) at least a portion of the stitched image corresponding to the overlapping field of view with a classifier to provide a list of detected objects from the environment at respective locations in the stitched image; determining (430) whether any detected object in the list of detected objects is a duplicate of another object in the list of detected objects; and reporting any objects that are determined to be duplicates.