Vehicle Surround View Stitching With 3D Object Contour Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle surround view systems suffer from objects disappearing or generating multiple display ghosts due to differences in camera positions and fields of view, affecting parking safety.
Innovation Solution
A method that adaptively considers the position of foreground objects and imaging sources during the stitching of multiple exterior images, using geometric contours and 3D models to project objects accurately on a vehicle surround image, avoiding partial disappearance and multiple ghosts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If multiple exterior images from cameras with different fields of view are stitched together, then a complete vehicle surround image is obtained, but objects may disappear or generate multiple display ghosts
Solution Approach 1:
The patent introduces a 3D model of the vehicle as an intermediate dimensional space to map and integrate 2D images from multiple cameras. By projecting image data onto the 3D vehicle model and then rendering the surround view, the system resolves spatial ambiguities and ensures consistent object representation across different camera fields of view, eliminating disappearance and ghosting artifacts.
Solution Approach 2:
The patent employs a 3D vehicle model as an intermediary structure that mediates between multiple camera inputs and the final surround view output. This intermediary model provides a unified spatial reference frame, allowing the system to consistently track and represent objects across different camera perspectives without direct pixel-level stitching that causes artifacts.
2Productivity
If simple image stitching is used to generate vehicle surround image, then processing speed is fast, but object position accuracy deteriorates
Solution Approach 1:
The patent performs preliminary actions by pre-building a 3D vehicle model and pre-establishing projection relationships between camera views and the 3D model. During real-time processing, objects are tracked in the 3D space and projected to the appropriate view, which is computationally more efficient than performing complex geometric transformations and stitching operations on the final images, thus maintaining both speed and accuracy.
Data Source
AI summary
A processing method for vehicle surround view, a vehicle surround view system and an in-vehicle processing device are provided. The processing method for vehicle surround view includes the following steps: obtaining a first exterior image and a second exterior image; recognizing a first interest object in the first exterior image and a second interest object in the second exterior image; fitting the first interest object to a first geometric contour and fitting the second interest object to a second geometric contour; applying the first geometric contour and the second geometric contour on a 3D model, to obtain a ground point, a first contour position and a second contour position; obtaining a merged contour position according to the ground point; and projecting the first interest object or the second interest object at the merged contour position on a vehicle surround image.