Variable 3D AVM System Using Deep Learning for Terrain Adaptation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing 3D around view monitoring (3D AVM) technologies suffer from terrain distortion due to the projection of camera videos onto fixed 3D projection planes, which can lead to inaccurate representation of terrain features and potentially misleading information for drivers.
Innovation Solution
A variable-type 3D AVM system utilizing a deep-learning neural network model in an edge-cloud environment estimates terrain characteristics from camera videos and generates dynamic 3D projection planes based on terrain maps, eliminating the need for fixed projection shapes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a fixed 3D projection plane is used to project camera videos, then the 3D AVM system can be implemented with simple processing, but terrain distortion occurs and terrain features are not accurately represented
Solution Approach 1:
The patent applies the dynamics principle by transitioning from a fixed 3D projection plane to a dynamic terrain-adaptive projection plane. The system estimates terrain characteristics in real-time and adjusts the projection plane shape accordingly, allowing the projection geometry to adapt dynamically to different terrain conditions (narrow roads, parking lots, open areas) while maintaining accurate terrain feature representation without requiring complex manual configuration
Solution Approach 2:
The patent implements parameter changes by modifying the projection plane parameters (shape, curvature, orientation) based on estimated terrain characteristics. The system changes the projection parameters dynamically according to the detected terrain type, enabling accurate terrain representation across various environments while keeping the overall system implementation relatively simple through automated parameter adjustment
2Manufacturing precision
If terrain characteristic estimation using deep learning is implemented, then terrain distortion is eliminated, but system complexity and computational requirements increase
Solution Approach 1:
The patent applies preliminary action by pre-training the deep learning neural network model offline to perform terrain characteristic estimation. The pre-trained model is then deployed in the 3D AVM system, allowing real-time terrain adaptation without requiring complex online training procedures. This approach reduces the computational burden during actual operation while maintaining high accuracy in terrain feature representation
Solution Approach 2:
The patent introduces an intermediary deep learning neural network model that acts as a bridge between the camera videos and the 3D projection system. This intermediary component estimates terrain characteristics automatically, simplifying the overall system architecture by replacing complex manual terrain analysis and projection plane configuration with an automated AI-based intermediary layer
Data Source
AI summary
The present invention generally relates to a 3D around view monitoring (3D AVM) technology for a vehicle. In particular, the invention relates to a technology enabling a distortion-free 3D AVM image plane to be formed by estimating a terrain characteristic from camera videos by using a deep-learning neural network model in an edge-cloud environment and making and utilizing a 3D map. The invention has an advantage of enhancing the convenience of driving a vehicle by forming a 3D AVM image plane from which a terrain distortion is eliminated.