Stereo-Assisted Monocular Depth Estimation on Non-Flat Roads
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current depth estimation methods for semi-autonomous and autonomous vehicles, such as LiDAR sensors, are costly and sensitive to weather conditions, while stereo cameras face errors that grow quadratically with distance and are computationally expensive, making them unsuitable for real-time long-distance depth estimation on non-flat roads.
Innovation Solution
A system utilizing a processor and stereo camera pair to calculate road profiles and determine object distances based on disparity between images, incorporating map data and machine learning models to improve depth estimation accuracy and reduce computational costs, allowing for accurate depth estimation on non-flat roads without relying on stereo depth estimation for long-distance objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If stereo cameras are used for depth estimation, then cost is reduced compared to LiDAR, but depth estimation error grows quadratically with distance
Solution Approach 1:
The system segments the depth estimation task into two parts: near-field depth estimation using stereo cameras and far-field depth estimation using monocular camera with map data. This allows each method to operate in its optimal range, preventing the quadratic error growth of stereo cameras at long distances while maintaining cost effectiveness.
Solution Approach 2:
Map data serves as an intermediary to assist the monocular camera in estimating far-field depth. By combining map information with monocular vision, the system achieves accurate long-distance depth estimation without relying on stereo cameras, thus avoiding their quadratic error problem while maintaining low cost.
2Ease of manufacture
If stereo cameras are used for depth estimation, then depth can be estimated without LiDAR, but computational complexity increases
Solution Approach 1:
The system applies stereo depth estimation only partially - specifically for near-field objects where it is most effective. For far-field objects, it switches to monocular estimation with map data assistance, reducing overall computational complexity while maintaining accuracy where it matters most.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables accurate and efficient depth estimation on non-flat roads, reducing errors and computational power requirements, facilitating safe and efficient autonomous vehicle operation by integrating map data and machine learning for improved localization and obstacle detection.
Implementation Method 1
Stereo vision is similar to three-dimensional (3D) perception in human vision and is based on the triangulation of rays from multiple viewpoints
Data Source
AI summary
A non-transitory processor-readable medium stores code representing instructions to be executed by the processor. The code comprises code to cause the processor to receive a first image and a second image from a stereo camera pair disposed with a vehicle. The code causes the processor to detect, using a machine learning model, an object based on the first image, the object located within a pre-defined area within a vicinity of the vehicle. The code causes the processor to determine a distance between the object and the vehicle based on disparity between the first image and the second image. The code causes the processor to determine a longitudinal value of the vehicle based on the distance and a height of the vehicle. The code causes the processor to send an instruction to facilitate driving of the vehicle based on a road profile associated with the longitudinal value.