Stereo Depth Hazard Detection for Small Distant Road Obstacles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional hazard detection systems for autonomous and semi-autonomous vehicles are inadequate due to reliance on extensive training data, inaccurate assumptions, high costs, and limited detection capabilities, particularly for small hazards at a distance.
Innovation Solution
A multi-view geometry-based system using synchronized stereo cameras to generate depth maps and identify discontinuities in disparity values, cropping images to focus on the driving surface, and applying a disparity threshold to detect hazards.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep neural network approaches are used for hazard detection, then detection capability is improved, but training data requirements and system complexity increase significantly
Solution Approach 1:
The patent extracts and removes the need for extensive training data by using multi-view geometry and disparity analysis. Instead of relying on DNN training, the system directly computes depth discontinuities from stereo camera pairs, eliminating the training data burden while maintaining hazard detection capability.
Solution Approach 2:
The patent replaces the computational mechanics of DNN training with geometric mechanics of stereo vision. By substituting the learning-based approach with direct geometric computation of disparity maps, the system achieves hazard detection without requiring large training datasets.
2Ease of manufacture
If single camera planar-parallax-based approaches are used, then system cost is reduced, but detection accuracy deteriorates due to inaccurate assumptions about roadway geometry
Solution Approach 1:
The patent segments the scene into distinct depth layers by analyzing disparity discontinuities in the depth map. This segmentation allows the system to identify hazards at depth boundaries without assuming the roadway is piecewise planar, thereby improving detection accuracy while maintaining single-camera feasibility.
Solution Approach 2:
The patent transitions from 2D image plane analysis to 3D depth space by computing disparity maps. This dimensional transformation enables accurate hazard detection at various distances without relying on inaccurate planar roadway assumptions, resolving the accuracy-cost contradiction.
3Length of stationary object
If LiDAR systems are used for hazard detection, then detection range is improved, but system cost increases prohibitively
Solution Approach 1:
The patent creates a computational copy of LiDAR-like depth information using inexpensive stereo camera pairs. By synthesizing depth maps through multi-view geometry, the system achieves LiDAR-equivalent detection range and capability at a fraction of the cost, eliminating the need for actual LiDAR hardware.
Solution Approach 2:
The patent substitutes the physical LiDAR scanning mechanism with optical stereo vision and computational disparity analysis. This replacement maintains long-range detection capability while dramatically reducing system cost, as cameras and processors are far cheaper than LiDAR systems.
4Device complexity
If conventional hazard detection systems are used, then implementation is simpler, but detection of small distant hazards is inadequate
Solution Approach 1:
The patent enhances small hazard detection by computing disparity maps that provide explicit depth information. This dimensional transformation from 2D to 3D space allows the system to detect small distant hazards that would be imperceptible in standard images, while maintaining computational feasibility through efficient stereo matching algorithms.
Data Source
AI summary
In various examples, systems and methods are disclosed that detect hazards on a roadway by identifying discontinuities between pixels on a depth map. For example, two synchronized stereo cameras mounted on an ego-machine may generate images that may be used extract depth or disparity information. Because a hazard's height may cause an occlusion of the driving surface behind the hazard from a perspective of a camera(s), a discontinuity in disparity values may indicate the presence of a hazard. For example, the system may analyze pairs of pixels on the depth map and, when the system determines that a disparity between a pair of pixels satisfies a disparity threshold, the system may identify the pixel nearest the ego-machine as a hazard pixel.