Stereo Disparity Maps for Hazard Detection on Uneven Paths
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing hazard detection technologies for autonomous vehicles are inadequate due to reliance on inaccurate assumptions, high training data requirements, and high costs, and struggle with detecting small hazards at a distance, particularly when road surfaces are not flat.
Innovation Solution
Utilizing stereo disparity between a pair of images to generate a baseline road disparity model, identifying hazards from deviations in the disparity map, and applying blockwise division and clustering algorithms to enhance detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If parallax-based image processing is used for hazard detection, then the system can detect hazards using a single camera, but the system requires the vehicle to travel a sufficient distance with sufficient time latency between capturing image frames, reducing detection speed and responsiveness
Solution Approach 1:
The patent divides the hazard detection task into two independent components: (1) road surface modeling using stereo disparity to establish expected depth patterns, and (2) hazard detection by identifying deviations from the model. This segmentation allows simultaneous processing of road geometry and hazard identification, eliminating the time latency required by parallax-based methods that process sequential frames.
Solution Approach 2:
The system performs preliminary action by pre-establishing a road surface disparity model from stereo image pairs before hazard detection. This model captures the expected depth variations of the road surface, allowing the system to immediately identify hazards as deviations from this pre-computed model, rather than waiting for vehicle movement to create parallax effects.
2Ease of manufacture
If parallax-based image processing is used for hazard detection, then the system uses a single camera, but the system is highly dependent on a flat road assumption which is not always case in real situations, reducing detection accuracy
Solution Approach 1:
The patent applies local quality by creating a localized road surface model for each region of the image using stereo disparity. Instead of assuming a globally flat road surface, the system computes depth information locally for each pixel or pixel group, capturing local variations in road geometry such as slopes, curves, and uneven surfaces. This local modeling approach maintains detection accuracy on non-flat roads while keeping the system relatively simple.
3Measurement precision
If LiDAR sensors are used for hazard detection, then the system can actively sense 3D surroundings and detect hazards above road surface, but the sensors are expensive and may not be economically practical
Solution Approach 1:
The patent creates a computational copy of the 3D road surface model using stereo vision algorithms. Instead of using expensive LiDAR hardware to directly measure 3D depth, the system computes depth information by processing pairs of 2D images from stereo cameras through disparity mapping. This computational approach replicates the 3D sensing capability of LiDAR using affordable camera hardware, making the system economically practical while maintaining measurement precision.
4Measurement precision
If LiDAR sensors are used for hazard detection, then the system can detect hazards at distance, but the density of the point cloud generated by LiDAR reduces with range, making it difficult to detect small hazards and estimate hazard size
Solution Approach 1:
The patent merges multiple 2D image frames captured from stereo cameras to construct a comprehensive 3D road surface model. By integrating disparity information from multiple views and combining data across different spatial locations, the system maintains high point density in the reconstructed 3D model even at long ranges. This merging approach compensates for the natural reduction in individual pixel density with distance, enabling detection of small hazards that would be lost in sparse LiDAR point clouds.
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
Provides robust and precise hazard detection by distinguishing road disparities from hazard pixels, enabling effective navigation around obstacles, reducing false positives, and optimizing computational resources.
Implementation Method 1
stereo disparity between a pair of images to generate a baseline road disparity model
Data Source
AI summary
In examples, system and methods for stereo disparity based hazard detection for autonomous machine applications are presented. Example embodiments may assist an ego-machine in detecting hazards within its path of travel. The systems and methods may use disparity between a stereo pair of images to generate a baseline path disparity model and further identify hazards from detected disparities that deviate from that path disparity model. A disparity map for the image pair is constructed in which each pixel represents a disparity for a corresponding element of the image captured. Blockwise division may be optionally used to subdivide the disparity map into a plurality of smaller disparity maps, each corresponding to a block of pixels of the disparity map. A V-space disparity map, where a first axis corresponds to disparity values and the second axis corresponds to pixel rows, may be used to simplify estimation of the path disparity model.