Night Obstacle Detection via Shadow Thickness Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current driver assistance systems face limitations in detecting obstacles at night due to the range and accuracy of shadow detection, particularly with stereo cameras, which are restricted by focal length and baseline, and struggle to differentiate between shadows and other dark patches on the road.
Innovation Solution
A driver assistance system using a camera connected to a processor that captures and processes image frames to locate and track dark image patches, measuring their thicknesses to identify shadows cast by obstacles from headlights, allowing for the differentiation between obstacles and road marks, and providing audible warnings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If stereo camera is used for shadow detection, then depth information can be obtained, but detection range is limited by camera focal length and stereo baseline
Solution Approach 1:
The patent transitions from 2D image plane analysis to 3D spatial analysis by tracking shadow motion across multiple frames. Instead of relying on stereo baseline for depth, the system uses temporal dimension (multiple frames) to infer spatial relationships through shadow displacement patterns, effectively adding a time dimension to overcome the limited spatial baseline.
Solution Approach 2:
The system performs preliminary shadow detection and tracking in the image plane before attempting depth estimation. By first identifying and tracking dark patches across frames to establish their motion patterns, the system prepares the data foundation needed for accurate depth inference, rather than attempting direct depth measurement with limited stereo baseline.
2Device complexity
If stereo camera with side-by-side mounting is used, then device complexity is reduced, but depth information on shadow edge is only available when light sources are displaced along the stereo baseline
Solution Approach 1:
The patent introduces shadow motion tracking as an intermediary process between the fixed stereo camera configuration and depth estimation. Instead of directly measuring shadow edge depth with limited stereo displacement, the system uses the temporal displacement of shadows across frames as an intermediate measurement that can be converted into depth information, bridging the gap between fixed mounting and accurate depth measurement.
3Object-affected harmful factors
If shadow detection is performed at night, then obstacle detection is enabled in low light conditions, but difficulty differentiating between shadows and other dark patches on the road increases
Solution Approach 1:
The patent applies dynamics by tracking shadow motion across multiple frames rather than analyzing static images. Shadows cast by obstacles exhibit characteristic motion patterns as the vehicle moves forward, while dark road patches remain stationary. This temporal dynamic analysis enables differentiation between moving shadows and static dark areas, solving the discrimination problem in low light conditions.
Solution Approach 2:
The system uses feedback from multi-frame tracking to continuously refine shadow identification. By comparing shadow position, shape, and motion across successive frames, the system feedback-corrects its identification of true shadows versus false dark patches, improving differentiation accuracy over time rather than relying on single-frame analysis.
Data Source
Figure 1~2
Figure 3a~3b
Figure 4a~4b
AI summary
A driver assistance systems mountable in a host vehicle while the host vehicle is moving forward with headlights on for detection of obstacles based on shadows. The driver assistance system includes a camera (12) operatively connectible to a processor (30). A first image frame and a second image frame are captured of a road . A first dark image patch and a second dark image patch include intensity values less than a threshold value. The first and the second dark image patches are tracked from the first image frame to the second image frame as corresponding images of the same portion of the road. Respective thicknesses in vertical image coordinates are measured of the first and second dark image patches responsive to the tracking.