Monocular 3D Obstacle Detection via Local Depth Map Computation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing driver assistance systems face computational overhead challenges in detecting three-dimensional structures from depth map images in real-time, particularly for objects like guardrails and curbs, due to the high computational cost of computing dense depth maps at high resolutions.
Innovation Solution
A method that uses monocular cues to detect candidate regions for three-dimensional structures by computing depth maps locally around detected features, adjusting resolution based on the accuracy required, and comparing these maps with pre-defined models to determine if objects are obstacles, thereby reducing computational overhead and enabling real-time processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If dense depth maps are computed at high resolution using multi-camera stereo or structure from motion, then measurement precision of three-dimensional structures is improved, but computational cost and processing time increase significantly
Solution Approach 1:
The patent divides the image into multiple regions of interest (ROIs) based on detected features. Depth maps are computed only for these specific ROIs rather than the entire image, significantly reducing computational load while maintaining precision where needed. This segmentation allows the system to process depth information locally around important features without the overhead of full-image processing.
Solution Approach 2:
The patent applies different processing quality levels to different regions of the image. High-resolution depth maps are computed only for regions containing detected features, while other regions receive lower resolution or no depth processing. This local quality approach ensures measurement precision is maintained where it matters most while reducing overall computational cost.
2Reliability
If depth maps are computed for the entire image at full resolution, then detection accuracy of three-dimensional obstacles is improved, but device complexity and computational overhead increase
Solution Approach 1:
The patent performs preliminary feature detection on the input image before computing depth maps. This preliminary action identifies regions of interest that contain potential obstacles or important features. By detecting features first and then computing depth maps only for those specific regions, the system reduces computational overhead while maintaining detection accuracy for critical areas.
Solution Approach 2:
The patent computes depth maps partially, only for specific regions of interest rather than the entire image. This partial action approach processes only the necessary portions of the image that contain detected features, reducing computational complexity while maintaining sufficient detection accuracy for obstacle identification.
Data Source
AI summary
Detection of three dimensional obstacles using a system mountable in a host vehicle including a camera connectible to a processor. Multiple image frames are captured in the field of view of the camera. In the image frames, an imaged feature is detected of an object in the environment of the vehicle. The image frames are portioned locally around the imaged feature to produce imaged portions of the image frames including the imaged feature. The image frames are processed to compute a depth map locally around the detected imaged feature in the image portions. Responsive to the depth map, it is determined if the object is an obstacle to the motion of the vehicle.