Open Door Detection From Sparse Vehicle Image Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in navigating due to the vast volume of data from sensors and traditional mapping technologies, which can limit navigation accuracy and efficiency.
Innovation Solution
A system and method for autonomous vehicle navigation using sparse representations of image frames, object detection networks, and open door detection networks to determine navigational actions based on image analysis, enabling efficient data processing and decision-making.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional mapping technology and full image data are used for navigation, then comprehensive environmental information is obtained, but data processing complexity and computational load increase significantly
Solution Approach 1:
The patent extracts only the essential and relevant features from image frames (sparse representation) rather than processing complete image data. This selective extraction of key visual information reduces computational load while maintaining navigation accuracy by focusing on critical environmental cues.
Solution Approach 2:
The patent applies different processing quality levels to different regions of the image frame. High-detail processing is applied only to regions containing relevant navigation information, while other regions receive minimal or no processing. This local differentiation reduces overall computational complexity while preserving navigation-critical details.
2Measurement precision
If comprehensive image data is processed in real-time, then accurate environmental perception is achieved, but processing time increases
Solution Approach 1:
The system extracts only the most salient features from image frames for processing, eliminating redundant data. This sparse representation approach maintains environmental perception accuracy by focusing on critical features while dramatically reducing the time required to process visual information.
Solution Approach 2:
The patent applies partial processing to image data, focusing computational resources only on the portions of the image that contain navigation-relevant information. This selective processing achieves sufficient environmental perception accuracy without the time cost of processing the entire image frame.
3Measurement precision
If complete image frames are analyzed for object detection, then detection accuracy is improved, but computational resources required increase
Solution Approach 1:
The system extracts only the essential visual features needed for object detection from complete image frames. This sparse representation maintains detection accuracy for navigation-critical objects while significantly reducing the computational energy required to process and analyze visual data.
Data Source
AI summary
A computer-implemented method for navigating a host vehicle includes receiving an image frame from an image capture device, the image frame representing an environment of the host vehicle and including a representation of a target vehicle; analyzing the image frame to determine a sparse representation of a portion of the image frame; providing the sparse representation to an object detection network; receiving an identifier of a candidate region identified by the object detection network; based on the identifier, extracting the candidate region from the image frame; providing the candidate region to an open door detection network; determining a navigational action for the host vehicle in response to an indication from the open door detection network that the candidate region includes a representation of a door of the target vehicle in an open condition; and causing an actuator associated with the host vehicle to implement the navigational action.