Autonomous Vehicle Object Detection Using Virtual Road Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in detecting objects at long ranges due to limitations in LIDAR and RADAR systems, and camera processing is complex, especially under varying lighting conditions and shadows.
Innovation Solution
The system uses image data from cameras to identify known structures in an environment, processing statistics to differentiate between road features and unknown objects, allowing for effective object detection and navigation by comparing image data portions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Length of stationary object
If LIDAR and RADAR systems are used for long-range object detection, then detection range is improved, but system complexity and cost increase
Solution Approach 1:
The patent creates a virtual model (copy) of the road geometry and features based on data from known structures, then uses this virtual model to detect objects in camera images. This copying approach enables long-range detection capability without requiring complex LIDAR or RADAR hardware systems.
2Device complexity
If camera processing is used for object detection, then device complexity is reduced, but processing complexity under varying lighting conditions increases
Solution Approach 1:
The patent introduces an intermediary virtual model of the road that serves as a reference between the camera images and the object detection process. This intermediary model, built from known road structures, helps filter out lighting variations and shadows, making the detection process simpler and more reliable.
Solution Approach 2:
The patent extracts and isolates the road geometry and features from the complex image data by using known road structures as reference points. This extraction process separates the road information from other elements like shadows and lighting variations, simplifying the overall processing.
3Device complexity
If camera processing is used for object detection, then device complexity is reduced, but detection precision at long ranges deteriorates
Solution Approach 1:
The patent performs preliminary actions by pre-building a virtual model of the road geometry and features using known road structures before the actual object detection takes place. This preliminary modeling enhances the precision of long-range detection by providing a reference framework that compensates for the limitations of camera-based systems.
Data Source
AI summary
An autonomous vehicle may be configured to detect objects based on known structures of an environment. The vehicle may be configured to obtain image data from a sensor and be configured to operate in an autonomous mode. The image data may include data indicative of a known structure in the environment. The vehicle may include a computer system. The computer system may determine, based on a first portion of the image data, information indicative of an appearance of the known structure. The computer system may determine, based on a second portion of the image data, information indicative of an appearance of an unknown object in the environment. The computer system may also compare the information indicative of the appearance of the known structure with the information indicative of the appearance of the unknown object and provide instructions to control the vehicle in the autonomous mode based on the comparison.