Lane Detection Using Ontology-Based Semantic Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems for detecting lanes in a driving scene rely heavily on map data, which can be unavailable in certain areas, leading to increased computational costs and potential misinterpretation of road topologies, especially in cases involving bike lanes and complex intersections.
Innovation Solution
The use of an ontology enhanced for deriving semantic context allows for the segmentation of road regions in a driving scene without relying on map data. This approach processes images to identify lane lines and road regions, computes pixel depth, and derives three-dimensional context, enabling accurate lane detection and 3D context inference.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If map data is used for lane detection, then accuracy of road attribute detection is improved, but computational costs and system complexity increase
Solution Approach 1:
The patent extracts and removes the dependency on map data from the lane detection system. Instead of using complex HD map data, the system processes only sensor data (camera and LIDAR) to detect lane lines and road attributes, thereby reducing system complexity while maintaining detection accuracy through improved semantic segmentation and 3D context inference
Solution Approach 2:
The system creates a virtual 3D representation of the road scene by inferring depth and spatial relationships from 2D sensor images. This virtual copy replaces the need for actual map data, allowing the system to achieve accurate road attribute detection without the computational burden of processing complex map information
2Measurement precision
If HD map data is used for lane detection, then accuracy of road attribute detection is improved, but computational costs increase
Solution Approach 1:
The system replaces expensive HD map data with inexpensive real-time sensor processing. Instead of relying on pre-computed high-definition maps that require significant computational resources to process and match, the system uses readily available camera and LIDAR data that can be processed more efficiently, reducing computational costs while maintaining detection accuracy
Solution Approach 2:
The system performs preliminary processing of sensor data by inferring 3D context and semantic meanings from 2D images before lane detection. This preliminary action of creating depth maps and semantic segmentations prepares the data in advance, making subsequent lane detection more efficient and reducing the computational burden that would otherwise be required to process HD map data
3Measurement precision
If map data is used for lane detection, then accuracy of road attribute detection is improved, but system availability decreases due to map unavailability in certain areas
Solution Approach 1:
The system makes itself independent of external map data by using its own sensors (camera and LIDAR) to detect and infer all necessary road attributes. This self-service capability allows the system to operate autonomously in areas where map data is unavailable, outdated, or inaccurate, thereby improving system availability and reliability without sacrificing detection accuracy
Solution Approach 2:
Instead of relying on pre-existing map data to guide detection, the system inverts the approach by using sensor data to create its own understanding of the road environment in real-time. This inversion eliminates the dependency on map availability and allows the system to adapt to any geographic location, improving both availability and accuracy
4Speed
If conventional lane detection systems are used, then processing speed is maintained, but accuracy of semantic identification decreases
Solution Approach 1:
The system transitions from 2D image processing to 3D context inference by estimating depth information from monocular camera images and integrating it with LIDAR data. This dimensional enhancement allows for more accurate semantic identification of road regions and lane lines while maintaining processing speed through efficient depth estimation algorithms and optimized neural network architectures
Data Source
AI summary
Systems, methods, and other embodiments described herein relate to the detection of lanes in a driving scene through segmenting road regions using an ontology enhanced to derive semantic context. In one embodiment, a method includes segmenting an image of a driving scene, independent of maps, by lane lines and road regions defined by an ontology and a pixel subset from the image has semantics of lane information from the ontology. The method also includes computing pixel depth from the image for the lane lines and the road regions using a model. The method also includes deriving 3D context using relations between the semantics and the pixel depth, the relations infer a driving lane for a vehicle from types of the lane lines and the road regions adjacent to the driving lane. The method also includes executing a task to control the vehicle on the driving lane using the 3D context.