Lane Change Detection via 2D Trajectory and Lane Boundary Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current lane change detection systems in ego vehicles rely on turn signal indicators and 3D lane model projections, which are prone to errors and insufficient for complex driving scenarios, especially when determining lane changes without driver input or accurately predicting vehicle movement.
Innovation Solution
The method involves using an image sensor to detect traffic lane boundaries and project the vehicle's expected trajectory into the same measuring space, comparing these to determine if the vehicle will stay in or change lanes, with distance analysis and potential use of neural networks for accurate lane boundary detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a 3D lane model projection method is used to detect lane changes, then the system can provide lane change detection capability, but the measurement precision deteriorates due to errors in projecting 2D image data to 3D world coordinates
Solution Approach 1:
The patent inverts the traditional approach by not projecting 2D image data to 3D world coordinates, but rather projecting 3D trajectory and lane boundary data into the 2D image coordinate system. This dimensionality inversion eliminates projection errors and allows direct comparison in the image plane, significantly improving measurement precision while maintaining lane change detection capability.
2Loss of information
If turn signal indicator information is used to detect lane changes, then the system can identify driver intention, but the loss of information increases because it cannot determine when or how quickly the vehicle will change lanes
Solution Approach 1:
The system continuously monitors the actual vehicle trajectory and compares it with the expected trajectory in real-time. When deviations indicating lane change intent are detected, the system can determine not only that a lane change is intended but also predict when and how quickly it will occur, providing feedback that eliminates information loss about timing and speed of lane changes.
3Measurement precision
If lane markings detected by image sensor are used to establish lane model, then the system can obtain lane boundary information, but the device complexity increases due to the need for multiple assumptions and model hypotheses
Solution Approach 1:
The patent extracts only the essential lane boundary information directly from the image data without requiring complex 3D model hypotheses about lane width, plane geometry, or other assumptions. By working directly in the image coordinate system, the system takes out the necessary lane boundary detection capability while eliminating the complexity of multiple model assumptions.
Data Source
AI summary
A method for detecting whether an ego vehicle will leave a currently traveled traffic lane of a roadway to the left or right or whether it will stay in the currently traveled traffic lane. In the method, an image of a measuring space, which includes the vehicle area in front of the ego vehicle, is generated using an image sensor; an expected trajectory of the ego vehicle is projected into the image; at least one traffic lane boundary laterally adjacent to the trajectory is detected; and a decision is made whether the traffic lane will be changed or maintained by comparing the trajectory to the at least one detected traffic lane boundary.