Lane Boundary Detection Using Polyline Clustering Across Frames
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for detecting road and lane boundaries in autonomous vehicles are prone to errors due to occlusions, noise in sensor data, and complex road topologies, leading to inaccurate delineation of boundaries and failure in handling forks and joins.
Innovation Solution
A system that determines road and lane boundaries by computing pair-wise distance measures for polylines based on similarity, curvature difference, and perpendicular distance, and clusters these polylines to accurately identify and track boundaries across multiple frames, using sensors like cameras and LiDAR.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional boundary detection methods are used, then the system is simple to implement, but the detection accuracy deteriorates due to occlusions, noise, and complex road topologies
Solution Approach 1:
The patent segments the boundary detection problem into multiple stages: initial boundary detection, polyline generation, pair-wise distance computation, and clustering. Each stage processes specific aspects of the detection task, allowing complex operations to be broken down into manageable components that can be implemented systematically while maintaining high accuracy.
Solution Approach 2:
The patent introduces pair-wise distance measures that compute relationships between boundaries in multiple dimensions (spatial distance, curvature difference, similarity metrics). This multi-dimensional approach allows the system to distinguish between true boundaries and false detections by evaluating them from multiple geometric perspectives simultaneously, significantly improving detection accuracy in complex scenarios.
2Measurement precision
If pair-wise distance measures and clustering are computed for all polylines, then the accuracy of boundary detection is improved, but the computational time and resources increase
Solution Approach 1:
The patent applies clustering algorithms selectively rather than uniformly to all detected polylines. By identifying candidate boundaries first and then applying computationally intensive pair-wise distance measures and clustering only to relevant subsets, the system achieves high detection accuracy while avoiding unnecessary computational overhead on clearly valid or obviously incorrect boundaries.
Solution Approach 2:
The patent computes pair-wise distance measures with different weights and criteria for different regions and types of boundaries. For example, curvature difference may be weighted more heavily for curved road sections, while perpendicular distance may be more important for straight sections. This localized adaptation of computation intensity optimizes the balance between accuracy and processing time for different road scenarios.
3Adaptability or versatility
If the system processes multiple polylines and performs clustering, then the ability to handle complex road scenarios (forks and joins) is improved, but the device complexity increases
Solution Approach 1:
The patent performs preliminary organization of detected boundaries into polylines before applying clustering algorithms. By pre-processing the boundary data to group collinear or near-collinear segments into candidate polylines, the system creates a structured input for the clustering stage that simplifies the handling of complex scenarios like forks and joins, reducing the computational complexity of the overall system.
Solution Approach 2:
The patent employs iterative clustering where the results from pair-wise distance computation feed back into the clustering process, allowing the system to refine its understanding of boundary relationships. This feedback mechanism enables the system to correctly identify forks and joins by repeatedly evaluating boundary relationships and adjusting cluster assignments based on accumulated evidence from multiple distance measurements.
Data Source
AI summary
A computing system may determine, for a plurality of polylines associated with one or more pictures of a video, a plurality of pair-wise distance measures for pairs of polylines from the plurality of polylines. The computing system may cluster, based on the plurality of pair-wise distance measures, a subset of polylines from the plurality of polylines to generate a polyline cluster. The computing system may determine a lane boundary or a road boundary that corresponds to the polyline cluster and may output information indicative of the lane boundary or the road boundary.