Lane Marking Classification via Cell Segmentation and Marklet Features
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing lane classification methods for self-driving vehicles often fail to accurately distinguish between solid and dashed lane markings, especially when markings are worn, damaged, or occluded, leading to incorrect classifications due to their reliance on continuous parameterized curves that cannot effectively model defects or occlusions.
Innovation Solution
An image processing method that divides candidate lane markings into cells, determines local features of each cell using marklets, aggregates these features to derive global features, and employs a machine learning classifier to classify lane markings as solid or dashed, while being robust to local variations and occlusions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If continuous parameterized curves are used to model lane markings, then the model is simple and computationally efficient, but it cannot effectively model defects or occlusions leading to incorrect classifications
Solution Approach 1:
The patent divides the candidate lane marking into multiple cells along its length. Each cell is independently analyzed to determine local features such as line segment connectivity and length. This segmentation allows the system to handle occlusions and defects by focusing on local characteristics rather than requiring the entire lane marking to fit a continuous curve model, thereby improving classification reliability while maintaining computational efficiency.
2Reliability
If line segment connectivity is checked to classify lane markings, then the method is intuitive and easy to implement, but it fails when lines are worn, damaged or occluded
Solution Approach 1:
The patent evaluates local features within each cell independently, such as the connectivity and length of line segments within that specific cell. By determining an inlier ratio for each cell based on local geometric constraints, the system can identify valid lane marking segments even when other parts are occluded or damaged. This local quality approach makes the detection more robust to defects while managing complexity through localized analysis.
3Measurement precision
If the entire candidate track is analyzed as a single unit, then the processing is simple and fast, but it cannot distinguish local defects from global patterns
Solution Approach 1:
The patent segments the candidate lane marking track into multiple cells and determines local features for each cell independently. This segmentation enables precise measurement of local characteristics such as line segment connectivity and inlier ratios within each cell. The systematic processing of individual cells allows for precise defect detection while maintaining processing efficiency through localized analysis rather than analyzing the entire track as a single unit.
4Reliability
If parameterized curves without length limitations are used, then the model is flexible and adaptable, but it cannot distinguish true lane markings from false ones like guardrails or curbs
Solution Approach 1:
The patent applies local geometric constraints and inlier ratio calculations within each cell to validate whether detected features match expected lane marking characteristics. By checking local properties such as line segment connectivity, length, and orientation within each cell, the system can distinguish true lane markings from false positives like guardrails or curbs. This local validation approach improves reliability in rejecting false positives while managing complexity through localized geometric checks rather than global model validation.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present invention relates to an image processing method for lane classification, comprising: determining a candidate track in an image of a road, wherein the candidate track is modelled as a parameterized line or curve corresponding to a candidate lane marking in the image of a road; dividing the candidate track into a plurality of cells, each cell corresponding to a segment of the candidate track; determining at least one marklet for a plurality of said cells, wherein each marklet of a cell corresponds to a line or curve connecting left and right edges of the candidate lane marking; determining at least one local feature of each of said plurality of cells based on characteristics of said marklets of each respective cell; determining at least one global feature of the candidate track by aggregating the local features of the plurality of cells; and determining if the candidate lane marking represents a lane marking based on the at least one global feature.