Drivable Road Segment Annotation Using Segmentation Mask Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing methods for annotating drivable road segments in maps are time-consuming and require significant resources, both in terms of time and computing power, making them inefficient and labor-intensive.
Innovation Solution
A multi-stage annotation system that uses sensor data to generate maps, divides them into sub-maps, infers segmentation masks for drivable road segments, filters and smooths them to produce continuous annotations, leveraging machine learning algorithms and cloud computing for efficient and accurate annotation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual user input is used to annotate drivable road segments, then annotation accuracy can be maintained, but the process becomes time-consuming and resource-intensive
Solution Approach 1:
The system enables maps to annotate themselves by automatically generating segmentation masks from sensor data collected during autonomous vehicle operation. The vehicle's own sensors and processing systems are used to create annotated maps without external human intervention, transforming the mapping process into a self-service operation that eliminates manual annotation time while maintaining accuracy through multiple processing stages including filtering and smoothing
Solution Approach 2:
The patent replaces the mechanical process of manual user input with an automated computer vision system that uses sensor data, machine learning models, and image processing algorithms to generate segmentation masks. This substitution transforms the annotation process from a labor-intensive manual operation to an automated computational process that operates continuously without human intervention
2Manufacturing precision
If manual user input is used to annotate drivable road segments, then annotation quality can be ensured, but computing and financial resources are significantly consumed
Solution Approach 1:
The system divides the annotation process into distinct stages: generating initial segmentation masks from sensor data, filtering masks through multiple criteria (drivability, continuity, connectivity), and smoothing to produce final high-quality annotations. This segmentation of the annotation process into manageable stages reduces computational complexity at each step while maintaining overall annotation quality through cumulative refinement
Solution Approach 2:
The system performs preliminary filtering and validation of segmentation masks before final annotation generation. By pre-processing the raw segmentation data through drivability checks, continuity verification, and connectivity analysis, the system eliminates low-quality candidates early in the process, reducing the computational burden on subsequent processing stages and ensuring only high-quality annotations proceed to final output
3Productivity
If automated methods are used for annotation, then time and resources are reduced, but ensuring continuous and accurate drivable road segment identification becomes more difficult
Solution Approach 1:
The system incorporates multiple feedback mechanisms including drivability checks that verify whether identified segments are actually traversable, continuity verification that ensures road segments form unbroken paths, and connectivity analysis that confirms proper linkage between segments. These feedback loops continuously validate and refine the automated annotation output, ensuring high reliability while maintaining productivity
Solution Approach 2:
The annotation system dynamically adjusts processing parameters and filtering criteria based on the specific characteristics of each map region and sensor data quality. The system adapts its segmentation and filtering strategies in real-time to handle varying road conditions, terrain types, and environmental factors, maintaining both efficiency and accuracy across diverse scenarios
Data Source
AI summary
Provided are techniques for automatic annotation of drivable road segments, including but not limited to: receiving sensor data, generating a map, dividing the map into disjoint sub-maps, inferring sub-map segmentation masks, constructing an inferred segmentation mask, filtering the inferred segmentation mask, smoothing the filtered segmentation mask, planning a path for a vehicle and controller the vehicle.