Voting-Based Triangulation for Urban 3D Object Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for localizing static 3D objects in urban environments face challenges due to the need for manual labeling, which is time and cost-intensive, and the limitations of GPS accuracy in dense urban areas, especially in differentiating similar objects under varying conditions like lighting and weather.
Innovation Solution
A voting-based triangulation technique using 2D image data from multiple cameras to automatically detect and determine the 3D positions of static objects, such as traffic lights, by generating a data set of 2D static object detections and performing distributed voting to recover accurate 3D positions, reducing the reliance on manual labeling and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual labeling is used to annotate map features, then accuracy of semantic components is improved, but time consumption and cost increase significantly
Solution Approach 1:
The system enables automatic self-labeling of map features by using machine learning models that detect and annotate static objects such as traffic lights, road signs, and road markings from image data without requiring manual human intervention, thereby resolving the contradiction between labeling accuracy and time consumption
Solution Approach 2:
The patent replaces the manual mechanical labeling process with automated computer vision and machine learning systems that process image data to automatically identify and annotate semantic features, eliminating the need for human annotators while maintaining or improving accuracy
2Loss of information
If GPS systems are used to provide location information for sensors, then location data is obtained, but accuracy deteriorates in dense urban environments
Solution Approach 1:
The system introduces map features such as road signs, traffic lights, and road markings as intermediary reference points that serve as mediators between GPS systems and the actual location, enabling more accurate positioning by matching detected features with pre-stored map data even when GPS signals are weak or blocked in urban canyons
3Measurement precision
If triangulation techniques are used to determine 3D positions from 2D images, then 3D localization capability is improved, but difficulty in detecting and measuring increases due to similar-looking objects
Solution Approach 1:
The system employs machine learning models that perform multiple functions simultaneously: detecting objects, classifying them by type, and determining their 3D positions, thereby handling the complexity of differentiating similar objects while maintaining triangulation accuracy through integrated multi-functional processing
Data Source
AI summary
The present invention relates to a method and system for automatic localisation of static objects in an urban environment. More particularly, the present invention relates to the use of noisy 2-Dimensional (2D) image data to identify and determine 3-Dimensional (3D) positions of objects in large scale urban or city environments. Aspects and/or embodiments seek to provide a method, system, and vehicle for automatically locating static 3D objects in urban environments by using a voting-based triangulation technique. Aspects and/or embodiments also provide a method for updating map data after automatically new 3D static objects in an environment.