Automated Image Annotation via Map Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manual annotation of images for autonomous vehicles is time-consuming and prone to human error, requiring efficient and accurate methods for annotating static and dynamic objects in complex environments.
Innovation Solution
The method leverages state-of-the-art vehicle localization techniques and high-definition maps to automate or semi-automate image annotation by projecting object locations from a map into images, using ego localization data and object projections to annotate known objects with high precision, allowing for minimal manual correction and interpolation/extrapolation across image sequences.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual annotation is used to annotate images with object locations, then annotation accuracy can be maintained through human review, but annotation time and labor costs increase significantly
Solution Approach 1:
The system performs preliminary localization of the vehicle on the road map and pre-computes object projections into image planes before annotation is needed. This preliminary action provides pre-positioned annotation data that can be directly used or minimally adjusted, eliminating the need for time-consuming manual annotation while maintaining accuracy through the precision of localization and projection calculations
Solution Approach 2:
The system creates accurate copies of object locations from the map frame of reference and projects them into the image plane frame of reference. This copying process generates annotation data that replicates ground truth information without requiring manual creation, thus reducing annotation time while preserving accuracy through mathematical projection transformations
2Productivity
If automated annotation using localization and projection is used, then annotation speed increases significantly, but annotation accuracy may deteriorate due to computational errors
Solution Approach 1:
The system uses feedback from localization data quality assessment to determine whether automated projection is sufficient or if manual verification is needed. By monitoring the reliability of ego localization data and comparing projected annotations against expected patterns, the system can identify and correct computational errors, maintaining accuracy while preserving the speed benefits of automation
Solution Approach 2:
The system applies partial manual correction only where needed rather than requiring complete manual annotation. By using automated projection for the majority of annotations and applying human review only to cases where localization uncertainty or projection errors are detected, the system achieves high annotation speed while maintaining accuracy through targeted manual intervention
3Measurement precision
If high-definition maps with centimetre-level accuracy are used for annotation, then annotation precision improves, but system complexity and computational requirements increase
Solution Approach 1:
The system extracts only the necessary localization information and object projection data from high-definition maps, rather than processing the entire map dataset. By selecting and using only the relevant centimetre-level accuracy data needed for specific annotation tasks, the system achieves high annotation precision while reducing computational complexity through targeted data extraction
Solution Approach 2:
The annotation process is segmented into distinct computational stages: localization in map frame, object identification, projection into image plane, and validation. This segmentation allows each stage to be optimized independently and processed efficiently, reducing overall system complexity while maintaining the precision benefits of high-definition maps through structured modular computation
Data Source
AI summary
A method of annotating known objects in road images captured from a sensor-equipped vehicle, the method implemented in an annotation system and comprising: receiving at the annotation system a road image containing a view of a known object; receiving ego localization data, as computed in a map frame of reference, via localization applied to sensor data captured by the sensor-equipped vehicle, the ego localization data indicating an image capture pose of the road image in the map frame of reference; determining, from a predetermined road map, an object location of the known object in the map frame of reference, the predetermined road map representing a road layout the map frame of reference, wherein the known object is one of: a piece of road structure, and an object on or adjacent a road; computing, in an image plane defined by the image capture pose, an object projection, by projecting an object model of the known object from the object location into the image plane; and storing, in an image database, image data of the road image, in association with annotation data of the object projection for annotating the image data with a location of the known object in the image plane.