Traffic Light Occlusion Detection Using 3D-to-2D Range Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face inaccuracies in traffic light detection due to occlusions, particularly when occluding objects temporarily block the view of camera sensor systems, leading to erroneous control instructions, and existing three-dimensional sensor systems have slower processing times compared to two-dimensional systems.
Innovation Solution
A traffic light detection system that converts three-dimensional point clouds into two-dimensional range images with distance information, using an occlusion polygon to exclude occluding objects and determine the traffic light configuration, allowing for real-time control of vehicle systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a three-dimensional sensor system (lidar) is used to detect occlusions, then measurement precision is improved, but processing time increases causing lag
Solution Approach 1:
The patent segments the three-dimensional sensor data into two-dimensional image planes corresponding to the camera's field of view. By processing data in segmented 2D planes rather than full 3D point clouds, the system achieves occlusion detection accuracy comparable to 3D systems while maintaining the faster processing speeds of 2D image processing.
Solution Approach 2:
The patent introduces an intermediary representation called a range image, which serves as a bridge between 3D lidar data and 2D camera images. The range image encodes distance information in a 2D format that can be processed quickly by camera systems while retaining the depth information needed for occlusion detection.
2Productivity
If camera sensor systems are used for traffic light detection, then processing speed is maintained, but reliability deteriorates due to occlusion misclassification
Solution Approach 1:
The patent merges data from two different sensor systems - camera images and lidar range images - into a unified detection framework. By combining the fast processing capability of cameras with the occlusion detection capability of lidar, the system achieves both high processing speed and high reliability in traffic light detection.
Solution Approach 2:
The patent applies local quality enhancement by adding depth information specifically to regions where occlusions are detected in the camera image. Rather than processing entire 3D point clouds, the system selectively enhances depth information only in relevant areas, maintaining fast processing while improving detection reliability where needed.
Data Source
AI summary
An occlusion detection system for an autonomous vehicle is described herein, where a signal conversion system receives a three-dimensional sensor signal from a sensor system and projects the three-dimensional sensor signal into a two-dimensional range image having a plurality of pixel values that include distance information to objects captured in the range image. A localization system detects a first object in the range image, such as a traffic light, having first distance information and a second object in the range image, such as a foreground object, having second distance information. An occlusion polygon is defined around the second object and the range image is provided to an object perception system that excludes information within the occlusion polygon to determine a configuration of the first object. A directive is output by the object perception system to control the autonomous vehicle based upon occlusion detection.