3D Object Detection Fusion for Elevated Roadside Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
LiDAR sensors have a limited vertical field of view, making it challenging to detect objects at higher elevations, such as traffic lights and signs, which is crucial for autonomous vehicles and robotics systems.
Innovation Solution
Integrate point cloud data from LiDAR, camera data, and high-definition map data using a graph neural network to enhance 3D object detection, leveraging spatial priors and temporal context for robust elevated object detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If LiDAR sensors are used for object detection, then spatial precision is improved, but vertical field of view is limited
Solution Approach 1:
The patent combines LiDAR point cloud data with camera image data in a unified 3D detection framework. The LiDAR provides precise spatial measurements while the camera provides wide vertical field of view, and their fusion resolves the contradiction between precision and coverage area.
Solution Approach 2:
The patent introduces camera data as an intermediary to extend the vertical field of view beyond LiDAR's limitations. The camera captures elevated objects that LiDAR cannot detect, and this intermediate data source bridges the gap in vertical coverage.
2Area of moving object
If camera sensors with wide field of view are used, then detection coverage is improved, but spatial precision is reduced
Solution Approach 1:
The patent merges camera data with LiDAR point cloud data, where the camera provides wide field of view coverage and the LiDAR provides precise spatial measurements. The combination allows the system to achieve both wide coverage and high precision simultaneously.
3Measurement precision
If multi-sensor fusion is implemented, then detection accuracy is improved, but system complexity is increased
Solution Approach 1:
The patent segments the detection process into distinct modules: point cloud processing, image processing, and fused detection. This segmentation manages complexity by organizing the multi-sensor fusion into manageable, modular components with clear interfaces.
Solution Approach 2:
The patent creates a unified detection framework that handles both LiDAR point cloud data and camera image data through a single multi-sensor fusion architecture. This universal system processes multiple data types simultaneously, reducing overall system complexity compared to separate processing pipelines.
Data Source
AI summary
Systems and techniques are described for object detection. For example, a device can obtain point cloud data of an environment of a device. The point cloud data includes point cloud(s) obtained using sensor(s) and a respective field of view of each sensor. The device can obtain, from camera sensor(s), camera data of the environment. Each camera sensor includes a respective field of view, where a respective vertical field of view of each camera sensor is greater than a respective vertical field of view of each sensor. The device can obtain map data of the environment that includes one or more spatial priors indicative of at least one of elevated object patterns or locations. The device can determine, using a trained machine learning system, a location of an object based on the point cloud data, the camera data, and the map data.