Virtual Point Fusion for Sparse LiDAR Mobility Tracking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Lidar systems provide sparse point cloud measurements, making it difficult to accurately determine mobility parameters of objects, such as orientation and velocity, which can lead to suboptimal tracking performance in autonomous systems.
Innovation Solution
A multi-modal fusion pipeline system that generates virtual points and associates them with lidar points to create a denser point cloud, allowing for more accurate determination of mobility parameters by predicting object positions in earlier semantic images and generating an object sequence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Lidar systems are used to detect objects, then object detection capability is provided, but measurement precision of mobility parameters deteriorates due to sparse point cloud measurements
Solution Approach 1:
The patent introduces virtual points as an intermediary element between the sparse Lidar point cloud and the object mobility parameters. These virtual points are generated based on the spatial relationship between multiple Lidar points and the object, effectively mediating the information gap caused by sparse measurements. The virtual points serve as additional measurement data that improves mobility parameter estimation without requiring denser physical Lidar point cloud coverage.
2Reliability
If sparse Lidar point cloud measurements are used, then device complexity is reduced, but tracking performance deteriorates due to insufficient mobility parameter data
Solution Approach 1:
The patent performs preliminary action by pre-calculating virtual points based on the current Lidar point cloud data and object position. These virtual points are computed in advance before the actual mobility parameter estimation is performed, allowing the system to have ready-to-use enhanced measurement data. This preliminary computation of virtual points improves tracking reliability without requiring complex real-time processing during the measurement phase.
3Measurement precision
If virtual points are generated and associated with Lidar points, then measurement precision of mobility parameters is improved, but device complexity increases due to multi-modal fusion pipeline
Solution Approach 1:
The patent creates copies of the Lidar point cloud data by generating virtual points that replicate and extend the information from existing Lidar points. Instead of requiring a completely new sensing modality or complex processing architecture, the system generates synthetic copies of measurement data through mathematical relationships. This copying approach improves measurement precision while keeping the processing system relatively simple, as it reuses existing Lidar data in transformed form.
Data Source
AI summary
A system can use semantic images, lidar images, and/or 3D bounding boxes to determine mobility parameters for objects in the semantic image. In some cases, the system can generate virtual points for an object in a semantic image and associate the virtual points with lidar points to form denser point clouds for the object. The denser point clouds can be used to estimate the mobility parameters for the object. In certain cases, the system can use semantic images, lidar images, and/or 3D bounding boxes to determine an object sequence for an object. The object sequence can indicate a location of the particular object at different times. The system can use the object sequence to estimate the mobility parameters for the object.