3D Point Cloud Reference Tracking for Occluded Object Motion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in accurately identifying and tracking objects over time, especially when objects are partially occluded or move out of the sensor's field of view, which impairs estimation of speed and heading.
Innovation Solution
A method that involves determining stable reference points by matching 3D point clouds captured at different times using a LIDAR device, allowing for the estimation of object motion characteristics even when reference points are occluded, by projecting reference points from one view to another based on transformations calculated between consecutive point clouds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If autonomous vehicles use standard object tracking methods, then tracking is simple when objects are fully visible, but tracking accuracy deteriorates when objects are partially occluded or move out of field of view
Solution Approach 1:
The patent transitions from 2D image processing to 3D point cloud processing by using LIDAR data. This dimensional change allows the system to capture spatial information in three dimensions, enabling more robust tracking of objects even when partially occluded in any single view, as the 3D structure provides additional geometric constraints for reconstruction
Solution Approach 2:
The system performs preliminary action by capturing multiple 3D point clouds at different time instances before the object becomes completely occluded or moves out of view. By accumulating temporal data points in advance and using transformation calculations between consecutive point clouds, the system prepares reference information that enables accurate tracking even when direct observation is temporarily lost
2Measurement precision
If the vehicle uses multiple sensors to improve detection accuracy, then measurement precision improves, but device complexity increases
Solution Approach 1:
The LIDAR sensor performs multiple functions: it captures 3D spatial information, provides temporal sequences of point clouds, and enables both object detection and motion characteristic estimation simultaneously. This multi-functionality reduces the need for separate specialized sensors while maintaining high measurement precision through a single versatile sensing system
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables accurate estimation of heading and velocity of objects, improving situational awareness and navigation safety for autonomous vehicles by accounting for partial occlusions and changing views.
Implementation Method 1
A method that involves determining stable reference points by matching 3D point clouds captured at different times using a LIDAR device
Data Source
AI summary
A computing device may identify an object in an environment of a vehicle and receive a first three-dimensional (3D) point cloud depicting a first view of the object. The computing device may determine a reference point on the object in the first 3D point cloud, and receive a second 3D point cloud depicting a second view of the object. The computing device may determine a transformation between the first view and the second view, and estimate a projection of the reference point from the first view relative to the second view based on the transformation so as to trace the reference point from the first view to the second view. The computing device may determine one or more motion characteristics of the object based on the projection of the reference point.