Vehicle Object Tracking Using Space-Time Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing driver assistance systems face challenges in efficiently associating newly detected objects with tracked objects due to computation resource limitations and modeling assumptions, leading to suboptimal tracking performance.
Innovation Solution
A method and device for tracking objects using a space-time representation that considers both spatial and temporal coordinates, employing a distance metric that accounts for the ego vehicle's movement, and utilizes Kalman filtering to associate and update object positions, enabling robust object tracking.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If filtering techniques like Kalman filtering are used to associate objects, then tracking stability is improved, but computation resources are consumed
Solution Approach 1:
The patent changes the parameter space from purely spatial coordinates to space-time coordinates by adding temporal information. This allows the system to associate objects based on both spatial proximity and temporal continuity, improving tracking stability while enabling more efficient computation by reducing the need for complex multi-sensor fusion operations.
Solution Approach 2:
The patent introduces a new dimension (time) to the object association problem. By representing objects in space-time coordinates rather than just spatial coordinates, the system can leverage temporal information to improve association accuracy. This dimensional extension allows for simpler association logic that achieves better tracking performance without proportionally increasing computational complexity.
2Measurement precision
If multiple sensors of different types are used, then detection capability is improved, but system complexity increases
Solution Approach 1:
The patent creates a universal space-time coordinate representation that can accommodate data from multiple sensor types (radar, camera, lidar, infrared). By transforming all sensor measurements into a common space-time framework, the system achieves multi-functionality without requiring separate processing pipelines for each sensor type, thereby reducing overall system complexity.
Solution Approach 2:
The patent merges the processing of multiple sensor types into a unified space-time association framework. Instead of handling radar, camera, and lidar data separately, the system combines them into a single coordinate system and uses a unified association algorithm, simplifying the architecture while maintaining the detection advantages of multiple sensors.
3Ease of operation
If dynamic entities with predefined lifespan are created for tracking, then association problem is addressed, but tracking accuracy decreases due to modeling assumptions
Solution Approach 1:
The patent changes the parameters used for object representation from static spatial coordinates to dynamic space-time coordinates. This allows the system to track objects based on their temporal behavior patterns rather than relying on predefined lifespans, improving tracking accuracy by adapting to actual object motion patterns without artificial constraints.
Solution Approach 2:
The patent introduces dynamic adaptability to the tracking system by using space-time coordinates that automatically adjust to object motion characteristics. Instead of using fixed predefined lifespans, the system dynamically determines object persistence based on temporal continuity in the space-time representation, improving accuracy while maintaining ease of operation.
Data Source
Figure 1~2
Figure 3
AI summary
The invention provides a device for tracking objects in a surrounding of a vehicle, comprising an interface configured to receive, from at least one vehicle sensor of the vehicle, newly acquired measurement data. The device further comprises a computing device. The computing device is configured to determine time coordinates and spatial coordinates of newly measured objects in a surrounding of the vehicle based on the newly acquired measurement data, wherein the spatial coordinates of the newly measured objects are defined relative to a global coordinate system. The computing device is further configured to associate each newly measured object with at most one tracked object of a plurality of tracked objects based on distances between the newly measured object and the plurality of tracked objects, wherein the distance depends on both a difference between the spatial coordinates of the newly measured object and the tracked object and a difference between the time coordinates of the newly measured object and the tracked object. The computing device updates spatial coordinates of the tracked objects using a Kalman filtering method, using the associated newly measured objects.