Multi-Sensor Object Tracking With Trace-Based Cluster Association
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for object detection and tracking in vehicles using multiple perception sensors face challenges such as duplicate detection, ghost objects, difficulty in understanding object shape and size, and increased accident risk due to incomplete tracking and missed object detection.
Innovation Solution
A method involving a control arrangement that processes sensor data from multiple perception sensors to generate clusters at different time points, determines a trace using these clusters, calculates a distance measure, and tracks objects based on this measure to ensure accurate aggregation and reduce computational resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If point cloud-based perception sensors are used to detect objects in the vehicle environment, then a 3D representation of the full environment can be obtained, but duplicate detection and ghost objects may occur due to incomplete aggregation of sensor data from multiple sensors
Solution Approach 1:
The patent merges sensor data from multiple perception sensors by aggregating point clouds and clustering points that belong to the same object across different sensors and time points. This combining approach ensures complete object representation while eliminating duplicate detections and ghost objects through unified object identification.
2Area of stationary object
If multiple perception sensors are mounted on different locations on the vehicle to achieve full environment coverage, then the fields of view become at least partially different, but this increases the complexity of data aggregation and trace determination
Solution Approach 1:
The patent introduces a temporal dimension to the aggregation process by considering sensor data at multiple time points (first, second, and third time points). This allows the system to track object movement trajectories and determine whether points from different sensors with different fields of view belong to the same moving object, simplifying the aggregation complexity.
3Measurement precision
If all points representing a tracked object are considered in the aggregated point cloud, then accurate object tracking can be achieved, but computational resources and processing time increase
Solution Approach 1:
The patent segments the aggregated point cloud into distinct clusters, where each cluster represents points belonging to the same object. By clustering points spatially and temporally, the system efficiently identifies and tracks objects without processing every individual point, reducing computational burden while maintaining tracking precision.
4Ease of manufacture
If conventional aggregation methods are used without considering object movement between time points, then processing is simpler, but object shape and size understanding becomes difficult and tracking accuracy decreases
Solution Approach 1:
The patent performs preliminary actions by determining the trace (movement trajectory) of objects between the first and second time points before aggregating data at the third time point. This preliminary trace determination enables the system to correctly associate points from different time points with the same moving object, improving shape and size understanding while maintaining implementation feasibility.
Data Source
AI summary
A method for tracking an object using sensors by: obtaining from a first sensor first sensor data comprising first and second clusters of data detected at respective first and second times, obtaining second data from a second sensor indicative of a candidate object comprising a third cluster of data detected at a third time between the first and second times, determining a trace using the first and second clusters, calculating a distance between the third cluster of data and the trace, determining that the third cluster of data is indicative of the tracked object if the calculated distance measure is below a distance threshold, and tracking the object using the first, second, and third clusters, or if the third cluster of data is not indicative of the tracked object, tracking the object using only the first and second clusters.