Cluster Trajectory Orientation for Non-Point Target Velocity Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional object detection and tracking systems using high-resolution radar or LIDAR face inaccuracies when dealing with non-point objects due to multiple reflections from different parts, leading to biased velocity estimation.
Innovation Solution
A system and method that uses a cluster trajectory orientation process with principal component analysis to estimate the heading and velocity of moving targets by generating a position data matrix, computing average coordinates, and identifying the major eigenvector direction to determine the heading, combining it with other velocity estimation methods for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a point object model is used for tracking, then the system is simple to implement, but velocity estimation becomes biased when tracking non-point objects
Solution Approach 1:
The patent segments the non-point object into multiple detected points or clusters of reflections, treating each reflection as a separate detection. By analyzing the spatial distribution and temporal evolution of these segmented points, the system can estimate the object's velocity more accurately without requiring a complex physical model of the entire object.
Solution Approach 2:
The patent transitions from treating the target as a single-point entity to analyzing it in multiple dimensions by considering the spatial distribution of multiple reflections across different radar beams. This dimensional expansion allows the system to extract velocity information from the relative motion patterns of different reflection points, improving accuracy for extended objects.
2Loss of information
If multiple reflections from different parts of a non-point object are detected, then more information about the object is obtained, but velocity estimation accuracy deteriorates due to bias introduction
Solution Approach 1:
The patent employs feedback mechanisms where the detected cluster of reflections is continuously tracked across multiple time steps. The system uses the temporal evolution of the reflection cluster's spatial distribution to refine velocity estimates, feeding back this information to correct for biases introduced by detecting multiple reflections from different object parts.
3Productivity
If conventional point-object tracking is used, then processing is computationally efficient, but tracking accuracy deteriorates for large objects like trucks
Solution Approach 1:
The patent applies partial action by focusing computational resources on analyzing only the essential characteristics of the reflection cluster (such as the centroid position and spread) rather than fully modeling every aspect of the extended object. This approach maintains computational efficiency while capturing sufficient information to improve velocity estimation accuracy for large objects.
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
This approach provides robust and accurate object velocity estimation, reducing interference and improving tracking performance by approximating the target shape and aligning with the object's orientation, while being less affected by outlying data and partial occlusions.
Implementation Method 1
a receiver for receiving reflected signals generated by reflection of the transmitted signals from a plurality of points on the moving target
Implementation Method 2
a second velocity generated from a Doppler-azimuth profile process
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system for characterizing a moving object performs a cluster trajectory orientation process associated wi th clusters of detected points in each of a set. of scans to estimate the heading of a non-point target. The cluster txajectory orientation process performs a principal component analysis on a corresponding position data matrix representing coordinates of the clusters of points for each of a set of scans and compares resulting eigenvec tors to a heading of the cluster averages to generate a heading estimate. The heading estimate is combined with velocity estimates from a point target based tracking process and a Doppler-azinnith profile process in a weighted combination based on target attributes to improve the accuracy and performance of the system.