LiDAR Object Tracking with Cluster Weighting for Irregular Shapes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing object tracking methods using LiDAR systems are destabilized by irregular measurements, particularly from protruding parts like side mirrors and towbars, leading to non-optimal matching and tracking issues.
Innovation Solution
The method involves determining bounding data by grouping target data into clusters and applying weighting values based on criteria such as number, spatial extent, alignment, and statistical methods to correct for irregular shapes, thereby stabilizing the tracking process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If irregular measurements from protruding parts like side mirrors and towbars are included in object tracking, then the completeness of object detection is improved, but the stability of tracking is worsened
Solution Approach 1:
The patent segments the detected object into multiple clusters based on spatial proximity and characteristics of measurement points. By dividing the object into clusters, the system can identify and separate protruding parts (like side mirrors and towbars) from the main object body, allowing selective weighting to stabilize tracking while preserving complete object information.
Solution Approach 2:
The patent applies different weighting values to different clusters of measurement points based on their local characteristics. Clusters representing protruding parts receive lower weights while clusters representing the main object body receive higher weights. This local differentiation stabilizes tracking by reducing the influence of irregular measurements while maintaining detection completeness.
2Measurement precision
If all target data is used for determining bounding data, then the accuracy of object representation is improved, but the robustness to irregular shapes is worsened
Solution Approach 1:
The patent performs preliminary clustering and weighting of measurement points before determining the bounding box. By pre-processing the target data to identify and weight clusters appropriately, the system prepares the data in advance to reduce the influence of irregular shapes, thereby improving robustness while maintaining accuracy in the final bounding data determination.
Solution Approach 2:
The patent changes the parameter weights assigned to different measurement point clusters based on their characteristics. By dynamically adjusting weighting values for different clusters (representing different parts of the object), the system optimizes the balance between bounding data accuracy and robustness to irregular shapes like side mirrors and towbars.
3Reliability
If weighting values are applied to target data clusters, then the stability of tracking is improved, but the complexity of data processing is worsened
Solution Approach 1:
The patent segments measurement points into clusters based on spatial and characteristic similarity, which simplifies the weighting process by treating each cluster as a unit. This segmentation reduces the overall complexity compared to individually weighting each measurement point, while still achieving tracking stability through cluster-level weighting.
Solution Approach 2:
The system automatically determines weighting values for each cluster based on intrinsic characteristics of the measurement data itself, such as cluster size, density, and spatial distribution. This self-service approach to weighting reduces the need for manual parameter tuning and external complexity, achieving tracking stability through data-driven automatic weighting.
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 enhances the robustness of object tracking by directly correcting measurement irregularities, reducing the influence of protruding parts, and improving the accuracy of tracking and segmentation.
Implementation Method 1
means for sending electromagnetic scanning signals, means for receiving reflected electromagnetic scanning signals
Data Source
AI summary
A method for tracking of an object with a detection device, using electromagnetic scanning signals, a plurality of target data is determined, where each target data comprises a position value. From at least a part of the target data a bounding data is determined. The object is tracked by determining a development of the bounding data. Where for determining the bounding data at least a part of the target data is grouped into clusters, a weighting value for the target data of at least one cluster is determined from the target data of the respective cluster, the bounding data is determined from weighted target data, where the weighted target data each are determined from the respective target data weighted by the weighting value of the cluster to which the target data belongs.