Group Object Tracking via Feature Clustering and Similarity Transformations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle navigation systems face challenges in accurately tracking groups of pedestrians, particularly when sub-groups move in different directions and speeds, requiring significant computing resources and leading to inaccurate tracking due to partial occlusion.
Innovation Solution
The method involves identifying object features using ORB detectors, forming clusters based on pixel distances, and applying similarity transformations to determine inlier features, with tracking data provided to vehicles using k-nearest neighbor, particle filter, or Kalman filter algorithms to predict object locations and avoid collisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If individual pedestrians in a group are tracked separately, then tracking precision may be maintained, but computing resources required increase significantly
Solution Approach 1:
The patent merges multiple pedestrian targets into a single group object for tracking purposes. By detecting common motion patterns and spatial relationships among multiple pedestrians, the system treats them as one unified target, thereby reducing the computational burden of tracking each individual separately while maintaining adequate tracking precision for safety-critical applications.
Solution Approach 2:
The patent segments the tracking process into two levels: group-level tracking for the entire pedestrian cluster and individual-level detection only when necessary. This hierarchical segmentation allows the system to maintain lower computational complexity by default while preserving the ability to resolve individual pedestrians when occlusion or safety concerns arise.
2Measurement precision
If sub-groups moving in different directions are tracked as separate entities, then tracking accuracy improves, but the complexity of determining group structure increases
Solution Approach 1:
The patent implements dynamic group structure determination by continuously monitoring motion patterns and spatial relationships. The system adapts the grouping configuration in real-time based on detected motion coherence, automatically separating sub-groups when different motion patterns are identified while maintaining a unified tracking framework. This dynamic approach balances tracking accuracy with computational complexity by only performing detailed structure analysis when necessary.
3Measurement precision
If traditional object detection methods are used for groups of pedestrians, then individual detection may be achieved, but partial occlusion leads to highly inaccurate tracking
Solution Approach 1:
The patent merges the visual information of partially occluded pedestrians into a unified group representation. By detecting the collective motion pattern and spatial coherence of the pedestrian group, the system can track the group center and trajectory even when individual pedestrians are occluded, thereby maintaining tracking accuracy despite partial occlusion of group members.
Data Source
AI summary
A system, including a processor and a memory, the memory including instructions to be executed by the processor to identify first object features from sensor data acquired by a stationary sensor at a first time step, determine second object features at a second time step. The instructions can include further instructions to determine one or more object clusters of first object features by determining distances measured in pixels between the first object features and corresponding second object features and comparing the distances to one or more mean distances and determine one or more object groups of inlier first object features in the one or more object clusters by determining a plurality of similarity transformations for a plurality of random samples of first object features and determining inlier first object features based on maximizing the number of first object features included in a similarity transformation.