Object Tracking Kinematic Model Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle detection systems, such as those using Hough match methods and parametric L-shape methods, face limitations in accurately describing the size and shape of objects, are computationally expensive, and struggle with stability, especially for curved profiles and non-rectangular objects.
Innovation Solution
A computerized object-tracking system that generates kinematic models by processing scan data from sensors like rangefinders, using a data collection, clustering, data association, and tracking module to identify and track objects with improved accuracy and stability, directly computing kinetic parameters from scanning points.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Hough match methods are used for object detection, then object profiles can be generated, but the computational cost increases and the resolution becomes discrete and fixed
Solution Approach 1:
The patent segments the object detection process into distinct modules: data collection module receives scan data, clustering module groups points into clusters, data association module links clusters to objects, and tracking module generates kinematic models. This segmentation allows each module to specialize in one aspect, improving overall precision while managing computational complexity through distributed processing.
Solution Approach 2:
The patent transitions from 2D scan data to 3D kinematic models by adding temporal dimension (tracking over time) and spatial dimension (clustering points in 3D space). This dimensional expansion enables more accurate object representation while the modular architecture manages the increased computational requirements.
2Adaptability or versatility
If parametric L-shape methods are used for object tracking, then rectangular objects can be detected, but the method becomes unstable and fails for curved profiles
Solution Approach 1:
The patent implements dynamic tracking by continuously updating kinematic models over time. The tracking module generates new kinematic models for each new frame and associates them with existing objects, allowing the system to adapt to moving objects with changing orientations and positions. This dynamic approach replaces the static L-shape method, providing stability for curved profiles while maintaining versatility.
Solution Approach 2:
The patent changes the fundamental parameters used for object representation from fixed L-shape parameters to dynamic kinematic model parameters including position, velocity, acceleration, and orientation. This parameter transformation enables accurate tracking of objects with any profile shape, particularly curved profiles, while maintaining tracking stability through continuous parameter updates.
3Productivity
If discrete resolution profiles are used for object representation, then computation is simplified, but the size and shape description becomes inaccurate
Solution Approach 1:
The patent performs preliminary clustering of scan points into groups before detailed object representation. The clustering module pre-processes the scan data by grouping points that belong to the same object, which simplifies subsequent processing while preserving accurate size and shape information. This preliminary action maintains productivity by reducing the complexity of later stages.
Solution Approach 2:
The patent creates simplified kinematic model copies that represent the essential characteristics of objects without requiring full discrete profile data. The tracking module generates compact kinematic models containing position, velocity, acceleration, and orientation parameters, which copy the essential motion characteristics while enabling faster processing compared to detailed discrete profiles.
Data Source
AI summary
A method for tracking an object, by generating kinematic models corresponding to the object using a computerized object-tracking system, is disclosed. The method includes receiving, by a data collection module, scan data associated with an object, and generating by the module, using the scan data, a new frame F, associated with a new time t+dt and including new points X. A clustering module identifies a new group G of new points Xi of the new frame F. A data association module associates the new group G with the object based on previous information associated with the object. A tracking module determines a new kinematic model M corresponding to the object based on the new group G.