Two-Stage Object Tracking for Linear and Complex Motion

Solution Overview

Problem

Existing object tracking techniques face challenges in balancing between simple linear motion and complex dynamic motion, particularly in scenarios involving cars and pedestrians, with filtering-based approaches struggling in non-linear scenarios and transformer-based approaches requiring excessive computational resources.

Innovation Solution

A two-stage object tracking system utilizing a gradient-boosting ensemble approach, combining a filtering-based first stage and a transformer-based second stage, where the first stage handles linear motion and the second stage addresses complex motion, with a non-maximum suppression method to refine bounding boxes and tracklets.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If transformer-based approaches are used for object tracking, then tracking accuracy in complex dynamic scenarios is improved, but computational resource consumption increases excessively

Engineering Contradiction:

Improvetracking accuracyVSAvoidcomputational resource consumption

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

The tracking system is divided into two independent stages: a filtering-based first stage for linear motion and a transformer-based second stage for complex dynamic motion. This segmentation allows each stage to use the most appropriate algorithm for its specific task, avoiding the need to deploy computationally intensive transformer models in all scenarios while maintaining high tracking accuracy when needed.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system dynamically switches between different tracking algorithms based on the motion characteristics of detected objects. The gradient-boosting ensemble classifier determines whether an object exhibits linear or complex dynamic motion, enabling the system to adaptively select the appropriate tracking stage and optimize computational resource allocation according to actual tracking needs.

Inventive Principle:

Principle #15Dynamics

2Productivity

If filtering-based approaches are used for object tracking, then computational efficiency is improved, but tracking accuracy deteriorates in non-linear scenarios

Engineering Contradiction:

Improvecomputational efficiencyVSAvoidtracking accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The tracking system is divided into two independent stages: a filtering-based first stage for linear motion and a transformer-based second stage for complex dynamic motion. This segmentation allows each stage to use the most appropriate algorithm for its specific task, avoiding the need to deploy computationally intensive transformer models in all scenarios while maintaining high tracking accuracy when needed.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

A gradient-boosting ensemble classifier serves as an intermediary between the filtering-based first stage and the transformer-based second stage. This intermediary evaluates motion characteristics and determines whether to escalate to the more accurate but computationally intensive second stage, enabling the system to maintain high accuracy in complex scenarios while preserving computational efficiency in simpler cases.

Inventive Principle:

Principle #24Intermediary (Mediator)

3Device complexity

If a single-stage tracking system is used, then device complexity is reduced, but the system cannot effectively handle both linear and complex dynamic motion

Engineering Contradiction:

Improvesystem structureVSAvoidmotion handling capability

Core Design Contradiction:

Device complexityVSAdaptability or versatility

Solution Approach 1:

The tracking system is divided into two independent stages: a filtering-based first stage for linear motion and a transformer-based second stage for complex dynamic motion. This segmentation allows each stage to use the most appropriate algorithm for its specific task, avoiding the need to deploy computationally intensive transformer models in all scenarios while maintaining high tracking accuracy when needed.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The gradient-boosting ensemble classifier provides universal functionality by handling both linear and complex dynamic motion classification. This intermediary component enables the system to adaptively select appropriate tracking strategies for different motion types, giving the overall system the versatility to handle diverse scenarios while maintaining relatively simple device architecture through conditional routing.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Data Source

PatentUS20260051066A1Tracking objects

Publication Date: 2026.02.19 QUALCOMM INC

AI summary

Systems and techniques are described herein for tracking objects. For instance, a method for tracking objects is provided. The method may include generating features based on a sensor-data frame; detecting an object based on the features; generating a bounding box based on the object; tracking the bounding box over a plurality of sensor-data frames to generate a tracklet, wherein the tracklet comprises a respective bounding box for each sensor-data frame of the plurality of sensor-data frames and an identifier; and combining the bounding box and a bounding box of the tracklet to generate an output bounding box.