Real-Time Moving Object Tracking on Smartphone Using Cradle Head Servo Motor.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Mobile Computing

Background:

• Growing demand for AI-powered smartphone accessories necessitates efficient real-time moving object detection.
• Current tracking systems face challenges with coexisting small/large objects and fast motion, leading to inaccuracies.
• Developing smartphone-specific tracking requires adaptability across operating systems and environments.

Purpose of the Study:

• To develop a novel deep learning-based real-time moving object tracking system for smartphones.
• To address limitations in handling varying object scales and high-speed movements.
• To offer distinct modes optimizing for either accuracy or speed.

Main Methods:

• Proposed a deep learning framework with two modes: accuracy priority and speed priority.
• Accuracy priority mode utilizes CSPNet with ResNet for high precision.
• Speed priority mode simplifies convolutional layers for faster inference, ideal for fast-moving objects.

Main Results:

• The system demonstrated effective real-time moving object tracking capabilities on smartphones.
• Evaluated performance across both accuracy and speed metrics for the two proposed modes.
• The accuracy priority mode balanced high accuracy with necessary speed for smartphone applications.

Conclusions:

• The developed deep learning system offers a viable solution for real-time moving object tracking on smartphones.
• The dual-mode approach provides flexibility to adapt to different tracking scenarios and requirements.
• This work advances mobile AI by improving object detection and tracking efficiency for dynamic environments.