Motion Object Segmentation, Recognition and Tracking

1
Motion Object Segmentation, Recognition and
Tracking

• Huiqiong Chen Yun Zhang Derek Rivait
• Faculty of Computer Science
• Dalhousie University

2
Aims

• Goal of this research
• To achieve a robust, low-complexity and accurate
  method for motion segmentation by using
  perceptual organization principles
• Motivation
• The role of Perceptual organization in vision is
  critical to success.
• Proposed method GET based motion segmentation
• Applications
• Video coding and compression
• Video surveillance
• Military target detection
• Medical Imaging
• Traffic Monitoring

3
GET-based Motion Segmentation System Architecture
4
System Data Flow
5
Sample 1 Walk Man Sequence
6
Sample 2 Express Way Sequence
7
License Recognition and Tracking

• Goal
• develop a practical solution to extract license
  plate of moving vehicles so that the license
  plate of each vehicle passing by can be
  identified automatically.
• Key idea
• combine motion tracking with region detection
• use application specific knowledge to guide for
  the target region detection region shape, ratio
  of width to height
• use knowledge previously discovered to generate a
  Region of Interest which focuses tracking to
  relevant areas.

8
License Recognition and Tracking (Contd)
GET feature map
Original frame
9
License Recognition and Tracking (at night)
10
License Recognition and Tracking (During the Day)
License plate