Logo image
3D pedestrian tracking and frontal face image capture based on head point detection
Journal article   Peer reviewed

3D pedestrian tracking and frontal face image capture based on head point detection

Zhongchuan Zhang and Fernand Cohen
Multimedia tools and applications, v 79(1-2), pp 737-764
01 Jan 2020

Abstract

Computer Science Computer Science, Information Systems Computer Science, Software Engineering Computer Science, Theory & Methods Engineering Engineering, Electrical & Electronic Science & Technology Technology
This paper proposes a method to track pedestrians in crowded scenes and capture the close-up frontal face images of a person of interest (POI) for recognition. Pedestrians are tracked via 3D positions of the head points (the highest point of a person) using 2 static overhead cameras. Head points are located and tracked based on the geometric and color cues in the scene. Possible head areas in a frame acquired from one of the overhead cameras are determined based on projective geometry. Head areas belonging to a person are clustered. Without creating a full disparity map of the scene, the 3D position of a pedestrian is obtained by utilizing the disparity along the line segment that passes through his/her head top. The 3D head position is then tracked using common assumptions on motion velocity. If the tracking is not accurate enough, the color distribution of a head top is integrated as a complementary method. With the 3D head point information, a set of pan-tilt-zoom (PTZ) cameras are scheduled to capture the frontal face images of POI. A most suitable PTZ camera is selected by evaluating the capture quality of each PTZ camera and its current state. The approach is tested using a publicly available visual surveillance simulation test bed. The experiments show that the 3D tracking errors are around 4 cm and high quality frontal face images are captured.

Metrics

16 Record Views
2 citations in Scopus

Details

UN Sustainable Development Goals (SDGs)

This publication has contributed to the advancement of the following goals:

#11 Sustainable Cities and Communities

InCites Highlights

Data related to this publication, from InCites Benchmarking & Analytics tool:

Collaboration types
Industry collaboration
Domestic collaboration
Web of Science research areas
Computer Science, Information Systems
Computer Science, Software Engineering
Computer Science, Theory & Methods
Engineering, Electrical & Electronic
Logo image