Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Partially Occluded Object Detection

Sun, Di (2016) Partially Occluded Object Detection. Master's Thesis, University of Pittsburgh. (Unpublished)

[img]
Preview
PDF
Draft Version

Download (2MB)

Abstract

Object detection is a fairly important field in computer vision and image processing, and there are some domains that have been well researched, such like face detection and pedestrian detection, while these detections are based on situation that there are no shelters on objects in images or videos. But most objects are partially occluded in real world. And this becomes one major obstacle in object detection. In this paper, I utilize a state-of-art model, Deformable Part Model (DPM), which present a mixture model contains multi-scale deformable parts. And attach visibility flags to both cells and part of it. A visibility flag is used to indicate whether a specific cell or part is visible or occluded. By removing the contribution of occluded cells and parts to the whole detection score, the performance of partially occluded object detection can be significantly improved. To get the value of visibility flags, I use a mathematical method called Alternating Direction Method of Multipliers (ADMM). I experiment the model on different occlusion situation and change the occlusion conditional likelihood penalty to get better performance.


Share

Citation/Export:
Social Networking:
Share |

Details

Item Type: University of Pittsburgh ETD
Status: Unpublished
Creators/Authors:
CreatorsEmailPitt UsernameORCID
Sun, Didis40@pitt.eduDIS40
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairChen, Yiranyic52@pitt.eduYIC52
Committee MemberLi, Ching-Chungccl@pitt.eduCCL
Committee MemberLi, Haihal66@pitt.eduHAL66
Date: 15 June 2016
Date Type: Publication
Defense Date: 29 March 2016
Approval Date: 15 June 2016
Submission Date: 28 March 2016
Access Restriction: No restriction; Release the ETD for access worldwide immediately.
Number of Pages: 49
Institution: University of Pittsburgh
Schools and Programs: Swanson School of Engineering > Electrical and Computer Engineering
Degree: MS - Master of Science
Thesis Type: Master's Thesis
Refereed: Yes
Uncontrolled Keywords: Object detection, deformable part-based model, partially occluded object
Date Deposited: 15 Jun 2016 13:07
Last Modified: 15 Nov 2016 14:32
URI: http://d-scholarship.pitt.edu/id/eprint/27366

Metrics

Monthly Views for the past 3 years

Plum Analytics


Actions (login required)

View Item View Item