Document Details

Document Type : Thesis 
Document Title :
Person Re-Identification System
نظام لإعادة تعريف الشخص
 
Subject : Faculty of Computing and Information Technology 
Document Language : Arabic 
Abstract : Person re-identification is one of the major and critical tasks within smart surveillance systems. Person re-identification systems aim to recognize whether a person has been observed in another non-overlapping camera over large cameras network. It is a challenging task since the large variations in the appearance of persons across different non-overlapping cameras. Recently, instead of hand-crafted methods, deep learning models add significant improvement in many computer vision problems. Most of current state-of-the-art person re-identification systems reidentify a person in a short-term situation when a person did not change their appearance. However, these systems fail when reidentify a person in a long-term situation because these systems depend only on appearance features and the person is expected to change his appearance. The long-term person re-identification is very common in real-world and until now it is not widely explored. In this thesis, we proposed the first long-term person re-identification system based on deep learning by extracting discriminative human gait features to address the problem of appearance variations and support both short-term and long-term person re-identification. In our proposed model, the Resnet-50 is finetuning to extract discriminative gait features. A combination of instance normalization and batch-normalization is adopted in Resnet layers which make our model invariant to appearance changes. A combination of softmax and triplet loss functions is used for training the model. Our proposed model is evaluated on CASIA-B dataset which is challenging benchmark dataset that has many different appearances for each identity. A comprehensive evaluation shows that our proposed model outperforms the existing state-of-the-art systems, especially in rank-1 and rank-5. It achieved from 59.7% to 88.1% in rank-1 and from 80.05% to 96.25% in rank-5. Also, our proposed model is evaluated on short-term person re-identification dataset, Market1501 and it achieved 90.1% in rank-1. 
Supervisor : Dr. Lamia Al-Rifai 
Thesis Type : Master Thesis 
Publishing Year : 1441 AH
2020 AD 
Added Date : Monday, June 29, 2020 

Researchers

Researcher Name (Arabic)Researcher Name (English)Researcher TypeDr GradeEmail
منى عمر المساوىAl-Masawa, Muna OmarResearcherMaster 

Files

File NameTypeDescription
 46535.pdf pdf 

Back To Researches Page