Detection, appearance modeling, identification and tracking of people in surveillance videos | |
| Author | Jain, Sanjana |
| Call Number | AIT Thesis no.CS-17-04 |
| Subject(s) | Pattern recognition systems Machine learning Neural networks (Computer science) |
| Note | A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering in Computer Science, School of Engineering and Technology |
| Publisher | Asian Institute of Technology |
| Series Statement | Thesis ; no. CS-17-04 |
| Abstract | Applications for identifying and tracking target individuals in stores are now rapidly finding their way onto retail businesses’ agendas. Many retailers would like to implement a blacklisted tracking system, and other applications include identifying members of loyalty schemes and giving rewards to special customers. The goal of the thesis is to automate the process of detecting members of a database of target individuals and assist in tracking them through crowds and occlusions. I present an automated approach to face detection, best view analysis, and face verification integrated with tracking to monitor for target individuals. I perform face detection using a custom Viola and Jones AdaBoost detection cascade to extract face images of individuals arriving at the shop. I incorporate face detection with a tracking algorithm in order to assign tracks to faces. The best view face for each individual is then obtained using entropy-based analysis, sharpness analysis, fiducial point detection, and pose estimation. Finally, I perform face verification with the target database using “Siamese” convolutional neural network to track target individuals and raise an alert. Previous face recognition approaches have achieved high accuracy on the high quality Labeled Faces in the Wild (LFW) dataset. In my case study, I focus on blacklisted person alerting for the HomKrun coffee shop at AIT. Images obtained from surveillance cameras have lower quality. Nevertheless, I obtain acceptable results for each of the modules of my proposed system, so it will serve as a baseline for further enhancement of methods for monitoring for target individuals in surveillance videos. |
| Year | 2017 |
| Corresponding Series Added Entry | Asian Institute of Technology. Thesis : no. CS-17-04 |
| Type | Thesis |
| School | School of Engineering and Technology (SET) |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Computer Science (CS) |
| Chairperson(s) | Dailey, Matthew N.; |
| Examination Committee(s) | Mongkol Ekpanyapong;Guha, Sumanta; |
| Scholarship Donor(s) | Thailand (HM King); |
| Degree | Thesis (M. Eng.) - Asian Institute of Technology, 2017 |