Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


BhattaraiS.T (final).pdf (2.9 MB)

ETD Abstract Container

Abstract Header

Digital Architecture for real-time face detection for deep video packet inspection systems

Bhattarai, Smrity
http://rave.ohiolink.edu/etdc/view?acc_num=akron1492787219112947

Abstract Details

2017, Master of Science, University of Akron, Electrical Engineering.
Face detection and optional recognition is a highly researched area in digital image processing. Face detection allows gathering of statistical data from video sequences, with applications in a variety of areas such as bio-metrics, information security, and video surveillance. The growing abundance of video sensors that are connected to the internet require high-throughput real-time processing of a multitude of digital video feeds, where each feed provides independent real-time statistics of the number of persons shown in the feed. Typical applications include pedestrian counting, public transit monitoring, crowd control, and sporting events. Video surveillance and security applications in particular can benefit from real-time algorithms that can process large amounts of data. Thousands of video sources must be monitored for extracting situational awareness information for homeland security and public safety applications, and the manual monitoring of such a vast amount of data is nearly impossible. Algorithms for both face detection [1–4] and recognition [3, 5–7] take two main approaches involving the local detection of facial features based on a geometric model of the human face [8] and a holistic based feature recognition, where the image data is treated as an entity without isolating different regions of the face. The main challenge in feature based facial detection is identification and location of human faces regardless of their pose, facial expression, orientation, imaging condition or presence of structural components [9]. Some advanced image-based pattern recognition techniques have been developed to handle difficult scenarios like multiple faces, faces of different sizes, and even detection in heavily cluttered backgrounds. [8] In this thesis, we explore how hardware computing architecture for detection of an image, as a face or non-face, is designed. The computing architecture is first designed, modeled, and tested in MATLAB simulink using Xilinx blockset. Images were later tested using a Virtex-6 FPGA ML605 Evaluation Kit. A field-programmable gate array (FPGA) is an integrated circuit designed to be configured by a user or a designer after manufacturing. The system uses the features of a face and non-face, which were previously extracted by training the set of face and non-face patterns. The system is fully feature based and does not require any assumptions for processing. In this approach, all the images are treated in the same way. They are not separated into different categories before processing them. The system is basically a combination of different modules like convolution, sub-sampling, bias add, scaling, neuron and decision combined in a specific format to classify the images as a face or non-face on the basis of the output. The algorithm is simple without any need for preprocessing of the image. The performance trade-off exists between the computational precision, chip area, clock speed, and power consumption.
Dr. Arjuna Madanayake (Advisor)
Dr. Ryan C Toonen (Committee Member)
Dr. Kye-Shin Lee (Committee Member)
146 p.
Electrical Engineering
Face detection, Convolutional Neural Network, Image processing

Recommended Citations

Citations

  • Bhattarai, S. (2017). Digital Architecture for real-time face detection for deep video packet inspection systems [Master's thesis, University of Akron]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=akron1492787219112947

    APA Style (7th edition)

  • Bhattarai, Smrity. Digital Architecture for real-time face detection for deep video packet inspection systems . 2017. University of Akron, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=akron1492787219112947.

    MLA Style (8th edition)

  • Bhattarai, Smrity. "Digital Architecture for real-time face detection for deep video packet inspection systems ." Master's thesis, University of Akron, 2017. http://rave.ohiolink.edu/etdc/view?acc_num=akron1492787219112947

    Chicago Manual of Style (17th edition)

akron1492787219112947
414
© 2017, all rights reserved.
This open access ETD is published by University of Akron and OhioLINK.