Undergraduate-Research Mentored Projects (2007-2011)

Demos of Some Projects
 Intelligent Presentation Guru (IPG), 2010 

YouTube Video

 Gait Analysis for Human Identification (GAHI) System, 2009 

YouTube Video

 Optical Braille Recognizer system (OBRS), 2009
Published in IPCV 2000

YouTube Video

List of All Mentored Computer Vision Bachelor Projects and Their Description


+ Stereo-camera based object recognition system: This project is planned to perform object recognition based on depth information integrated and color information captured from a stereo camera. Firstly, a 3D Computer Vision technique will be used to extract depth information stereo camera images, a hybrid technique using depth and color information will be used to segment and recognize objects.

Intelligent Home Security system: This system provides security function using finger-print recognition and face recognition as identification components. Cameras will be installed in the entry points of the home and finger-print reader will be used to allow authenticated people to enter home by finger-print. Moreover, the system is going to built such that any suspicious action did in the home will be sent as a mobile message to the home owner(s); using cameras.


Intelligent Presentation Guru (IPG) T
his project focuses on the recognition of presenter activities. The system is able to recognize a set of gestures including 3D pointing, consistent motion, and other different gestures to control the PC (e.g. moving to and from the next slide). Kernel Density Estimation (KDE) is used to model both the background and the foreground leading to more robust results. The segmented image is then input to the 3D pointing and gesture recognition modules. The 3D pointing is based on the Pinhole Model to estimate the depth of the arm using two of the three cameras then a 3D vector is extracted indicating the direction of pointing. On the other hand, the gesture recognition module performs motion analysis on the presenter’s silhouettes to recognize the gestures

YouTube Video


+ SMart class Room System (SMRS): SMRS is composed of three modules: (1) The Networking module whose job is to manage the communication within an in-class environment gathering students and teachers or a remote environment, where students and even class teachers can communicate remotely. The intelligence in this system is centralized around recognizing students’ activities and behavior. (2) The Hand Raising Detection module (based on hands moving up) to monitor students’ interaction in the class. (3) The Face/ Facial Expression Recognition module to identify interactive students and to recognize facial expressions (sad, happy, bored, etc) remotely or in-class using a segmentation module followed by a Particle Filter for classification. This provides the teacher with a live-feedback for students’ interaction and their live-impression about the lecture, enabling him to live-redirect his approach of explanation and to track the pace of each student.


+ System-On-Chip Action Recognition: This project aimed to design a computer vision system on a single chip achieving real time processing and portability. The system utilized LeanXCam, A camera having a DSP (Digital Signal Processor) to be programmed, as the core HW on which a real-time application was implemented. We chose one of the ubiquitous actions in Arabian countries which is Muslim Prayer. The development environment of this project was totally under Linux.



+ Gait Analysis for Human Identification (GAHI) System: This system is able to recognize humans through analyzing the gait cycle. At first, a subtraction background model is developed to discriminate between the foreground and the background. The resultant binary image is grouped into blobs. These blobs are then classified to ignore non-human objects; using the Codebook Learning approach. Then, a Gait Energy Image (GEI) is generated for each human gait cycle. The GEI is used as input to PCA (Principle Component Analysis) for dimensionality reduction, followed by MDA (Multiple Discriminate Analysis) for clustering per human rather than per gait cycle, leading to more accuracy.

YouTube Video

+ Intelligent Information Discovery & Processing System (IIDPS)IIDPS is mainly a NLP system able to crawl the web semantically based on specific topic, and automatically generate a report for the user summarizing the achieved information. Firstly, the crawled page passes a cleaning phase (clearing irrelevant information, e.g. tags, ADs and etc). Then document vector is created and passed to a previously trained Neural Network to determine relevancy to the trained topic(s). The documents are viewed and ranked according to the relevancy, with the most relevant paragraph viewed in the summarized report.



+ Optical Braille Recognizer system (OBRS): OBRS is a system that dynamically converts Braille documents into plain text. Firstly, the scanned image is filtered for enhancement as a preprocessing step. Then the image is de-skewed to resolve any rotation problems. Next, Braille recto and verso dots are detected; based on a special thresholding module. Finally, Braille cells are recognized based on the detected dots.     

YouTube Video

+ Automatic Panoramic Image Stitching Engine (APISE): APISE is automated software for panoramic image stitching. Features are extracted from the images using SIFT then matching is performed using RANSAC. Afterwards, connected components are located which are further used for performing bundle adjustment and panoramic rendering using multi-band blending.

+ Medical Image Registration system (MIRS)MIRS is a system for non-linear registration of similar medical images for diagnostic purposes. A separate transformation matrix is built up for each pixel, nonlinearly, such that the transformational smoothness is ensured and the illumination difference is minimized. This optimization problem was formulated as an energy function that was solved numerically achieving reliable results.



+ Auto Spelling Robot: This project is actually an amalgamation of speech recognition, optical character recognition and hardware. The robot perceives Arabic speech using its speech recognition module. It then scans in-line cubes labeled with all Arabic characters, searching for a character belonging to the spelled word. Next, it moves the cube to its position in the composing area (the area where the new word is to be formed).