Karim Afdel

An efficient method for video shot boundary detection and keyframe extraction using SIFT-point distribution histogram

In today’s digital era, there are large volumes of long-duration videos resulting from movies, documentaries, sports and surveillance cameras floating over internet and video databases (YouTube). Since manual processing of these videos are difficult, time-consuming and expensive, an automatic technique of abstracting these long-duration videos are very much desirable. In this backdrop, this paper presents a novel and efficient approach of video shot boundary detection and keyframe extraction, which subsequently leads to a summarized and compact video. The proposed method detects video shot boundaries by extracting the SIFT-point distribution histogram (SIFT-PDH) from the frames as a combination of local and global features. In the subsequent step, using the distance of SIFT-PDH of consecutive frames and an adaptive threshold video shot boundaries are detected. Further, the keyframes representing the salient content of each segmented shot are extracted using entropy-based singular values measure. Thus, the summarized video is then generated by combining the extracted keyframes. The experimental results show that our method can efficiently detect shot boundaries under both abrupt and gradual transitions, and even under different levels of illumination, motion effects and camera operations (zoom in, zoom out and camera rotation). With the proposed method, the computational complexity is comparatively less and video summarization is very compact.

Exploiting web 2.0 technologies in promoting learning activities: E-learning — Web 2 platform

Many schools and universities around the world recognize the need to improve their process of teaching by using e-learning platform that integrates Web 2 technologies which interest new populations of students. The scope of this paper is to describe our experience at Ibnou Zohr University in Agadir by presenting two trainings. Our e-learning platform consists of moodle together with a set of chosen web2 tools.

Web 2.0 based e-learning: Moodle-openmeetings platform

While Web 2.0 technologies are beginning to have beneficial effects on trade, media and business in general, there are few papers describing the impact of technology on education. Currently, there are isolated cases of teachers who begin to explore the potential of blogs, sharing of media services, and other social software — which, even are not specifically designed for e-learning can be used to give teachers, tutors and learners the ability to create exciting new learning opportunities. Our approach is to continue to benefit from the traditional approach of E-learning by using the Moodle platform and at the same time encourage students to use Web 2.0 tools in their training in order to overcome the problem of isolation faced by learners in the remote training often leading to a high dropout rate due to some technical problems.

A reversible conversion methodology: Between XML and object-relational models

XML is a standard for data exchanging between sites and heterogeneous applications. To exploit these data by database systems based on relational model, algorithms and methods of conversion have been developed. To do same with object-relational systems representing an extension of relational systems we propose in this paper a methodology to convert a XML schema respecting a DTD (Document Type Definition) into a schema of object-relational model. This methodology is reversible so that the result of conversion can be used to rebuild the initial XML schema.

Transforming XML into Object-Relational Schema

Recently, there is a vast increase in the use of XML for describing and exchanging data. To manipulate efficiently these data, it would be wise to use database systems which represent an appropriate tool to store and manage data. To have this purpose, we need to transform XML schema into database models such as relational and Object-Relational (OR). The aim of this work is to present a methodology that transforms an XML schema into the OR model. To be automatic, the steps of this transformation are formalized by a mapping function defined from XML into the object-relational model. Among the advantages of this transformation is that it preserves the structure and constraints defined in XML schema which enables to retrieve the initial XML structure from its converted, OR schema.

Automatic face recognition with aging using the invariant features

In the field of automatic face recognition, transformations of facial features due to aging cause a problem. Due to small amounts of extracted features, the identity verification can be difficult. The feature-based methods that are present in the literature are still being developed, with unsatisfactory results caused by high rates of false matching. In this paper we present a new method of matching verification of SIFT extracted feature points, which uses both the positions and scales of feature points. By using this method and the SIFT descriptor, we develop an identity verification system robust to aged based facial features transformations. The application of our verification system in the FG-net database demonstrates our approachs performance. The experimental results show that if 16.66% false acceptance rate is admitted, 81.81% true matching rate is obtained.

An automatic video surveillance indexing based on facial feature descriptors

Automatic surveillance video footage indexing is much more desirable while providing an assistive tool for personnel security. Since the most relevant object that attracts our attention in surveillance videos is human face, we focus in this paper on building a system for indexing surveillance videos based on human face features. The proposed system has three main stages: Video Surveillance Summarisation, Face Detection, and Facial Feature Descriptors, and Indexing. A keyframe selection technique based on local foreground entropy is used for video surveillance summarisation. In the Face Detection stage, a skin color based method using measurements derived from the color-space components of the keyframe is used to locate eye, mouth and face boundary. Subsequently, SURF algorithm is applied to extract the feature descriptors of interest point from the detected face region. These descriptors are then indexed using vocabulary tree. The integration of the above-mentioned methods that are all good in their results, have made our overall system robust and efficient. Therefore, good results have been obtained while testing in ChokePoint public dataset contains 48 video sequences with a total of 179 349 frames including 64 204 face images.

Prototype of an embedded system using Stratix III FPGA for vehicle detection and traffic management

This paper presents a development of a system of management and control of traffic congestion, the role of this system is to alert decision makers at every moment of the threshold value reached by traffic congestion to take the necessary measures to resolve the problems that appear in traffic jams. One of the modules of that system presents an optimization of vision based vehicle detection and vehicle counting. In this module we will develop a program that detects the vehicles, counts them and calculates speed of each vehicle. The algorithms were implemented in the system using Intel OpenCV library for image processing. The second task in this work is the implementation of the algorithm using Stratix III FPGA, Raspberry Pi B and Raspberry Pi B Camera.

Towards a secure access to patient data in cloud computing environments

In the modern health service, data are stored in a Data Center and it can be accessed only by authorized users. However, this Data are prone to be exposed to a number of attacks, especially by the Cloud providers Personnel with privileged access. To avoid illegal access to comprehensive content of data center including patients information, we propose in this article a mechanism using the content-based watermarking technique. Information of patient and a digest are encrypted, before being embedded into LSBs bitplane of image associated to the patient. This image is integrated directly into the database. Hadoop system with the integrate functions. HDFS and MapReduce will play the key roles for our solution.

Real-Time Video Surveillance System for Traffic Management with Background Subtraction Using Codebook Model and Occlusion Handling

Abstract The scope of this paper is a video surveillance system constituted of three principal modules, segmentation module, vehicle classification and vehicle counting. The segmentation is based on a background subtraction by using the Codebooks method. This step aims to define the regions of interest associated with vehicles. To classify vehicles in their type, our system uses the histograms of oriented gradient followed by support vector machine. Counting and tracking vehicles will be the last task to be performed. The presence of partial occlusion involves the decrease of the accuracy of vehicle segmentation and classification, which directly impacts the robustness of a video surveillance system. Therefore, a novel method to handle the partial occlusions based on vehicle classification process have developed. The results achieved have shown that the accuracy of vehicle counting and classification exceeds the accuracy measured in some existing systems.