Sharif Digital Repository

Today, with the spread of various videos especially sports video and availability of them, the need for automated methods for search and retrieval of concepts from videos is strongly felt. On the other hand, summarizing video in order to making significant parts of them available is very important. One of the broadest scope that researchers have provided methods for extracting semantic concepts, is soccer videos. Often to extract high-level events, we need a definition and use of intermediate concepts. In this thesis "replay", "audio highlights" and "view type" are used to extract important events in the game. Here, two new methods are introduced for replay detection and view type... 

Image registration is the process of matching two or more images captured at different times, angles or even sensors. Image registration is finding a transformation between two images. Image registration has many applications in different fields of Image Processing and machine vision. In the field of remote sensing, image registration can be used for environmental monitoring, generating graphical maps and gathering data for geogeraphical information systems, also in the field of medical image processing it can be used to detect tumor growth. Generally whenever there is need to extract information from many images; image registration is considered as an important pre-processing step.... 

Gene regulatory network consists of a set of genes; interacting with each other via their protein products. Such interations lead to the regulation of the genes’ production rate. A breakdown in the regulatory process, may lead to some kinds of diseases. Therefore, understanding the gene regulatory process, is beneficial for both diagnosis and treatment. In this thesis, gene regulatory networks are modeled by the means of dynamic Bayesian networks. We have used sampling based methods, in order to learn the network structure. As these methos have a very high computational cost; we have used a correlation test to prune the search space. This way, an undirected network skeleton is obtained; for... 

Image classification is one of the most challenging issues in computer vision. One sort of such classifications is Scene Classification. To perform automatic classification reserchers used many aproches.The general approach used features directly extracted from the image, such as color and texture or features extracted by the SIFT algorithmetc. Another method is based on recognizing object of the Scene (espessially indoor scene). This method is based on finding of a limited number of prespecified objects. In the proposed method, first a window surrounding each objects, (regardless of the type of object) founded. Then the SIFT feature is extracted from that window. All features (corresponding... 

Price prediction in financial markets is an exciting problem for a vast majority of groups and people; however, investment portfolio managers and owners are always looking for holistic predic-tion approaches and tools having high functional accurate metrics. Strictly speaking, players in fi-nancial markets are always in search of methods and toolboxes since they need to overcome the un-certainty of their buy, sell, or hold decisions in order to reduce the investment risk. In this research, we have tried to deal with the stock price prediction problem as an asset pricing problem and find a novel approach to push forward the state-of-the-art of the problem based on the fundamental pric-ing... 

Wireless networks with the shared media require Medium Access Control (MAC) protocol that provides contention resolution mechanism. The design of MAC protocol is a challenging issue because of limited energy source and high propagation delay in UnderWater Acoustic Sensor Networks (UWASNs). The most important goal of the MAC design for underwater sensor networks is to resolve data packet collision efficiently with respect to energy consumption. So, selection of MAC protocol is an important task because it has significant impact on the system efficiency. A suitable MAC protocol must consider some factors such as fairness, energy consumption, latency and throughput based on various... 

In this research we explain a new method for detection of infraction behavior of drivers in highways. There are different approaches to find and register such abnormal and illegal behaviors, such as a camera-based monitoring system and Radar-based infraction detectors. But none of them can recognize and register the weave through traffic or spiral motions of vehicles, because the radar-based systems do not have adequate latitudinal–resolution while their longitudinal-accuracy is moderate. Moreover, the locations of radars are fixed and, the drivers adapt the velocity of their vehicles based on the position of the known radars. Consequently, these systems are inherently not suitable for our... 

It is believed that the synchronized activity of different brain areas, is the main cause of information binding inside the brain. Tis is definitely one of the most exciting challenges in modelling modern complex systems. Brain disorders such as schizophrenia,Alzheimer’s disease, epilepsy, autism and Parkinson’s disease are associated with abnormal synchronization abilities of neural networks. Functional connections can be assessed indirectly by measuring the electrophysiological criteria of ynchronization.Traditionally, in the study of neurophysiological, synchronizations are assessed by analyzing the coherence of frequency-domain characteristics of time series in standard methods for... 

Sensor networks is a set of sensors (nodes) which are communicating with each other, and gather information from different environments, and process them to give us valuable results. Sensors have a small energy source like a small battery which have to be used efficiently. Congestion detection and avoidance, together are of great importance in Wireless Sensor Networks (WSNs) in case of traffic load, data loss, and power consumption. In this work, we propose a new APS layer (application support sub-layer), which is another key standard component of Application layer, offering a well-defined interface and several control services. It functions as a bridge between the network layer and other... 

A Mobile Ad hoc NETwork (MANET) is a set of wireless mobile nodes that form a self configured network. MANETs do not have infrastructure and are not currently deployed on large scales. So research in this area is simulation based. Mobility model in a mobile ad hoc network explains the movement pattern of mobile users and is designed to describe the change of their location, velocity and acceleration over time. One of the challenges in this field is the definition of a common mobility model that provides an accurate and realistic movement description of mobile nodes. In this thesis we want to study the mobility pattern of vehicles on Kish Island and compare it with different kinds of... 

One of the major problems encountered in artificial intelligence is environment interaction. A sensible way for around-interaction is to collect information from sensors and make proper reactions. Generally, in the real world, an intelligent agent needs to know its pose. There are several ways to know the pose of an agent. One of them is based on the fusion of the output of an inertial sensor with the information retrieved from a camera. This leads to enhance the estimation of the pose of an agent. In this research different solutions for this problem are compared. Studies are done with the goal of attitude detection for vehicles and mini-robots, moving on planar surface. For this end, we... 

Widespread use of the interconnected devices has caused an increase in the number of online services and the Internet of Things applications. Some of the protocols used in networking equipment does not have the ability to manage the high volume of traffic, scalability and mobility. Existing networks with traditional protocols are rigid and the process of policy-making on network is slow. Manging the huge number of data flows is a complex and time-consuming task. To respond to such problems, in this study a Software defined architecture is proposed as an alternative to traditional network architecture. In this type of networks, administrators have overview on network controllers and network... 

Human face-to-face communication is an ideal model for designing a multimodal/media human-computer interface (HCI). Recent advances in image analysis and pattern recognition open up the possibility of automatic detection and classification of emotional and conversational facial signals. Automating facial expression analysis could bring facial expressions into man-machine interaction as a new modality and make the interaction tighter and more efficient. In this research an accurate real-time sequence-based system for representation, recognition and analysis of low-intensity facial expressions and facial action uints (FAUs) is presented. The feature extraction is done using facial feature... 

Sensor Localization is a crucial part of many location‐dependent applications that utilize wireless sensor networks (WSNs). Several approaches, including range‐based and range‐free, have been proposed to calculate the position of randomly deployed sensor nodes. With specific hardware, the range‐based schemes typically achieve high accuracy based on either node‐to-node distances or angles. On the other hand, the range‐free mechanisms support less positioning accuracy with less expense. The proposed scheme is based on range‐free localization, which utilizes the received signal strength (RSS) from the anchor nodes. In this work, genetic fuzzy and neuro-fuzzy systems are used to yield more... 

Mobile Ad-Hoc Networks (MANETs) have been studied as one of the most important technologies in the mid to late 1990s. There are several research works on types of network traffic modeling and prediction. Therefore, a very important issue is to make prediction on traffic-flows that each node handles. Because of this, prediction permits us to improve and increase the performance of the network. This project is a contributing effort to improve the traffic packets prediction by Neural Networks in MANET. The main goal of this thesis is about the recovery of data after crisis in phenomenal roads and highways. Our goal is recognizing phenomenal crisis-points in roads. In this thesis packets are... 

This thesis presents several methods time-frequency based approach for classifying the vibration signals of rotatery machine. It uses the features extracted from the time-frequency distribution (TFD) of the vibration signal segments. Results of applying the method to a database of real signals reveal that, for the given classification task, the selected features consistently exhibit a high degree of discrimination between the vibration signals collected from healthy and fault machine. A comparison between the performances of the features extracted from several TFDs shows that the STFT slightly outperforms other reduced interference TFDs.


 

Content-Based Image Retrieval has been an active research area in recent years, due to the vast amount of digital media available via the Internet. In this work we formulate contentbased image retrieval as machine learning problem using the relevance feedback technique and propose a learning algorithm adapted to specific properties of image retrieval. After studying specific properties of image retrieval as a machine learning problem, we propose a Bayesian framework for image retrieval based on one-class learning and test it on different image datasets. The proposed method is a kernel based approach and can also utilize domain knowledge in the form of prior knowledge in constructing a model... 

Crop classification is one of the most important applications of remote sensing in agriculture. Knowing what crops are on the farm is invaluable both on a micro and macro scale. For example, this information can be used to design and imple- ment agricultural policies, product management and ensure food security. Also, this information can be used as a prerequisite for implementing other programs at the farm scale, such as monitoring and detecting anomalies during the crop growth cycle. Most of the studies in this field are focused on the optical data of the Sentinel-2 satel- lite, but the optical data are vulnerable to atmospheric conditions, and on the other hand, there is valuable... 

Visual Question Answering (VQA) deep-learning systems tend to capture superficial statistical correlations in the training data because of strong language priors and fail to generalize to test data with a significantly different question-answer(QA) distribution. To address this issue, we introduce a Visually Directed Question Encoder to replace the commonly used RNNs in base models. our method uses visual features alongside word embeddings of question words to encode each word. As a result, the model is forced to look at the visual information relevant to each word and it no longer produces answers based on just the question itself. We evaluate our approach on the VQA generalization task... 

Throughout history, forming an optimal asset portfolio has been the primary goal of capital owners and managers of investment funds in any economic activity. Achieving this goal is equivalent to trying to minimize the risk caused by the inevitable fluctuations in the capital market and maximizing the overall investment return during the expected period. Investors can operate in various financial markets where there are different stocks and asset classes in each of these markets. The main goal of investors is to identify profitable stocks and form an optimal asset portfolio based on them.Based on this, during the past decades, many studies have been conducted to form and optimize the stock...