Sharif Digital Repository

Today with the availability of cheap depth sensor, processing point clouds produced by these sensors and extracting geometric features is an active field of computer vision. Object recognition is a basic computer vision issues that even with considerable research has remained as a challenge. In these thesis we have studied methods of utilizing depth images and geometric information of point clouds in order to extract geometric features from point clouds and have introduces a set of new geometric features using Normal Orientation Histogram. Also a novel and efficient method for segmentation of point cloud of indoor scenes is proposed. Experimental results depict that our proposed methods have... 

Digital videos is one of biggest parts of digital world. With rapid development of social networks, it is more challenging to protect media against copy right violations. Video fingerprint algorithms is one of the best tools which help detecting these violations. In this thesis, we propose two algorithms in compressed and uncompressed domains which are based on dividing videos in shots. We apply our algorithms on H264 Compression standard. In compressed domain, we use frame size in bytes in stream of video for extracting fingerprints. In Uncompressed domain, we apply multi kernel algorithm in image pyramid to extract fingerprint from videos. Results from Testing on our Dataset including 1278... 

Nowadays, electromagnetic waves with millimeter wavelengths have various usages including medical,non destructive testings, structure inspection and imaging usages. Millimeter waves go through most of common clothings, but unlike x-ray waves, these waves have no ionizing effects.Therefore, they have low health risks for humans. This characteristic makes the imaging systems based on millimeter wave suitable for security purposes, such as airport security. The inclined usage of these systems requires an increase in precision, speed, reduction of error and automatic object detection algorithms. An important way to increase precision is to change the imaging system structure, resulting in... 

With the advancement of Machine Vision and Image Processing systems, the need for conceptual interpretation is raising. For this interpretation, one should detect objects available in the image and then tries to find the relations between the objects. For a good interpretation of the image, the machine vision system should learn patterns available in the nature. Object recognition systems are also used in other vision tasks and they can be used for content-based image retrieval, control and surveillance, or human action and gesture recognition for a better and easier human-computer interface. In an object recognition system, first some features should be extracted from the input image and... 

Gesture and motion evaluation are needed for a variety of applications. The use of human hand motions as a natural interface tool has motivated researchers to conduct research in the modeling, analysis and evaluation of various hand movements. In particular, human-computer intelligent interaction has been a focus of research in vision-based gesture recognition. In this work, we introduce a 3-D hand model evaluation method that offers flexible and elaborate representation of hand motion. We used landmarked points on tips and joints of the fingers and calculated the 3-D coordinates of these points through a stereo vision system followed by an HMM (Hidden Markov Model) to recognize hand... 

In this research, we introduce a semi-supervised manifold regularization framework for hu- man pose estimation. Here we aim the three major challenges in discriminative human pose estimation. We utilize the unlabeled data to reduce the need to labeled data and compen- sate for the complexities in the input space. We model the underlying manifold by a nearest neighbor graph. Due to depth ambiguity which is the main challenge in this problem, the true underlying manifold of the data bends and gets too close to itself is some areas which results in poor graph construction. To solve this problem, we argue that the optimal graph is a subgraph of the k-nearest neighbor graph and employ an... 

Human facial activity recognition is one of the endeavors to improve human-computer interaction. Recognition of excitements and emotions on human face by machine and makinga corresponding reaction is essential for man machine intraction.The purpose of this project is recognizingactivities such as speaking, eating, laughing, agree and disagree which have more complexity than usualemotions such as fear and happinesscontained in common datasets.So, adataset in accordance with the above mentioned 5 activities was collected and the appropriate feature vector for analyzing these face activities were implemented.Distance between the interest points located on the face were used as parameters in... 

In this project an algorithm is designed for automatic landing of unmanned helicopter on a pad moving with six degrees of freedom. The designed controller is based on state dependent riccati equations (SDRE).by developing nonlinear mathematical model of helicopter and then converting this model to state dependent coefficient (SDC) form. A nonlinear compensator is added to controller to compensate effect of some nonlinear terms of model that are not able to translate and then state dependent riccati equations are solved.
The relative pose-estimation of landing pad is based on vision. Corner detection algorithm is used to identify and detect features by processing of image taken from... 

Semi-supervised Learning (SSL) is an extremely useful approach in many applications where unlabeled data can be easily obtained. Graph based methods are among the most studied branches in SSL. Since neighborhood graph is a key component in these methods, we focus on methods of graph construction in this project. Graph construction methods based on Euclidean distance have the common problem of creating shortcut edges. Shortcut edges refer to the edges which connect two nearby points that are far apart on the manifold. Specifically, we show both in theory and practice that using geodesic distance for selecting and weighting edges results in more appropriate neighborhood graphs. We propose an... 

In the first section, a novel No-Reference Video Quality Assessment (NR-VQA) method, based on Convolutional Neural Network (CNN) for High Efficiency Video Codec (HEVC) is presented. Deep Compressed-domain Video Quality (DCVQ) measures the video quality, with compressed domain features such as motion vector, bit allocation, partitioning and quantization parameter. For the training of the network, normalized PSNR is used due to the limitation of existing datasets. The evaluation of the proposed method shows that it has”96%” correlation to subjective quality assessment (MOS). The method can work simultaneously with the decoding process and measures the quality each frame in the different... 

Attitude estimation means calculating the state variables during the flight, especially in landing and takeoff phases. If we can extract the optical flow using the sensors mounted on the flying object, due to the fact that the optical flow is created by linear and rotational speed of the object relative to the surrounding, we are able to calculate the relative attitude by analyzing the optical flow. Indeed the purpose is developing this idea by using artificial neural networks.
First, we find the optical flow patterns for every attitude condition near the ground, using geometrical calculations. Then we produce an optimal neural network by these patterns. This network has the ability to... 

One of the applications of machine vision is to identify the indoor scenes. Identification of indoor scenes is the capability of detecting walls, floor and ceiling of an interior. Also, to distinguish the distinctions between these components by using some planes. There are several methods so that by employing machine vision it will be possible to identify indoor scenes and one of the most notable methods is fuzzy image processing. In fact, many applications of image processing require the knowledge of an expert. The theory of fuzzy set and fuzzy logic is a powerful tool to expressing and processing the human knowledge by “if … then …” rules.The purpose of this project is to present an... 

Nowadays, Human Action Recognition is one of the most widely used and active areas of research in computer vision. the purpose of Human Action Recognition is to label an action in a video. This field has numerous applications like human-computer interaction, video analysis, medical care, surveillance camerate, and etc. Like other subcategories of computer vision, today with the advent of deep learning networks and its development, considerable progress has been made in the accuracy and speed of the methods. The main purpose of this research is to improve human action recognition networks on RGB-D videos. In this study, three methods for action recognition using deep neural networks are... 

With the advent of technology, small in size sensors, memory, speeding up the processor and lowering the cost, it is possible to build an embedded camera system. The goal of this project is to design and build an embedded camera system so it can execute any set of necessary algorithms as depending on the application. In this project, two models of embedded camera systems have been presented as an integrated system and a system with independent units. To design the integrated embedded system, ZYNQ processor is used and two structures are presented in the form of hardware-software and hardware design. In hardware-software design, image processing operations are done by software and in hardware... 

Over recent years, ever increasing developing computer architecture and computer science made increase ability of computers for data processing and creating intelligent systems without human supervision. One of major and important usages of this systems is automatic supervise and analysis on video scene by image and video processing, machine vision and machine learning techniques. Developing this systems and increasing number of installed surveillance cameras on cities made employment of this systems for supervise on urban traffic events.Automatic number plate recognition, anomaly detection, accident detection, traffic sign detection and recognition and etc are example of Intelligent Traffic... 

The purpose of this paper is to quantify the extent of damage of rectangular reinforced concrete shear walls after an earthquake using surface crack patterns. One of the most important tasks after an earthquake is to assess the safety and classify the performance level of buildings. This assessment is usually performed by visual inspection that is prone to significant errors. In this research, an extensive database on the images of damaged rectangular reinforced concrete shear walls is collected from the literature. This database includes more than 200 images from experimental quasi-static cyclic tests. Using the concept of Fractal geometry, several probabilistic models are developed by... 

This project aims to highlight vision related tasks centered around "car". First, we gathered a dataset of 4343 front view car images, captured from the streets of Iran and Syria during daylight, the images of which are all manually cropped around their corresponding accurately chosen bounding boxes. we also extracted seven parts (i.e. left and right front lights, left and right mirrors, bumper, plate, and air intake) from each car image in the dataset. Our dataset is suitable for developing and testing bounding box extraction algorithms, holistic and part based analyses, occlusion handling algorithms, etc. next, we utilized Viola-Jones Detector to develop a system for car detection, in... 

In this work, we try to develop a method for searching for an object which is known to us and also returning it back where the search was begun, through shortest possible path. The surrounding environment was assumed unknown and only a stereo camera and an IMU exist to work with. The proposed method was developed in Linux using GPU and with C++ language and also achieved 58 fps. Parallel programming was utilized in different parts of the algorithm and obstacle avoidance is guarranteed. The surrounding environment and obstacles are assumed changing over the time