Dissertations / Theses on the topic 'Image-based positioning'

Position of the vehicle is essential to navigate the vehicle along a desired path without any human interference. Global Positioning System (GPS) loses significant power due to signal attenuation caused by construction buildings. A good positioning system should have both good positioning accuracy and reliability. The purpose of this thesis is to implement a new positioning system using camera and examine the accuracy of the estimated vehicle position on a real-time scenario. The major focus of the thesis is to develop two algorithms for estimation of the position of the vehicle using a static camera and to evaluate the performance of the proposed algorithms. The proposed positioning system is based on two different processes. First process uses center of mass to estimate the position, while the second one utilizes gradient information to estimate the position of the vehicle. Two versions of the positioning systems are implemented. One version uses center of mass concept and background subtraction to estimate the position of the vehicle and the other version calculates gradients to estimate the position of the vehicle. Both algorithms are sensitive to point of view of the image i.e height of the camera. On comparing both algorithms, gradient based algorithm is less sensitive to the camera view. Finally, the performance is greater dependent on the height of the camera position for center of mass positioning system, as compared to the gradient positioning system but the accuracy of the systems can be improved by increasing the height of the camera. In terms of the speed of processing, the gradient positioning system is faster than the center of mass positioning system. The first algorithm, based on center of mass has 89.75\% accuracy with a standard deviation of 3 pixels and the second algorithm has an accuracy of 92.26\%. Accuracy of the system is estimated from the number of false detected positions.

The Dynamic Positioning (DP) systems currently available employ a feedback controller based on the position and heading error, associated with a wind feed-forward control action to keep the vessel position. This last technology improves the eficiency of the DP by anticipating wind loads. However, there is no consolidated technology to feed the controller with wave loads information and the major issue is related to the wave measurement process. The present thesis aims at filling this technological gap by proposing an alternative approach for the measuring of waves near the vessel. Image-based measurement methods are noninvasive and can produce a spatial and temporal description of the surface, making them suitable for recovering the geometry of liquid surfaces. Nevertheless, these surface reconstruction methods, particularly those applying stereoscopic approaches, have dificulty in measuring waves produced in laboratory facilities, mainly because, in such conditions, the water surface is smooth, translucent and highly specular. Alternatively, intensity based image methods are capable of dealing with these surface characteristics, if employed under controlled conditions and if a suitable reflectance model is selected. In the present study, a well-known reflectance model is applied to recover the parameters of regular waves produced in an offshore basin. Firstly, an experimental setup is proposed, designed to grant appropriate conditions for the application of the reflectance model, even under conventional laboratory illumination. Later, the second set of experiments with a model scale DP vessel are presented, which applied the image-based method developed before as a wave feed-forward system. Three control methods are evaluated, namely: PID, PD (Proportional-Derivative) and the PD with the wave feed-forward. Results demonstrated that the presence of the wave feed-forward reduced the steady error of the PD controller. These are the first steps towards the practical use of the wave feed-forward, and several aspects remain pending. However, the promising results and discussions regarding the future steps define the contribution of this work.
Os sistemas atuais de Posicionamento Dinâmico (DP) empregam técnicas de controle baseadas na realimentação da posição e do aproamento, associadas à compensação antecipada das cargas de vento (wind feed-forward) para manter a posição da embarcação. Esta tecnologia melhora a experiência do DP, pois antecipa a ação do vento. Entretanto, não há tecnologia consolidada para a pré-compensação de forças ondas (wave feed-forward) e o maior desafio reside na medição desses agentes ambientais. A presente tese tem como objetivo preencher essa lacuna tecnológica propondo uma abordagem alternativa para a medição das ondas próxima ao casco. Métodos de medição baseados em imagem são não invasivos e produzem descrições espaciais e temporais da superfície analisada, tornando-os particularmente adequados à medição de superfícies líquidas. Entretanto, os métodos comumente empregados, principalmente aqueles baseados em imagens estéreo, são incapazes de medir ondas produzidas em ambiente de laboratório, pois, nestas condições, a superfície da água é lisa, apresenta transparência e comporta-se como um espelho. Por outro lado, métodos baseados na intensidade de luz são capazes de lidar com tais características, se aplicados em condições adequadas e quando empregam modelos apropriados. Neste trabalho, um modelo de reetância amplamente conhecido é empregado para extrair os parâmetros principais de ondas regulares produzidas em um tanque de provas offshore. Inicialmente, propõe-se um arranjo experimental que permita a aplicação do modelo de refetância construído, mesmo sob as condições de iluminação naturais do laboratório. Posteriormente, um segundo conjunto de experimentos com um modelo de embarcação DP é proposto, no qual se aplica o método baseado em imagem, testado anteriormente, como sistema de pré-compensação de forças de onda. Três métodos de controle são avaliados, quais sejam: PID (Proporcional-Integral-Derivativo), PD (Proporcional-Derivativo) e o PD-WFF (PD associado ao wave feed-forward). Os resultados demonstram que a presença pré-conpensação de forças de onda reduz o erro em regime do controlador PD. Estes são os primeiros passos em direção da incorporação da pré-compensação de forças de onda nos sistemas DP reais e diversos aspectos técnicos ainda estão pendentes. Entretanto, os resultados promissores e discussões acerca dos futuros passos da pesquisa definem a contribuição do presente trabalho no campo do Posicionamento Dinâmico.

Virtual camera realisation and the proposition of trapezoidal camera architecture are the two broad contributions of this thesis. Firstly, multiple camera and their arrangement constitute a critical component which affect the integrity of visual content acquisition for multi-view video. Currently, linear, convergence, and divergence arrays are the prominent camera topologies adopted. However, the large number of cameras required and their synchronisation are two of prominent challenges usually encountered. The use of virtual cameras can significantly reduce the number of physical cameras used with respect to any of the known camera structures, hence adequately reducing some of the other implementation issues. This thesis explores to use image-based rendering with and without geometry in the implementations leading to the realisation of virtual cameras. The virtual camera implementation was carried out from the perspective of depth map (geometry) and use of multiple image samples (no geometry). Prior to the virtual camera realisation, the generation of depth map was investigated using region match measures widely known for solving image point correspondence problem. The constructed depth maps have been compare with the ones generated using the dynamic programming approach. In both the geometry and no geometry approaches, the virtual cameras lead to the rendering of views from a textured depth map, construction of 3D panoramic image of a scene by stitching multiple image samples and performing superposition on them, and computation of virtual scene from a stereo pair of panoramic images. The quality of these rendered images were assessed through the use of either objective or subjective analysis in Imatest software. Further more, metric reconstruction of a scene was performed by re-projection of the pixel points from multiple image samples with a single centre of projection. This was done using sparse bundle adjustment algorithm. The statistical summary obtained after the application of this algorithm provides a gauge for the efficiency of the optimisation step. The optimised data was then visualised in Meshlab software environment, hence providing the reconstructed scene. Secondly, with any of the well-established camera arrangements, all cameras are usually constrained to the same horizontal plane. Therefore, occlusion becomes an extremely challenging problem, and a robust camera set-up is required in order to resolve strongly the hidden part of any scene objects. To adequately meet the visibility condition for scene objects and given that occlusion of the same scene objects can occur, a multi-plane camera structure is highly desirable. Therefore, this thesis also explore trapezoidal camera structure for image acquisition. The approach here is to assess the feasibility and potential of several physical cameras of the same model being sparsely arranged on the edge of an efficient trapezoid graph. This is implemented both Matlab and Maya. The quality of the depth maps rendered in Matlab are better in Quality.

The primary goal of this dissertation is to develop point-based rigid and non-rigid image registration methods that have better accuracy than existing methods. We first present point-based PoIRe, which provides the framework for point-based global rigid registrations. It allows a choice of different search strategies including (a) branch-and-bound, (b) probabilistic hill-climbing, and (c) a novel hybrid method that takes advantage of the best characteristics of the other two methods. We use a robust similarity measure that is insensitive to noise, which is often introduced during feature extraction. We show the robustness of PoIRe using it to register images obtained with an electronic portal imaging device (EPID), which have large amounts of scatter and low contrast. To evaluate PoIRe we used (a) simulated images and (b) images with fiducial markers; PoIRe was extensively tested with 2D EPID images and images generated by 3D Computer Tomography (CT) and Magnetic Resonance (MR) images. PoIRe was also evaluated using benchmark data sets from the blind retrospective evaluation project (RIRE). We show that PoIRe is better than existing methods such as Iterative Closest Point (ICP) and methods based on mutual information. We also present a novel point-based local non-rigid shape registration algorithm. We extend the robust similarity measure used in PoIRe to non-rigid registrations adapting it to a free form deformation (FFD) model and making it robust to local minima, which is a drawback common to existing non-rigid point-based methods. For non-rigid registrations we show that it performs better than existing methods and that is less sensitive to starting conditions. We test our non-rigid registration method using available benchmark data sets for shape registration. Finally, we also explore the extraction of features invariant to changes in perspective and illumination, and explore how they can help improve the accuracy of multi-modal registration. For multimodal registration of EPID-DRR images we present a method based on a local descriptor defined by a vector of complex responses to a circular Gabor filter.

Finding correspondences between images is a crucial activity in many overlapping fields of research, such as Image Retrieval and Pattern Recognition. Many existing techniques address this problem using local invariant image features, instead of color, shape and texture, that to some degree loose the large scale structure of the image. In this thesis, in order to account for spatial relations among the local invariant features and to improve the image representation, first a graph data structure is introduced, where local features are represented by nodes and spatial relations by edges; second an algorithm able to find matches between local invariant features, organized in graph structures, is built; third a mapping procedure from graph to vector space is proposed, in order to speed up the classification process. Effectiveness of the proposed framework is demonstrated through applications in image-based localization and art painting. The literature shows many approximate algorithms to solve these problems, so a comparison with the state of the art is performed in each step of the process. By using both local and spatial information, the proposed framework outperforms its competitors for the image correspondence problems.

This thesis presents a prototype beacon-based indoor positioning system using IR-based triangulation together with various inertial sensors mounted onto the receiver. By applying a Kalman filter, the mobile receivers can estimate their position by fusing the data received from the two independent measurement systems. Furthermore, the system is aimed to operate and conduct all calculations using microcontrollers. Multiple IR beacons and an AGV were constructed to determine the systems performance. Empirical and practical experiments show that the proposed localisation system is capable centimeter accuracy. However, because of hardware limitation the system has lacking update frequency and range. With the limitations in mind, it can be established that the final sensor-fused solution shows great promise but requires an extended component assessment and more advanced localisation estimations method such as an Extended Kalman Filter or particle filter to increase reliability.

Thesis (MScEng (Mechanical and Mechatronic Engineering))--Stellenbosch University, 2008.
A fluoroscopy-guided needle-positioning system is designed and tested as a first prototype for aiding urologists in gaining fast, accurate and repeatable kidney calyx access during a PCNL procedure while also reducing radiation exposure of the people involved. Image guidance is realized by modelling the fluoroscopic system as an adapted pinhole camera model and utilizing stereo vision principles on a stereo image pair. Calibration, distortion correction and image processing algorithms are implemented on images of a designed calibration object. Thereafter the resulting variables are used in the targeting of the calyx with the aid of a graphical user interface. The required relative translation and rotation of the needle from its current position to the target is calculated and the system is adjusted accordingly. Using digital cameras, needle placement accuracies of 2.5 mm is achieved within the calibrated volume in a simulated environment. Similar results are achieved in the surgery room environment using the fluoroscopic system. Successful needle access in two porcine kidney calyxes concluded the testing

碩士
國立臺灣大學
資訊網路與多媒體研究所
103
This thesis proposes an image-based ego-positioning framework to achieve sub-meter accuracy, with compact reference database compared to previous work, which can be used for the positioning of Internet connected devices. Consumer GPS devices nowadays obtain accuracy around several meters, which could be worse or even unable to work depending on the connection with GPS satellites. Current researches on image-based positioning utilize images and the reference database in different ways, however their accuracy, time consumption or database size have set limitations of their application. Experiments on different scenes in this work demonstrate that the method we proposed effectively reduce the size of reference database while main- taining a sub-meter accuracy for positioning. With evaluated performance, influencing factors are investigated, open problems are discussed for further improvements and applications.

碩士
國立宜蘭大學
資訊工程學系碩士班
106
Global Positioning System (GPS) can provide instantly outdoor position, but GPS are affected by shelter or buildings when positioned in indoors. Signal error of GPS will become very large. Therefore, when the indoor environment cannot be located using GPS. Other indoor positioning technologies are needed to solve the indoor positioning problem. This study explores an indoor positioning method that can be easily used by users and managers. But each of the indoor positioning methods must causes some errors, this study will explore how to achieve the goal of precise positioning by combining other techniques to reduce the error value in the original error. This paper mainly uses Bluetooth Low Energy as the main architecture. On the device, Bluetooth Low Energy's technology provides an indoor positioning method that is compatible with most smart phones and tablets on the market, as well as low power consumption, low cost of construction and easy deployment. Using Beacon and Fingerprinting to detect indoor positioning, and then using Fast corner feature extraction and Hog feature comparison, finally, the angle difference and the scaling ratio are calculated to find the angle and displacement distance. According to the above that getting the actual location of the user and achieving the goal of accurate indoor position. According to the experimental results of this paper, after the user uses Beacon to build the environment, without additional equipment, just a smartphone or tablet can being used to perform the indoor position. That not only reducing the burden of cost but also through image processing and feature point matching to increase Beacon indoor positioning accuracy. For the problem of GPS indoor positioning error, it will be solved more efficiently.

碩士
逢甲大學
自動控制工程學系
89
In this thesis, a LCD positioning system based on Hough Transform is described. Traditionally, the center and edges of an aligning mark are used to analyze the orientation and position of the LCD flat plate. However, when mark is superposed each other or image is filled of noise and the edges are blurred, the problem of position verification becomes more complex. A common skew detecting technique can’t solve these problems. We assist an algorithm that can calculate the location and rotation angle for two marks precisely and quickly by using the Hough Transform. A mark position and orientation analyze method consists three steps. First step is an edge detection algorithm to overcome the image noise. Then to solve the defects of Hough Transform in line position storing and processing speed, a new Hough Transform parameter space and a FHT method based on block is established. Finally, a mark expected model is adopted. The mark expected technique would find LCD’s position and orientation; even a part of mark image is superposition. These methods are not only useful in mark recognition but also helpful in other study for analysis in Hough Transform parameter space.

碩士
國立成功大學
電腦與通信工程研究所
97
More and more inventions in uses of multimedia services and human machine interface (HMI) have been greatly created recently. Due to appealing appearances, versatile functionalities, and low prices, some of them are popular through mass productions. In this thesis, the main goal is to realize an interactive painting system, which is primarily based on image recognition. In order to achieve a steady and low cost system, the proposed system is integrated by a general camera and a liquid crystal display (LCD) monitor in hardware. By using the designed positioning pattern and recognition technique, we accomplished various advanced painting functions, such as capturing texture, auto-positioning, icon selection, and posting a small video, etc. The proposed painting system was practiced by kindergartners. Experimental result shows that image-based auto-positioning brush (IA-Brush), which integrates with a video camera, LCD monitor, and pattern recognition technique, can achieve a highly-acceptable interactive painting system. We believe that the proposed system will exhibit great contributions to many other applications in interactive multimedia education and entertainment systems.

碩士
國立宜蘭大學
多媒體網路通訊數位學習碩士在職專班
102
Due to the rapid development of information technology and the growth of network speed, we increasingly rely on the internet search engines in many aspects in our daily life, such as foods, clothing, housing, transportations, educations, and entertainments. Although the e-map provided on the internet is very convenient and powerful, but the GPS (satellite positioning) location readings provided by e-map and user’s camera are usually different. We hope to use image analysis technology to enhance the positioning accuracy. Image processing and analysis often use methods, such as edge detection, image segmentation, Fourier transform, image feature transform to analyze the degree of different between two images. This thesis aims to compare the image captured by camera with the street view images provided by Google. We compare the image differences through the analysis of image structure, color variance, spectral information, and ASIFT image matching techniques. By the results of image analysis, we expect to estimate the vehicle moving trajectory on the Google map. We proposed an approach combing all above mentioned image analysis methods. The experiment results demonstrate that our method can improve the positioning accuracy.

碩士
國立中山大學
海下科技研究所
106
Baseline acoustic positioning systems can obtain the absolute positions of underwater targets, but its update rate is slow and calibration is often complicated and tedious. In addition, temporal and spatial variations of water column sound speed profile significantly affect positioning accuracy of the baseline system. Doppler velocity log (DVL) and inertial navigation system (INS) have higher update rate than the baseline systems. But DVL and INS are dead reckoning systems which provide relative positioning only and suffer from time-dependent drift error. Considering that video cameras are standard equipment on almost underwater vehicles, it is easy to collect seafloor videos for a vehicle while conducting seafloor survey. With the advantages of high resolution and high frame rate, the seafloor video has great potential for accurately positioning an underwater vehicle based on image feature detection and matching. Therefore, in this study, the feature-based image matching algorithm for positioning an underwater vehicle is proposed. The seafloor videos collected off southwestern Taiwan at a depth of about 1000 meters by the deep-towed vehicle FITS (Fiber-optical Instrumentation Towed System) are used for evaluating the performance of the proposed algorithm. As the success of image feature detection and matching depends on various factors such as seafloor richness and roughness, descriptor of feature detection algorithms, threshold for Hessian keypoint detector, overlapping area of two images, and illumination, considerable effort in this study was made to assess the effects of various factors on the performance of the proposed algorithm. Performance evaluations were conducted by comparing the estimates of the proposed algorithm to the measurements of the DVL onboard the FITS.

碩士
國立中興大學
電機工程學系所
96
This thesis proposes a quick and precise method for target positioning using multiple image coordinate extraction (MICE) technique based on the simple target estimation (STE). In this approach, an aerial vehicle cruises through the target area in a simulated 3D graphic scenario, taking multiple images of the target. Then the obtained images and the associated aerial vehicle’s position data are recorded. After that, the MICE technique is performed to extract the target coordinates. It is hard to get the global solution using the MICE technique to estimate target positions. Generally, the solution can be achieved by a method of try and error with lots of time. In this thesis, the boundaries of solutions are first limited by the STE technique, which estimates the target positions approximately. Then, the MICE approach can speed up the estimation of the target coordinates precisely. In this thesis, a de-interlacing technique on the basis of direction-oriented interpolation is introduced. It is used to suppress the effect of the interlaced images on the target positioning. In addition, a concept on high density terrain data transformed from low density terrain data is made. Experimental results demonstrate that the STE approach can effectively speed up the MICE technique in precise estimation of target positioning.

碩士
國立雲林科技大學
電機工程系碩士班
100
This thesis proposes an image processing scheme on a cell phone equipped with Android operating system to positioning control for an autonomous rotorcraft. The scheme also provides an effective method for the transmission of the real-time image captured by the camera installed on the craft via 3G wireless network to the ground station. The 3G wireless techniques are considered appropriate for the requirement of long-distance transmission of real-time images captured by the rotorcraft. By contrast to the previous work in which the image processing was carried out in the computer system of the ground station, the present research focusing on the development of image processing techniques on the Android platform in a cell phone installed on the vehicle. The information derived from the images is fed to the flight controller via Bluetooth data communication. This video servoing technique has been proved very effective, efficient and low-cost. It not only resolves the long-distance communication problem, but reduces the complexity of the on-line image processing platform on the plane, which is usually expensive. Specifically, the conventional labeling algorithm of global search is not employed in this research for simplicity; instead, the so-called block matching technique is adopted to speed up the object searching process. This has been shown to have improved at least 50% processing speed than the conventional global labeling methods.

碩士
國立臺灣科技大學
機械工程系
106
This research provides two solutions to solve two exist problems in the automated production process. Firstly, we provide a solution to improve the efficiency and accuracy of a manual camera calibration process by using a robot-arm operated automatic calibration procedure. With the robot arm holding the camera in a predefined positions and orientation to take the pictures on the calibration board and calibrate the camera intrinsic parameters, the lens distortion parameters and hand-eye transform matrix at the same time in a fully autonomous manner. Secondly, in order to tackle the difficulty resulted from poor object positioning situations in the automated production process, we develop a new solution by attaching some artificial landmarks on the target object and using the camera system to perform the calibration based on the image containing the landmarks. In the case that the object has some position errors, the system can compute the compensation displacement needed by the robot arm by comparison of the two images of the artificial landmarks. The results in the experiments show both systems have successfully shown high performance, high accuracy and high implementation potential.

碩士
國立交通大學
機械工程系所
95
In this thesis, an vision-based positioning system is implemented on two-axis linear motor stage. Two sub-systems adopted herein are image processing and motion positioning subsystem. In the image processing subsystem, the raw vision of the moving object is captured from a CCD(Charge-Coupled Device) and the image is processed from the DSP-Based image card(DM64xMM).After several image processing procedures, such as the gray-scale processing, image enhancement, thresholding, and edge detection, the image data is transmitted to PC by USB interface to compute the centroid of the moving object. In the motion positioning subsystem, the object is controlled to move to the desired target by using a motion card(PISO-PS200).After integrating these two subsystems, the vision-based positoning system can move the object to target within one image pixel resolution.

碩士
國立虎尾科技大學
光電與材料科技研究所
95
In this thesis, we propose a FPGA (Field Programmable Gate Array) chip module to implement the light spot positioning system and its applications. We use image sensor to capture two successive images by controlling the power of laser pointer or LED (light emitter diode) ON and OFF. Then a simple subtraction operation is used to eliminate the background and avoid the light noise. By this method, we can get a resulted image with noting but a clear light spot and it becomes very easy to locate its coordinate. We also design an algorithm suitable for FPGA-based system to locate the coordinate of light spot center. The conventional PC-based light spot locating method records all the coordinates of the spot pixel in the image, then compute the average center to get the coordinate of the light spot. Our algorithm uses the replaced method to save memory and logic elements in the FPGA module. The total used logic elements (LE) of our algorithm are only 16.6% of that of the conventional method. To interact with the computer, we use FPGA to control the USB micro-controller in a commercial USB mouse to make the compute cursor to follow the light spot motion. Besides, we put a diffractive optical element (DOE) in front of the image sensor to blur the background image and transform the light spot into a diffractive circle pattern. This approach can eliminate the subtraction noise caused by the two successive images without similar back-ground images.

碩士
樹德科技大學
資訊管理系碩士班
106
In the event of an accident, the people fled in order to survive. The route of their journey was greatly affected by the environment and the surrounding conditions. The accident was often accompanied by smoke caused by various factors, which made it impossible to know effectively through the on-site surveillance camera. Personnel travel and escape, the personnel may be different in body shape or body state, and the moving process has different speeds of movement and collision with the push and the person in the image overlaps, so that the subsequent analysis and research on its travel dynamics Not easy. This study uses the artificial intelligence deep learning training model combined with image recognition technology to divide the Gongge area into 5 square meters for people to move in their area, and place smoke cakes for smoking, create a smoke scene of the disaster scene, and simulate The indoor camera monitors the camera, while the inverted U-shaped route simulates the overlap between the person''s escape and the obscuration of the smoke mask for image recognition in different route segments. The captured image is obtained by Tensorflow''s Object Detection API combined with SSD and Fast R-CNN training algorithms to detect the impact of smoke screen on personnel image recognition. After processing the frame image and data The chart shows that there is a significant difference in the recognition rate between the two, and the Fast R-CNN model is used to take the number and position information from the image from the comparison results. It is hoped that the research will improve the current disaster prevention route planning and post-disaster analysis. Integrity.

碩士
國立虎尾科技大學
光電與材料科技研究所
97
This paper presents that the laser light position using computer calculation make mouse index move to laser light position and then achieve the interactions between humans and machines. The calculation is the eliminations of odd frame and even frame due to light and dark changes from illuminant of laser pen. The light position system operation and control axis are designed by a commercial software package, Lab-VIEW (Laboratory Virtual Instrument Engineering Workbench). According to the light and dark changes and bright spread from laser pen, web camera gets two continuous frames, odd frame with laser spots and even frame without laser spots, and then gets one laser spot image from the elimination of two frames. The lightest position on CMOS sensor X/Y coordinates in laser image is calculated by Lab-view. In order to get the opposite positions between projection screen effect area and the coordinates of computer screen, the projection screen and the web camera resolution are adjusted. As the lightest position of laser point in CMOS sensor is recognized, the mouse index can be moved to the opposite position of computer screen and achieve the interactions between humans and machines. Additionally, all calculations are through computer and simplify other external circuit, for example, to process images by FPGA or DSP (Digital signal processing). The processes are three steps. First, the computer call DLL to get image by connecting web camera. Second, the laser spot coordinate position on web camera could be got by image process identification. Finally, the mouse index is moved to laser point position for chasing the laser index.