Search

Most augmented reality solutions are based on computer vision techniques to detect and track objects using a camera and display a computer generated virtual content. However, main issues of such techniques are image transformations, chaotic environment and illumination level. In general, such noisy environment causes complex or even impossible task for object tracking in real-time. In case of computer vision tracking techniques, occlusions causes 3D virtual content to disappear, therefore it has negative impact for the usability of applications. Because of the named reasons, most widely used modern feature detection and description techniques were analyzed that allows to solve partial occlusion problems and properly display virtual content. In this research robustness comparison according to repeatability criteria was accomplished using different type of image transformations sets by verifying specific image pairs. Computer vision techniques performance estimates in respect of speed was also analyzed to inspect suitability for the augmented reality. Research results are critical to propose improved tacking techniques in augmented reality field, also, by integrating inertial sensor information. Quaternion application for microelectromechanical sensor data were also thoroughly analyzed to estimate camera orientation. With this solution it is aimed to eliminate separate sensor disadvantages and improve camera orientation tracking accuracy, reliability and virtual content representation in augmented reality, without using any computer vision technique. The proposed solution was adapted not only from tracking perspective, but also for the development of spherical form interaction device. Detailed Kalman filter analysis and practical application are also presented for orientation data acquired from real sensors, which are expressed in quaternions. Analysis of gradient descent method and practical application for sensor data fusion were also accomplished. The proposed hybrid tracking technique for augmented reality was proposed, where orientation and position estimates from computer vision and sensor fusion techniques were analyzed in a systemic approach. Sensor array was used to eliminate separate sensor disadvantages and improve tracking without using or partially using computer vision data. Provided research experiments has predefined assumptions and orientation-position data were imitated in respect to computer vision and sensor fusion. Additional conditions were introduced to acquire optimal orientation-position estimates and results were compared. Complementary properties of Kalman filter hybrid tracking technique were demonstrated. Solution was adapted for mobile devices in the field of augmented reality for internet of things. To achieve more accurate and reliable camera orientation and position estimates in space and expand augmented reality application possibilities, in this dissertation computer vision and sensor fusion technique efficiency were analyzed and a new hybrid object tracking technique using Kalman Filter was proposed. For this purpose. The main use of presented technique is to expand augmented reality application fields, also improve object orientation and position tracking for reliable virtual content representation even in cases, when useful information from feature extraction and matching from image is not available.

Most augmented reality solutions are based on computer vision techniques to detect and track objects using a camera and display a computer generated virtual content. However, main issues of such techniques are image transformations, chaotic environment and illumination level. In general, such noisy environment causes complex or even impossible task for object tracking in real-time. In case of computer vision tracking techniques, occlusions causes 3D virtual content to disappear, therefore it has negative impact for the usability of applications. Because of the named reasons, most widely used modern feature detection and description techniques were analyzed that allows to solve partial occlusion problems and properly display virtual content. In this research robustness comparison according to repeatability criteria was accomplished using different type of image transformations sets by verifying specific image pairs. Computer vision techniques performance estimates in respect of speed was also analyzed to inspect suitability for the augmented reality. Research results are critical to propose improved tacking techniques in augmented reality field, also, by integrating inertial sensor information. Quaternion application for microelectromechanical sensor data were also thoroughly analyzed to estimate camera orientation. With this solution it is aimed to eliminate separate sensor disadvantages and improve camera orientation tracking accuracy, reliability and virtual content representation in augmented reality, without using any computer vision technique. The proposed solution was adapted not only from tracking perspective, but also for the development of spherical form interaction device. Detailed Kalman filter analysis and practical application are also presented for orientation data acquired from real sensors, which are expressed in quaternions. Analysis of gradient descent method and practical application for sensor data fusion were also accomplished. The proposed hybrid tracking technique for augmented reality was proposed, where orientation and position estimates from computer vision and sensor fusion techniques were analyzed in a systemic approach. Sensor array was used to eliminate separate sensor disadvantages and improve tracking without using or partially using computer vision data. Provided research experiments has predefined assumptions and orientation-position data were imitated in respect to computer vision and sensor fusion. Additional conditions were introduced to acquire optimal orientation-position estimates and results were compared. Complementary properties of Kalman filter hybrid tracking technique were demonstrated. Solution was adapted for mobile devices in the field of augmented reality for internet of things. To achieve more accurate and reliable camera orientation and position estimates in space and expand augmented reality application possibilities, in this dissertation computer vision and sensor fusion technique efficiency were analyzed and a new hybrid object tracking technique using Kalman Filter was proposed. For this purpose. The main use of presented technique is to expand augmented reality application fields, also improve object orientation and position tracking for reliable virtual content representation even in cases, when useful information from feature extraction and matching from image is not available.

Currently, such microelectromechanical sensors as accelerometers, gyroscopes and magnetometers are the dominant sensors in mobile devices. The market of mobile devices is constantly expanding and focused on sensor integration process by adding supplementary functionality for the applications; therefore, it is possible to adapt these sensors for augmented reality technology solutions. Many augmented reality solutions are based on computer vision processing methods in order to identify and track markers or other objects. However, the main problem is chaotic environment, lighting conditions where object recognition and tracking in real-time becomes difficult and sometimes is an impossible process. This paper analyses possibilities to apply microelectromechanical sensors. Additionally, it investigates quaternion use for sensor data to estimate reliable and accurate camera orientation and represent virtual content in augmented reality technology.

Papildytoji realybė yra vizualizacijos metodas, plačiai naudojamas daugybėje programų, dažniausiai sekant dirbtinius objektus, pavyzdžiui, kvadratinius žymeklius. Žymeklis leidžia nustatyti padėtį ir orientaciją erdvėje, norint įtraukti trimatį virtualų turinį į realią aplinką. Pagrindinės problemos naudojant žymeklius – vartotojų ar aplinkos objektų sukeliamos okliuzijos. Dėl šių okliuzijų virtualus trimatis turinys dažniausiai dingsta ir tai turi neigiamą įtaką programos naudojamumui. Šiame darbe atliktas vaizdo apdorojimo metodų tyrimas ir pateikta išsami požymių aptikimo ir aprašymo metodų FAST, SIFT ir SURF analizė. Šie metodai leidžia spręsti dalinės okliuzijos problemas atvaizduojant virtualų turinį tinkamoje vietoje. Atliktas stabilumo palyginimas pagal atkartojamumo kriterijų naudojant skirtingo tipo vaizdų transformacijų rinkinius. Įvertintas vaizdo apdorojimo metodų greitis ir tinkamumas naudoti su papildytosios realybės technologija. Tyrimo rezultatai svarbūs norint parinkti tinkamą vaizdo apdorojimo metodą papildytosios realybės technologijos sprendimams ir atlikti tobulinimus integruojant inercinių jutiklių informaciją. Augmented reality is a widely used visualization technique in various application fields which usually tracks artificial objects, for instance, markers that let users add 3D virtual content into the real world. The main problem with using markers is the occlusions from users or objects in the environment. These occlusions usually cause virtual content to disappear; therefore, it has negative impact to the usability of the application. For this reason the analysis of computer vision methods FAST (Feature from Accelerated Segment Test), SIFT (Scale Invariant Feature Transform) and SURF (Speeded Up Robust Features) are presented that can solve partial occlusion problems. Robustness comparison was based on a repeatability criteria using different types of image transformation sets. Speed and suitability of computer vision methods were also evaluated for augmented reality technology. The research results are essential in order to select the proper computer vision technique for augmented reality solutions and further improvements.

In this work the analysis of interaction techniques, devices and its’ possibilities were accomplished. It was determined that the problem, which many researchers tries to solve – more natural interaction between users and computers. Interaction system in augmented reality environment using Leap Motion controller was developed. To achieve this goal augmented reality NyARToolkit and Leap Motion controller libraries were used. Solution ensures extensive information about hand, fingers or finger-like tools orientation, direction, position, velocity, acceleration etc. that no other methods or devices on the market can provide. Solution also provides user interaction in augmented reality environment improved and more intuitive, that lets manipulate with virtual content as simple as possible.

Baigiamojo darbo tikslas yra sukurti manipuliatoriaus Išmanioji pirštinė modelį. Baigiamajame darbe aptariama manipuliatoriaus Išmanioji pirštinė modelio samprata, apžvelgiami ir palyginami manipuliatoriaus Išmanioji pirštinė modelio analogai bei apžvelgiami jų valdymo būdai, realizuojamas, testuojamas ir įvertinamas manipuliatoriaus Išmanioji pirštinė modelis., The Goal of graduation thesis is to create model of manipulator Smart Glove. In the graduation thesis the concept of model of manipulator Smart Glove is discussed, alternative analogs are compared and their working methods are discused, model of manipulator Smart Glove is designed, developed, tested and evaluated.