İnsansız hava aracı gözlemlerinde gerçek-zamanlı görüntü mozaiklemesi ve stabilizasyonu

Using mini Unmanned Aerial Vehicles (UAVs) equipped with camera for aerial surveillance is an gaining popularity worldwide. However severe vibrations and fast movements coupled with size and weight constraints presents a limit for effectiveness of UAV surveillance. Although there are several studies investigating image stabilization and mosaicing to provide a solution to this problem, most of them are at experimental phase requiring movement constraints or additional hardware installed on UAV. In order to provide a hardware independent solution to work at actual operational conditions, a novel real-time aerial image stabilization and mosaicing system is developed. System is developed for Baykar mini IHA which is the main UAV used by Turkish Military on rural operations. In order to achieve required standards, factors affecting the performance of real-time operation in real-world conditions were analyzed. Classifications of scenery encountered during flights and differences between infrared and day light images were investigated. A survey on current state of art registration algorithms is conducted and selected algorithms are tested in both in-door experiments and flight tests. Necessary optimizations and modifications are done in order to achieve a robust, accurate, real-time mosaicing and stabilization algorithm. Comparisons of several different approaches are done by using a novel mosaic quality measurement method employing printed high resolution images for ?Ground Data? and 5 axis CNC for positioning. Resultant methods are able to increase effectiveness of mini UAV surveillance beyond its current limitations and can be applied to any basic UAV configuration having a Ground Control Station computer.