ON GEOMETRIC PROCESSING OF MULTI-TEMPORAL IMAGE DATA COLLECTED BY LIGHT UAV SYSTEMS

Abstract. Data collection under highly variable weather and illumination conditions around the year will be necessary in many applications of UAV imaging systems. This is a new feature in rigorous photogrammetric and remote sensing processing. We studied performance of two georeferencing and point cloud generation approaches using image data sets collected in four seasons (winter, spring, summer and autumn) and under different imaging conditions (sunny, cloudy, different solar elevations). We used light, quadrocopter UAVs equipped with consumer cameras. In general, matching of image blocks collected with high overlaps provided high quality point clouds. All of the before mentioned factors influenced the point cloud quality. In winter time, the point cloud generation failed on uniform snow surfaces in many situations, and during leaf-off season the point cloud generation was not successful over deciduous trees. The images collected under cloudy conditions provided better point clouds than the images collected in sunny weather in shadowed regions and of tree surfaces. On homogeneous surfaces (e.g. asphalt) the images collected under sunny conditions outperformed cloudy data. The tested factors did not influence the general block adjustment results. The radiometric sensor performance (especially signal-to-noise ratio) is a critical factor in all weather data collection and point cloud generation; at the moment, high quality, light weight imaging sensors are still largely missing; sensitivity to wind is another potential limitation. There lies a great potential in low flying, low cost UAVs especially in applications requiring rapid aerial imaging for frequent monitoring.