ABSTRACT:In this work, an automated approach for 3D building roof modelling is presented. The method consists of two main parts, namely roof detection and 3D geometric modelling. For the detection, a combined approach of four methods achieved the best results, using slope-based DSM filtering as well as classification of multispectral images, elevation data and vertical LiDAR point density. In the evaluation, the combination of the four methods yields 94% correct detection at an omission error of 12%. Roof modelling is done by plane detection with RANSAC, followed by geometric refinement and merging of neighbouring segments to clean up oversegmentation. Walls are then detected and excluded, and the roof shapes are vectorised with the alpha-shape method. The resulting polygons are refined using 3D straight edges reconstructed by automatic straight edge extraction and matching, as well as 3D corner points constructed by intersection of the 3D edges. The results are quantitatively assessed by comparing to ground truth manually extracted from high-quality images, using several metrics for both the correctness and completeness of the roof polygons and for their geometric accuracy. The median value of correctness of the roof polygons is calculated as 96%, while the median value of completeness is 88%.
Commission I, WG I/V KEY WORDS: Adjustment, Matching, Photogrammetry, Point Cloud, Surface, Test, UAVs ABSTRACT:Nowadays, small size UAVs (Unmanned Aerial Vehicles) have reached a level of practical reliability and functionality that enables this technology to enter the geomatics market as an additional platform for spatial data acquisition. Though one could imagine a wide variety of interesting sensors to be mounted on such a device, here we will focus on photogrammetric applications using digital cameras. In praxis, UAV-based photogrammetry will only be accepted if it a) provides the required accuracy and an additional value and b) if it is competitive in terms of economic application compared to other measurement technologies. While a) was already proven by the scientific community and results were published comprehensively during the last decade, b) still has to be verified under real conditions. For this purpose, a test data set representing a realistic scenario provided by ETH Zurich was used to investigate cost effectiveness and to identify weak points in the processing chain that require further development. Our investigations are limited to UAVs carrying digital consumer cameras, for larger UAVs equipped with medium format cameras the situation has to be considered as significantly different. Image data was acquired during flights using a microdrones MD4-1000 quadrocopter equipped with an Olympus PE-1 digital compact camera. From these images, a subset of 5 images was selected for processing in order to register the effort of time required for the whole production chain of photogrammetric products. We see the potential of mini UAV-based photogrammetry mainly in smaller areas, up to a size of ca. 100 hectares. Larger areas can be efficiently covered by small airplanes with few images, reducing processing effort drastically. In case of smaller areas of a few hectares only, it depends more on the products required. UAVs can be an enhancement or alternative to GNSS measurements, terrestrial laser scanning and ground based photogrammetry. We selected the above mentioned test data from a project featuring an area of interest within the practical range for mini UAVs. While flight planning and flight operation are already quite efficient processes, the bottlenecks identified are mainly related to image processing. Although we used specific software for image processing, the identified gaps in the processing chain today are valid for most commercial photogrammetric software systems on the market. An outlook proposing improvements for a practicable workflow applicable in projects in private economy will be given.
A challenging problem in the field of avian ecology is deriving information on bird population movement trends. This necessitates the regular counting of birds which is usually not an easily-achievable task. A promising attempt towards solving the bird counting problem in a more consistent and fast way is to predict the number of birds in different regions from their photos. For this purpose, we exploit the ability of computers to learn from past data through deep learning which has been a leading sub-field of AI for image understanding. Our data source is a collection of on-ground photos taken during our long run of birding activity. We employ several state-of-the-art generic object-detection algorithms to learn to detect birds, each being a member of one of the 38 identified species, in natural scenes. The experiments revealed that computer-aided counting outperformed the manual counting with respect to both accuracy and time. As a real-world application of image-based bird counting, we prepared the spatial bird order distribution and species diversity maps of Turkey by utilizing the geographic information system (GIS) technology. Our results suggested that deep learning can assist humans in bird monitoring activities and increase citizen scientists’ participation in large-scale bird surveys.
In this study, we have evaluated the use of unmanned air vehicle (UAV) photogrammetry for the monitoring of a wheat experiment under field conditions, filtered a digital surface model (DSM) to derive the wheat plant heights, and compared the results with the field measurements. The images were acquired with the use of a low-cost UAV Walkera QR350 and GoProHero3+ action camera in May 2015. In total, 477 images were acquired for quality assessment of the proposed method, and a reference dataset was collected with terrestrial fieldwork. For a comparison of field measurements with DSM-derived plant heights, the maximum calculated plant height in the plot was selected. The mean, median, and standard deviation were calculated as 4.66, 3.75, and 13.78 cm. Regarding the statistical t-test between the field measurements and plant heights from the DSM, the t-value was calculated as 1.82, and the p-value was 0.071. Because the t-value was larger than 0.50, the values between the traditional method and our approach were highly correlated.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.