Buildings are not only central to the day-to-day activities but also serve as critical indicators of urban development and transformation. The automatic extraction of building footprints from high-resolution Remote Sensing Imagery (RSI) has emerged as an important and popular tool in urban studies. It helps to enhance the understanding and management of urban sprawl, urban planning, population estimation, resource allocation, and post-disaster damage assessment. In this context, having an automated and robust building detection model is crucial. Deep Learning (DL) model and Object-Based Image Analysis (OBIA) techniques are the main and commonly used for automated building detection. This study investigates the efficacy of a pre-trained DL model and a rule-based model OBIA techniques in building detection across varied resolutions and geographic settings. Employing orthophotos from Luleå, Gävle, and Stockholm, the research assesses the adaptability and robustness of these methods under image properties and urban densities. The DL model was initially trained on 0.25m resolution data of Sweden by Lantmäteriet (Sweden mapping agency). The rule-based model was developed by applying OBIA techniques on behalf of this study. Models were analyzed through six feature agreement statistics including Critical Success Index (CSI), Precision, and Detection Probability (POD). The findings reveal that the DL model consistently outperformed the OBIA approach across all study areas, particularly at the original 25 cm resolution. Gävle showed superior precision with a CSI of 0.8139 for the DL model against a CSI of 0.7493 for OBIA at 25 cm.  The evaluation was improved by considering 50*50 sq. m subsets and building sizes. These evaluations highlight that building size and urban density significantly influence detection accuracy. Larger (> 2500 sq. m) buildings and less dense areas tend to yield higher accuracy across both detection methods. The DL model exhibited high CSI values for very large buildings (>5500 sq. m) in Gävle, surpassing 0.8, while the detection of very small (< 50 sq. m) buildings remained challenging for both methods. Overall, the pre-trained DL model is very sensitive to resolution changes compared to OBIA. Importantly, both give their best performance at the original resolution while DL is superior than OBIA. A rule-based OBIA model is affected by the geographical characteristics more heavily than a DL model. Both models have their best performance in the area with medium building density when medium to very large buildings exist. This study highlights how big the impact of building size, geographic characteristics, and image resolution on the performance of DL and OBIA techniques. However, further investigation is recommended to draw a strong conclusion regarding the impact of resolution on the model performance.

In this thesis, an approach to automatically derive information about land cover from the remotely sensed data is presented. The data interpretation was done with classification process and performed in software eCognition Developer. The Object-based image analysis, which assignes the classes - for example land cover types, to clusters of pixels (=objects), was used. For the classification, products of two different data sources were combined - the orthophotos generated from aerial imagery and Normalized Digital surface model derived from LiDAR data. Five types of landscape elements were identified and classified.

Zur Verminderung des Rauschens sehr hochauflösender Satellitenbilder existieren eine Vielzahl von Glättungsalgorithmen. Die Wirkung verschiedener Tiefpaß- und kantenerhaltender Filter auf das Verhalten eines objektorientierten Segmentierungsverfahrens wird anhand zweier synthetischer Grauwertbilder und einer IKONOS-Aufnahme untersucht. Als Rauschmaß hat sich ein modifiziertes, ursprünglich von Baltsavias et al. [2001] vorgeschlagenes Verfahren bewährt, in dem je Grauwert nur die Standardabweichungen der gleichförmigsten Gebiete berücksichtigt werden. In Vergleich mit synthetisch verrauschten Bildern zeigt sich jedoch, daß auf diese Weise das Rauschen im Bild systematisch um fast den Faktor zwei unterschätzt wird. Einfache Filter wie Mittelwertfilter und davon abgeleitete Verfahren verschlechtern die Präzision der Objekterkennung dramatisch, kantenerhaltende Filter können bei stärker verrauschten Daten vorteilhaft sein.Als bester Filter, der bei Ansprüchen an präzise Segmentgrenzen im Pixelbereich sinnvoll einzusetzen ist und dabei mit nur einem Parameter gesteuert werden kann, erweist sich der modifizierte EPOS-Filter, ursprünglich vorgestellt von Haag und Sties [1994, 1996]. Die generellen Bildparameter, wie Standardabweichung oder Histogramm werden durch diesen kantenerhaltenden Filter nur unwesentlich beeinflußt.

Land managers need to rapidly assess vegetation composition and bare ground to effectively evaluate, manage, and restore shrub steppe communities that have been encroached by pinyon and juniper (P-J) trees. A major part of this process is assessing where to apply mechanical and prescribed fire treatments to reduce fuel loads and maintain or restore sagebrush steppe rangelands. Geospatial technologies, particularly remote sensing, offers an efficient option to assess rangelands across multiple spatial scales while reducing the need for ground-based sampling measurements. High-spatial resolution color-infrared imagery (0.06-m pixels) was acquired for sagebrush steppe communities invaded by P-J trees at five sites in Oregon, California, Nevada, and Utah with a Vexcel Ultra CamX digital camera in June/July 2009. In addition to untreated P-J woodlands, imagery was acquired over P-J woodlands where fuels were reduced by either prescribed fire, tree cutting, or mastication treatments. Ground measurements were simultaneously collected at each site in 2009 on 0.1-hectare subplots as part of the Sagebrush Steppe Treatment Evaluation Project (SageSTEP). We used Trimble eCognition Developer to 1) develop efficient methods to estimate land cover classes found in P-J woodlands; 2) determine the relationship between ground measurements and object-based image analysis (OBIA) land cover measurements for the following classes: trees (live, burned, cut, and masticated), shrubs, perennial herbaceous vegetation, litter (including annual species), and bare ground; and 3) evaluate eCognition rule-sets (models) across four spatial scales (subplot, site, region, and network) using untreated P-J woodland imagery. At the site scale, the overall accuracy of our thematic maps for untreated P-J woodlands was 84% with a kappa statistic of 0.80. For treatments, the overall accuracy and kappa statistic for prescribed fire was 85% and 0.81; cut and fell 82% and 0.77, and mastication 84% and 0.80, respectively, each indicating strong agreement between OBIA classification and ground measured data. Differences between mean cover estimates using OBIA and ground-measurements were not consistently higher or lower for any land cover class and when evaluated for individual sites, were within 5% of each other; all regional and network OBIA mean cover estimates were within 10% of the ground measurements. The trade-off for decreased precision over a larger area (region and network scale) may be useful to prioritize fuel-management strategies but will unlikely capture subtle shifts in understory plant communities that site and subplot spatial scales often capture. Although cover assessments from OBIA differed somewhat from ground measurements, they were accurate enough for many landscape-assessment applications such as evaluating treatment success and assessing the spatial distribution of fuels following fuel-reduction treatments on a site scale.

No description available.

No description available.