Accurate and up-to-date road information is essential for both effective urban planning and disaster management. Today, very high resolution (VHR) imagery acquired by airborne and spaceborne imaging sensors is the primary source for the acquisition of spatial information of increasingly growing road networks. Given the increased availability of the aerial and satellite images, it is necessary to develop computer-aided techniques to improve the efficiency and reduce the cost of road extraction tasks. Therefore, automation of image-based road extraction is a very active research topic. This thesis deals with the development and implementation aspects of a semi-automatic road extraction strategy, which includes two key approaches: multidirectional and single-direction road extraction. It requires a human operator to initialize a seed circle on a road and specify a extraction approach before the road is extracted by automatic algorithms using multiple vision cues. The multidirectional approach is used to detect roads with different materials, widths, intersection shapes, and degrees of noise, but sometimes it also interprets parking lots as road areas. Different from the multidirectional approach, the single-direction approach can detect roads with few mistakes, but each seed circle can only be used to detect one road. In accordance with this strategy, a RoadModeler prototype was developed. Both aerial and GeoEye-1 satellite images of seven different types of scenes with various road shapes in rural, downtown, and residential areas were used to evaluate the performance of the RoadModeler. The experimental results demonstrated that the RoadModeler is reliable and easy-to-use by a non-expert operator. Therefore, the RoadModeler is much better than the object-oriented classification. Its average road completeness, correctness, and quality achieved 94%, 97%, and 94%, respectively. These results are higher than those of Hu et al. (2007), which are 91%, 90%, and 85%, respectively. The successful development of the RoadModeler suggests that the integration of multiple vision cues potentially offers a solution to simple and fast acquisition of road information. Recommendations are given for further research to be conducted to ensure that this progress goes beyond the prototype stage and towards everyday use.

Accurate and up-to-date road information is essential for both effective urban planning and disaster management. Today, very high resolution (VHR) imagery acquired by airborne and spaceborne imaging sensors is the primary source for the acquisition of spatial information of increasingly growing road networks. Given the increased availability of the aerial and satellite images, it is necessary to develop computer-aided techniques to improve the efficiency and reduce the cost of road extraction tasks. Therefore, automation of image-based road extraction is a very active research topic. This thesis deals with the development and implementation aspects of a semi-automatic road extraction strategy, which includes two key approaches: multidirectional and single-direction road extraction. It requires a human operator to initialize a seed circle on a road and specify a extraction approach before the road is extracted by automatic algorithms using multiple vision cues. The multidirectional approach is used to detect roads with different materials, widths, intersection shapes, and degrees of noise, but sometimes it also interprets parking lots as road areas. Different from the multidirectional approach, the single-direction approach can detect roads with few mistakes, but each seed circle can only be used to detect one road. In accordance with this strategy, a RoadModeler prototype was developed. Both aerial and GeoEye-1 satellite images of seven different types of scenes with various road shapes in rural, downtown, and residential areas were used to evaluate the performance of the RoadModeler. The experimental results demonstrated that the RoadModeler is reliable and easy-to-use by a non-expert operator. Therefore, the RoadModeler is much better than the object-oriented classification. Its average road completeness, correctness, and quality achieved 94%, 97%, and 94%, respectively. These results are higher than those of Hu et al. (2007), which are 91%, 90%, and 85%, respectively. The successful development of the RoadModeler suggests that the integration of multiple vision cues potentially offers a solution to simple and fast acquisition of road information. Recommendations are given for further research to be conducted to ensure that this progress goes beyond the prototype stage and towards everyday use.

The political history of U.S. commercial remote sensing began in 1984 when the U.S. government first attempted to commercialize its civil earth observation satellite system " Landsat. Since then, the high technology of earth imaging satellite systems has generated intense debates and policy conflicts, primarily centered on U.S. government concerns over the national security and foreign policy implications of high-resolution commercial satellite systems. Conversely, proponents of commercial observation satellites have urged U.S. policymakers to recognize the scientific and socio-economic utility of commercial remote sensing and thus craft and implement regulatory regimes that allow for a greater degree of information openness and transparency in using earth observation satellite imagery. This dissertation traces and analyzes that tumultuous political history and examines the policy issues and social construction of commercial remote sensing to determine the role of knowledge in the effective crafting and execution of commercial remote sensing laws and policies. Although individual and organizational perspectives, interests, missions, and cultures play a significant role in the social construction of commercial observation satellite systems and programs, the problem of insufficient knowledge of the myriad dimensions and complex nature of commercial remote sensing is a little studied but important component of this social construction process. Knowledge gaps concerning commercial remote sensing extend to various dimensions of the subject matter, such as the global, economic, technical, and legal/policy aspects. Numerous examples of knowledge voids are examined to suggest a connection between deficient knowledge and divergent policy perceptions as they relate to commercial remote sensing. Relevant knowledge voids are then structurally categorized to demonstrate the vastness and complexity of commercial remote sensing policy issues and to offer recommendations on how to fill such knowledge gaps to effect increased collaboration between the US government and the U.S. commercial remote sensing industry. Finally, the dissertation offers suggestions for future STS studies on policy issues, particularly those that focus on the global dimensions of commercial remote sensing or on applying the knowledge gap concept advanced by this dissertation to other areas of science and technology policymaking. / Ph. D.

Thesis (MSc)--Stellenbosch University, 2015. / ENGLISH ABSTRACT: Urban planning and management require up-to-date information about urban land cover. Producing such geospatial information is time consuming as it is usually done manually. The classification of such information from satellite imagery is challenging owing to the difficulties associated with distinguishing urban features having similar spectral properties. Therefore, this study evaluates the combination of a digital surface model (DSM) derived from LiDAR data and very high-resolution GeoEye-1 satellite imagery for classifying urban land cover in Cape Town. The value of the DSM was assessed by comparing a land cover product obtained from the GeoEye-1 image to a map produced using both the GeoEye-1 image and the DSM. A systematic segmentation procedure for the two classifications scenarios preceded a supervised (using a support vector machine, K nearest neighbour and classification and regression algorithm tree classifiers) and rule-based classification. The various approaches were evaluated using a combination of methods. When including the DSM in the supervised and rule-based classifications, the overall accuracy and kappa vary between 80% to 83% and 0.74 to 0.77 respectively. When the DSM is excluded, the overall accuracy ranges between 49 to 64% whereas kappa ranges between 0.32 to 0.53 for the two classification approaches. The accuracies obtained are always about 20% higher when the DSM is included. The normalised DSM (nDSM) enabled accurate discrimination of elevated (e.g. buildings) and non-elevated (e.g. paved surfaces) urban features having similar spectral characteristics. The nDSM of at least one-metre resolution and one metre vertical accuracy influenced the accuracy of the results by correctly differentiating elevated from non-elevated. The rule-based approach was more effective than the supervised classification, particularly for extracting water bodies (dams and swimming pools) and bridges. Consequently, a rule-based approach using very high spatial resolution (EHSR) satellite imagery and a LiDAR-derived DSM is recommended for mapping urban land cover. / AFRIKAANSE OPSOMMING: Stedelike beplanning- en bestuur vereis dat inligting oor grondbedekking (land cover) op datum moet wees. Die vervaardiging van hierdie georuimtelike inligting is tydrowend omdat dit gewoonlik met die hand gedoen word. Die onttrekking van sulke inligting vanuit satellietbeelde bied ʼn groot uitdaging omdat stedelike voorwerpe met soortgelyke spektrale eienskappe moeilik is om van mekaar te onderskei. Hierdie studie evalueer die kombinasie van ʼn digitale oppervlak model (DOM) afkomstig van LiDAR-data en ʼn baie hoë resolusie GeoEye-1-satellietbeeld om stedelike grondbedekking in Kaapstad te klassifiseer. Die waarde van die DOM word bepaal deur ʼn grondbesettingsproduk wat vanuit ʼn GeoEye-1-beeld verkry is te vergelyk met ʼn grondbesettingsproduk wat verkry is deur beide die GeoEye-1-beeld en die DOM te gebruik. Sistematiese segmentasie word op die twee benaderings uitgeoefen en dit word gevolg deur ʼn gekontroleerde klassifikasie (steunvektormasjiene, k-naaste aangrensende waarde en klassifikasie en regressie algoritme) en ʼn reël-gebaseerde algoritme. Hierdie verskeie benaderings is geëvalueer met behulp van ʼn kombinasie van kwalitatiewe en kwantitatiewe metodes. Toe die DOM in die gekontroleerde en reël-gebaseerde klassifikasie ingesluit is, het die algehele akkuraatheid en kappa tussen 80% en 83%, en 74% en 77% gewissel. Toe die DOM uitgesluit is, het die algehele akkuraatheid en kappa tussen 49% en 64%, en 32% en 53% vir die twee klassifikasiebenaderings gewissel. Die behaalde akkurraatheidswaardes is altyd 20% hoër as die DOM ingesluit word. Dit is hoofsaaklik omdat die DOM akkurate onderskeiding tussen hoë (bv. geboue) en plat (bv. geplaveide oppervlaktes) stedelike bakens met gelyksoortige spektrale eienskappe in staat stel. Die kwaliteit van die DOM beïnvloed die akkuraatheid van die resultate. ʼn DOM van ten minste een meter resolusie, met een meter of beter vertikale akkuraatheid, word benodig om te verseker dat geboue en ander beboude bakens korrek van mekaar onderskei kan word. Die reël-gebaseerde benadering was meer effektief as die gekontroleerde klassifikasie, veral om waterliggame (damme en swembaddens) en brûe te identifiseer. Gevolglik word ʼn reël-gebaseerde benadering met die hoë resolusie satellietbeelde en ʼn LiDAR-afgeleide DOM aanbeveel om stedelike grondbesetting te karteer.