Goals of the task
The task is responsible for the processing of the spatial indices and USTs defined in Task 2.1 and 2.3. It is aimed to develop a semi-automatic workflow. The accuracy of the results is assessed with reference data collected in Task 2.2 and shortly demonstrated below. The results are provided to all project partners and local stakeholders in the case cities. They are also used to correlate and to spatially disaggregate the data collected in WP1 and WP3, and serve as important base dataset of the RP methodology (WP5).
Work done to achieve the goals
U-Tueb developed and continuously optimizes an object-based rule-set to classify LULC from satellite images, to extract building footprints and to identify USTs. The rule-set utilizes spectral information as well as relational parameters. Initially, it was not foreseen to derive information for the entire administrative areas, but only for the densely built-up parts of the case cities. Since several RP partners asked for an estimation of the number of dwellings for the entire city, U‑Tueb developed an approach for the peri-urban and rural parts of the case cities as well using medium resolution RapidEye data. Therefore, RapidEye satellite images with a resampled resolution of 5 m (originally 6.5 m) were acquired to cover the peri-urban and rural areas. Template matching methods were used to refine the results. Here, only the detection and counting of dwellings are relevant, since we assume relatively uniform building types in these areas.
For the densely built-up areas, U-Tueb utilizes high-resolution Pléiades data. With the object-based workflow LULC is classified first, then the building footprints are extracted, spatial indices calculated and processed to USTs. The Pléiades results were combined with the results of the RapidEye dataset. For all building blocks, indices/ parameters were used to process USTs.
The building footprints derived of the LULC result for the area covered by the Pléiades satellite scene were used to identify building types. A preliminary classification was conducted by applying thresholds to the building footprints. The area of the footprints was used, as well as the relation to the USTs. In a second step, a visual inspection of the results was conducted. If necessary, building classes were refined manually. Other LULC derived information like traffic infrastructure, vegetation cover and agriculture is handed to the respective tasks in WP1 and WP3.
The analyses are finished for the City of Kigali. Currently, the calculation of the USTs and building archetypes for Da Nang is in progress.
Results so far
The LULC was processed for Da Nang and Kigali (Fig. 23). The combined use of very high resolution Pléiades and high-resolution RapidEye images led to satisfying results concerning the identification of urban structures and the number of built-up structures. A transferable methodology to derive USTs of urban indices was developed. The USTs are closely related to the Local Climate Zones (LCZ) definition. For Kigali, nine classes were identified (Fig. 24 & Fig. 25).
Fig. 23: RapidEye – based LULC classification of Kigali
Fig. 24: Support Vector Machine – based UST classification result with nine training classes as input parameters
Besides the USTs identification, U-Tueb derived the buildings types of the LULC dataset. The Pléiades image covers the central and densely built-up part of the city. 151,360 built-up structures were identified and were classified to 10 building archetype classes with an overall accuracy of 91 % (Fig. 26 & Fig. 27). A RapidEye image was used to identify buildings in the sparsely and mostly rural areas of Kigali. 92,712 were identified with an overall accuracy of 83.6 %. An overall number of 244,072 built-up structures were identified by the combined methodology. A detailed description of the methodology and the results can be found in the combined deliverable D-2.7/ 2.8.
Fig. 25: USTs for the central part of Kigali
Fig. 26: Share of building types per building blocks of Kigali
Fig. 27: Result of building type classification for single building objects in Kigali