**4. Methods**

Study of Kibera informal settlement has two main aims: to derive detailed land use/land cover map that can supply population estimation, and to analyse the settlement growth and changes between 2006 and 2009.

Extracting data of urban land use structure from remote sensing imagery require methods that are able to provide appropriate level of details observed. Object based classification has been successfully implemented to obtain land cover information from urban VHR remote sensing applications. Thus, this approach was selected in the land use/cover classification of Kibera settlement with main aim to delimitate well residential objects from open areas, and potentially to obtain informal settlement structure in the microstructure level (distinguish individual houses). In addition to determination of detailed urban structures we were also interested in locating the step-wise expansion of informal residential areas, which was analyzed through comparison of images taken in different time using pixel-based multilevel image differencing approach.

include various Kibera specific phenomena mappings (www.mappingnobigdeal.com), though the assessment of potential of VHR satellite imagery for mapping purposes presents

Six VHR satellite images were available for our research (Table 1): one GeoEye image and five QuickBird images. Satellite images were partly (pre)processed. This means images were roughly georeferenced and corrected for sensor radiometry, also pan-sharpened, and

2006-03-27 QuickBird R-G-B 0.6 (pansharpened) minor Change detection

2008-08-10 QuickBird R-G-B 0.6 (pansharpened) present Change detection 2009-07-25 GeoEye R-G-B 0.5 (pansharpened) free Land use/cover

Besides different inherent spatial resolution the main differences among GeoEye and QuickBird images were sensor viewing angles, causing higher objects roof prints and shadows to have different positions among images. As Kibera informal settlement lies in a hilly terrain, the positional accuracy fit of geographical entities among images was not reached because much of distortion comes from the terrain as well. For the study no digital elevation model was available, thus ortorectification was not possible. However, GPS field

Study of Kibera informal settlement has two main aims: to derive detailed land use/land cover map that can supply population estimation, and to analyse the settlement growth and

Extracting data of urban land use structure from remote sensing imagery require methods that are able to provide appropriate level of details observed. Object based classification has been successfully implemented to obtain land cover information from urban VHR remote sensing applications. Thus, this approach was selected in the land use/cover classification of Kibera settlement with main aim to delimitate well residential objects from open areas, and potentially to obtain informal settlement structure in the microstructure level (distinguish individual houses). In addition to determination of detailed urban structures we were also interested in locating the step-wise expansion of informal residential areas, which was analyzed through comparison of images taken in different time using pixel-based multi-

**coverage** 

**Analysis performed** 

Change detection

one of the recent examinations of their use for Kibera community.

**Date Sensor Bands used Spatial resolution Cloud** 

2006-07-31 QuickBird R-G-B 0.6 (pansharpened) free 2007-01-22 QuickBird R-G-B 0.6 (pansharpened) present 2008-01-07 QuickBird R-G-B 0.6 (pansharpened) free

Table 1. List of available satellite images and their main characteristics.

walks tracks were available for the main roads and path-network in the area.

**3.3 Available VHR satellite data** 

**4. Methods** 

changes between 2006 and 2009.

level image differencing approach.

provided as a stack of three visible bands only.

Object based classification of the Kibera informal settlement was performed on GeoEye image since its characteristics (close to nadir viewing angle, good spatial resolution, and fine contrast) were most promising to obtain adequate details on object recognition within the informal settlement area. Rooftops are covered with different materials, ranging from new to rusty sheets, bricks and other materials, each of them having specific reflectance characteristics (spectral representation) on satellite image (Fig. 3, Fig. 6a). For population estimation study we need to differentiate well rooftops, unpaved roads and non-build land and therefore discriminate residential areas from open soils, respectively. Object based segmentation automatically delimits satellite image into homogeneous elements (segments), where close correspondence to the real (geographical) objects on the Earth's surface is expected. Usage of thus obtained image elements (segments) has a number of benefits, one of them is ability to incorporate spatial and contextual information such as size, shape, texture and topological relationships (Blaschke et al., 2004; Benz et al., 2004) in contextual classification. In the stage of classification all these segments are classified according to their attributes into most appropriate classes (representing various geographical objects under study consideration), while obtaining detailed classification of urban area land cover/use.

Fig. 3. Examples of rooftops, rooftops renovations and buildings constructions on VHR satellite imagery from three different dates.

With object-based analysis on rooftops morphology attributes we expected to improve the assessment of the potential population in slum areas. Since no complete and relevant field survey (official census) was recently performed, different density parameters were tested to approach the potential population and compared to other available population assessments.

#### **4.1 Data pre-processing and preparation**

Data preprocessing is important procedure in remote sensing technology. It meets issues that have to be carefully understood and solved before any data analysis process starts. In order to be able to compare satellite images taken for the same scene at different acquisition dates they have to be co-registered and radiometrically adjusted. Recent automatic registration algorithms can accomplish the task well when similar acquisition geometry among sensor systems is provided. Global geometric transformations are mostly appropriate for positional corrections in such cases. However, this was not the case with the imagery obtained for Kibera study (see section 3.3).

Obtained images were rectified but not precisely aligned one to another. Due to agitated terrain in Kibera and lack of any digital elevation model, semi-automated rigorous

Object-Based Image Analysis of VHR Satellite Imagery for

subset only.

**4.2 Land cover classification** 

belonging object class.

are shown on Fig.6.

Population Estimation in Informal Settlement Kibera-Nairobi, Kenya 415

the second approach is that it tries to globally adjust the given (to-be-adjusted) image or a chosen area with the reference image or a subset through a statistical approach. We applied linear regression for the relative adjustment of spectral bands between the images. Relative radiometric normalisation was done through local adjustments of QuickBird images onto GeoEye reference image, for Kibera settlement with 30 m buffer

Since GeoEye image of the whole Kibera informal settlement contains lots of information and the analysis of the total area would be too demanding in terms of computer processing, we divided image of Kibera into 12 smaller parts (according to 12 informal villages). The complete process of segmentation and classification was applied systematically for each informal village separately, with same parameter settings at each phase. This was possible due to relatively homogenous landscape over Kibera settlement. Thus we obtained 12 regional classification outputs, which were then merged in the final stage. To avoid erroneous classification on the edges of the splitted images, we applied 30 meter buffer

Object based classification consists of two stages. Image is first segmented into a set of segments (regions) that are considered to be homogeneous in terms of one or more spectral or spatial properties. Then follows classification where each segment is classified into

Supervised segmentation within software used (ENVI EX, Feature Extraction module) is defined by two segmentation parameters that influence an average size of segments: segmentation and merging. Setting different values for these two parameters causes change of size of segments, allowing for an image to be segmented at many different scales, so both parameter values influence classification results. Since structure of the Earh's surface is similar throughout the whole settlement, same general segmentation parameters were used for each of the 12 villages. Visual example of segment structures

While classifying Raila village we adapted segmentation parameters to best extract shapes of individual buildings inside informal settlement. Since segmentation parameters were adequate for one particular land use only (i.e. buildings), others were expected to be underor over-segmented. There is no single "optimal" scale for analysis of remote sensing images, rather there are many optimal scales that are specific to the image-objects that exist within a scale (Hay et al., 2003) and this is why using a multi-scale approach may often be preferable (Johnson and Xie, 2011). All spectral bands were used and given equal weight for image

segmentation and all available attributes were calculated for all segments.

when masking the village fragments out from the whole GeoEye image (Fig. 5).

Fig. 5. A buffer of 30 meters around the village border.

procedures of co-registration could not be applied. The non‐linear rubber‐sheeting method was the only possibility to obtain mutually aligned images. This procedure is effective, but very time consuming, since it demands manual selection of hundreds of control points for each image.

Fig. 4. Selecting the control points for rubber-sheeting method for image geo-referencing (AutoSync module of ERDAS Imagine).

GeoEye image taken on 25th of July 2009 was selected as the reference image considering its highest spatial resolution, good matching with GPS path-network tracks and the fact it is most recent. Then QuickBird images selected for analysis were manually registered to the reference, based on cca. 1,400 manually selected control points per image and using a piecewise transformation based on triangles formed from the tie points (Fig. 4). Resampling was nearest neighbour. An average RMSE is not reported as this is local approximation technique. Geo-corrected images were evaluated through detailed visual control.

Geometrically matched GeoEye and QuickBird images were then used for OBIA. Finally three images were selected for change detection due to their best results in geo-correction phase: GE2009-07-25, QB2008-08-10 and QB2006-03-27. For change detection analysis images were resampled to uniform 1 m resolution, to be prepared for radiometric standardisation.

After geometric adjustments there are still differences amongst the spectral properties of satellite images (spectral bands from the same or different images are not adjusted to each other). Hence, before pixel-based image comparisons (image differencing) the radiometric standardisation is needed. Most standardisation procedures derive from adjustments of invariant objects or from the least squares method (linear regression). The problem with the first method is that invariant areas should be verified with field measurements. Furthermore, the generally recognised invariant objects, such as deserts or light sandy beaches, do not come into play, since they cannot be found in Kibera area. The principle of the second approach is that it tries to globally adjust the given (to-be-adjusted) image or a chosen area with the reference image or a subset through a statistical approach. We applied linear regression for the relative adjustment of spectral bands between the images. Relative radiometric normalisation was done through local adjustments of QuickBird images onto GeoEye reference image, for Kibera settlement with 30 m buffer subset only.
