Object-Based Image Analysis (OBIA) has been successfully used to map slums. In general, the occurrence of uncertainties in producing geographic data is inevitable. However, most studies concentrated solely on assessing the classification accuracy and neglecting the inherent uncertainties. Our research analyses the impact of uncertainties in measuring the accuracy of OBIA-based slum detection. We selected Jakarta as our case study area because of a national policy of slum eradication, which is causing rapid changes in slum areas. Our research comprises of four parts: slum conceptualization, ruleset development, implementation, and accuracy and uncertainty measurements. Existential and extensional uncertainty arise when producing reference data. The comparison of a manual expert delineations of slums with OBIA slum classification results into four combinations: True Positive, False Positive, True Negative and False Negative. However, the higher the True Positive (which lead to a better accuracy), the lower the certainty of the results. This demonstrates the impact of extensional uncertainties. Our study also demonstrates the role of non-observable indicators (i.e., land tenure), to assist slum detection, particularly in areas where uncertainties exist. In conclusion, uncertainties are increasing when aiming to achieve a higher classification accuracy by matching manual delineation and OBIA classification.

Object-Based Image Analysis (OBIA) has been successfully used to map slums. In general, 12the occurrence of uncertainties in producing geographic data is inevitable. However, most studies

High temporal and spatial-resolution imageries are a valuable data source for slum monitoring. However, the transferability of OBIA methods across space and time remains problematic, due to the complexity of the term "slum". Hence, transparency is important when analysing the transferability of OBIA methods for slum mapping. Our research developed a framework for measuring the temporal transferability of OBIA methods employing the trajectory error matrix (TEM). We found relatively low trajectory accuracies indicating low temporal transferability of OBIA methods for slum monitoring using point-based assessment methods. However, the analysis of change needs to be combined with an analysis of the certainty of this change by considering the context of the change to deal with common problems such as variations of the viewing angles and uncertainties in producing reference data on slums.

The Generic Slum Ontology (GSO) was developed to assist the detection of slums using Geographic Object-Based Image Analysis (GEOBIA). When applying the GSO locally, uncertainties exist in slum detection and transferability. Slums often have fuzzy boundaries and different ways to conceptualise. This study focuses on inherent uncertainties when analysing the transferability of the GSO across space, time and conceptualizations in the city of Jakarta, Indonesia. To measure the transferability of the GSO, we developed quantitative and qualitative indicators in multi-temporal Pleiades imagery (2012-2015) of two purposely-selected subsets. This framework allows assessing whether the developed ruleset is transferable across different spatial and temporal images. We applied two classification stages: background removal with a low scale parameter (SP) followed by slum extraction with a coarser SP. Both quantitative and qualitative indicators showed limited spatial and temporal transferability. Three sources of uncertainties can explain this result. First, the static concept of the employed ruleset and dynamic changes of slums. Real-world objects evolve over time, but their description remains static. Second, the gap between the real world (subjective conceptualization of objects) and image domain (quantitative values). For instance, the roof materials of slums (i.e. asbestos) have a similar spectral property with parking lot (from concrete), which resulted in misclassification. Third, the use of references data from local experts and municipal data introduce uncertainties that related to local ground knowledge and politics of slum declarations. Thus, this research contributes to the development of transferability measurements for the GSO and the understanding of underlying uncertainties.

Mapping slums is vital for monitoring the Sustainable Development Goal (SDG) indicators. In the absence of reliable data, Remote Sensing (RS)-based approaches, particularly the Deep Learning (DL) methods, have gained recognition and high accuracies for slum mapping. However, using RS alone has its limitation in complex urban environments. Previous studies showed the added value of combining ground-level information with RS. Therefore, this research aims to integrate Remote Sensing Imagery (RSI) and Street View Images (SVI) for slum mapping. Jakarta city is the study area representing the challenge of distinguishing between slum and non-slum kampungs, and these kampungs accommodate approximately 60% of the population of Jakarta. This research compares the mapping results obtained by four DL networks: FCN-DK6 used only RSI, a VGG16 used only SVI, and two networks combined RSI and SVI (FCN-DK6-i and Modified FCN-DK6). Further, the Modified FCN-DK6 network was explored by integrating SVI at each convolutional layer, i.e., Modified FCN-DK6_1, Modified FCN-DK6_2, Modified FCN-DK6_3, Modified FCN-DK6_4, and Modified FCN-DK6_5. Experimental results demonstrate that combining RSI and SVI improves the accuracy, depending on how and at what level in the FCN network they are integrated. The Modified FCN-DK6_2 outperforms the rest in Modified FCN-DK6 experiments and FCN-DK6-i.

How to cite (APA 6th Style): Pratomo, J., and Widiastomo, T. (2016). Implementation of the Markov Random Field for urban land cover classification of UAV VHIR data.