Feature Extraction in Urban Areas Using UAV Data

As the rapid development is being focused in the urban area, there is a need for the utilisation of a system for updating this profile immediately. The usage of Uncrewed Aerial Vehicle (UAV) for mapping purposes is one of the current technologies being used in recent years. UAVs are widely used in a variety of domains due to their low price, ability to deliver very high resolution data, and ability to fly at low altitudes without being constrained by overcast weather. Typically, data extraction methods for UAVs are still quite limited, and traditional approaches are still used. For mapping applications, orthoimage features are often manually recognised and digitised using visual interpretation skills. Unfortunately, these approaches are time-consuming, costly, and repetitive. Pixel-based classification approach is frequently utilised to help extract low-level features, in which the image is categorised only based on spectral characteristics. The drawback of this approach is that the pixels in the overlapping region are misclassified as a result of class confusion. Moreover, pixel based classification performs very poorly in high resolution images. The Object-Based Image Analysis (OBIA) classification technique has large potential for automatic data extraction from Very High Resolution (VHR) images. OBIA techniques start with segmentation of image followed by classification and feature extraction using contextual information and rule base. In this study, an attempt is made to assess the capability of OBIA for detailed classification of highly dense urban areas mapped by UAV with a VHR imagery of the order better than 5 cm. The image is segmented using multiresolution image segmentation with a suitable scale, compactness and smoothness to form homogeneous image objects. Various parameters (spectral, texture, context and elevation) are computed for the VHR UAV Images. Rules are formulated to extract and categorise urban features specifically for roads and buildings. The segmented roads are classified into categories based on width and connectivity. Buildings extracted are categorised based on their elevation and size. The study efficiently demonstrates the potential of VHR orthoimage and Digital Surface Model (DSM) for urban classification using the OBIA techniques. The finest of details captured by UAV can be effectively classified using the segmentation and classification approach.