Urban expansion has created significant challenges in transportation planning, infrastructure management, and spatial analysis, particularly in developing cities. This study explores the application of Support Vector Machine (SVM) for the identification and mapping of urban road networks from high-resolution remotely sensed UAV imagery in parts of Benin City, Nigeria. The SVM algorithm, trained using a Radial Basis Function (RBF) kernel, was developed to classify road features from UAV images. Results demonstrated that SVM provided high classification accuracy and effectively distinguished road features from surrounding land uses, confirming its potential for scalable urban mapping solutions. The results were validated using reference data thereby yielding an overall classification accuracy of 89.7% with particularly robust performance in delineating major thoroughfares, clearly visible in the eastern portion of the study area where the primary arterial road exhibits clean, continuous vectorization and a Kappa coefficient of 0.85. The clear delineation of the primary road network (width > 8m) achieves 93.4% accuracy (σ = 2.1), while the more fragmented appearance of secondary roads in the central portion of the image corresponds to the lower accuracy rate of 85.7% (σ = 3.2). The mapped road network demonstrated strong correspondence with existing road infrastructure, including both major and minor roads, highlighting the capability of the SVM approach in delineating complex urban features. This study demonstrates the applicability of machine learning techniques in urban road mapping and provides a cost-effective, scalable solution for updating spatial road databases, particularly in developing urban environments with limited geospatial resources. The findings support informed urban planning and infrastructure development initiatives in Benin City and similar urban centers across sub-Saharan Africa.
Ovu et al. (Thu,) studied this question.