Abstract

Land administration in Cameroon still relies on paper-based land titles, causing inefficiencies, errors, and delays in verifying property rights. Extracting structured information from scanned historical documents is difficult due to OCR inaccuracies in numerical values (such as coordinates) and textual data, which can distort polygon geometries, misplace parcels, or alter recorded surface areas. This study introduces a hybrid framework that integrates computer vision, dual OCR pipelines, fuzzy logic reasoning, Tobler’s law–based spatial correction, and geospatial reprojection to automate the extraction and georeferencing of alphanumeric and coordinate data from scanned land titles. The system follows a five-layer architecture comprising input acquisition, image preprocessing, intelligent extraction, decision logic, and geospatial optimization. Tests on twenty scanned titles show 95–97% text extraction accuracy and an 8–14% improvement in coordinate consistency with fuzzy correction. Compared to OCR-only methods, the hybrid pipeline demonstrates greater robustness and supports large-scale digitization and modernization of land governance in Africa.

This work is licensed under a Creative Commons Attribution 4.0 License.