Cadastral Parcel Boundary Extraction from UAV Images

Abstract

Information extraction from unmanned aerial vehicle (UAV) images has progressed to a great extent. The focus for geospatial data sensing, representation, processing, visualization and information delineation has gradually shifted to UAV, because of its fast delivery with very high spatial resolution data. The users from various fields have started using the UAV data for potential field uses. The land records boundary information delineation is one such case where the UAV images are used for quick and accurate boundary extraction. The extracted boundary is further used as input for the geographical information system (GIS). The fast image acquisition also needs faster and automated information extraction for the technology to meet the societal need. After the image acquisition from the UAV image, the image is segmented using the mean-shift techniques. The best fit segmentation with minimum over and under-segmentation is taken up for automated boundary extraction. The segment boundary coordinates are extracted using the MATLAB image processing tools and represented in vector format using the Geospatial Data Abstraction Library tool. The extracted boundary is compared with the manually digitized boundary with the help of buffers around the reference data and the extracted data. The buffer of 2.0 m is taken up for the completeness, correctness and quality parameter assessment. The completeness, correctness and quality of the extraction are found to be 73.02%, 73.28% and 59.66%, respectively.