Parcel Extraction

Land parcels are defined pieces of land that mark property boundaries. They play a significant role in cadastral mapping, land administration, urban planning, property management, and in developing geospatial basemaps. Conventional parcel mapping has always relied on field surveys and surveying practices that ensure the determination of property boundaries through accurate positioning. Even though these practices can yield accurate results, they are both time-consuming and costly, especially if the areas are extensive.

The emergence of highly-resolved aerial and satellite imagery has made the development of effective image-based techniques a reality in parcel delineation. With advanced computer vision and machine learning technologies, such imagery can be used to identify and delineate parcels based on various landscape features such as roads, buildings, fences, walls, vegetation cover, driveways, among others. The automation of parcel delineation through these techniques would not only reduce the workload involved but would also greatly speed up the process of developing parcel data sets.

Nonetheless, there needs to be a distinction between visually inferred parcel boundaries and legal property boundaries. While parcels in residential and rural settings have visible features that give clues to where parcel boundaries may be located, there are other parcels that are legally recognized but have no visible features that indicate property boundaries.

Therefore, automated parcel mapping models are developed for creating probable or rough estimates of parcel boundaries and not the legally recognized parcel boundaries. In this regard, the output is to be regarded as a rough estimation of the parcel boundary data, which can be helpful in performing GIS mapping and geospatial analysis.

Automated parcel boundaries can be incorporated into the GIS process to create basemaps, property visualization, land use maps, urban planning datasets, and various other forms of GIS applications. Once the parcel boundaries are created, GIS experts can validate the boundaries using survey information, cadastre data, observations on the ground, or any other form of authoritative sources.

Through the incorporation of high-resolution images, deep learning, computer vision, and GIS technology, automated parcel delineation becomes a highly efficient way of creating preliminary parcel data sets while ensuring the