IRSAMap:Towards Large-Scale, High-Resolution Land Cover Map Vectorization

With the continuous development of remote sensing technology and its application requirements, higher requirements have been put forward for the accuracy and expression of land cover extraction and mapping. The traditional pixel based grid classification results have problems such as jagged edges and inconsistent border connections across image frames, making it difficult to directly express topological relationships such as road intersections and building adjacency. In subsequent applications, additional manual vectorization processing is often required to support GIS analysis, resulting in high process costs and low efficiency. Object level vector modeling expresses land features in vector forms such as "surfaces, lines, and points", which is naturally suitable for GIS and surveying applications. It can be directly used for area and length measurement, network connectivity analysis, and spatiotemporal change updates. To achieve this technological leap, there is an urgent need for high-quality datasets with large-scale, multi scenario, and standardized annotation as support. IRSAMap was born to provide a standardized data foundation for the research and application of vectorization methods.