Abstract: As the urbanization and the policy of smart city advance step by step, new challenges are put forward for the meticulous planning of cities. It is of great significance to clarify the division of urban spatial structure and strengthen the rational planning of urban functional areas. We obtain the semantic classification results of urban functional areas using the remote sensing technology and the semantic information mining method based on the GF-1 image, POI (point of interest) data, and road network data. Firstly, extraction of construction land in study area is based on object-oriented method, and the block partition is recognized by using road network data. Considering that the semantic features from POI data make fine classification of urban construction land, we estimate POI data for each category using kernel density analysis. Then the evaluation model of function area category is established based on the overlapping regions of multiple types of kernel density. Thus the function land classification of study area is completed. 360 blocks of plots are randomly selected for sample verification test. The results show that the definition of urban function areas is accurate and the accuracy of urban function zoning is as high as 87.5%.

Key words: urban functional areas, construction land extraction, POI data, kernel density estimation, semantic information mining