Authors: Yaping Cai*, University of Illinois, Shaowen Wang, University of Illinois, Kaiyu Guan, University of Illinois
Topics: Geographic Information Science and Systems, Remote Sensing, Agricultural Geography
Keywords: Crop type classification, high-performance, deep neural network, feasibility
Session Type: Paper
Start / End Time: 1:10 PM / 2:50 PM
Room: Roosevelt 5, Marriott, Exhibition Level
Presentation File: No File Uploaded
Accurate crop type classification based on remote sensing data is important for both scientific and practical purposes. The state-of-the-art research on crop-type classification has been shifted from relying on only spectral features of single static images to combining both spectral and time series information. Facing the data-intensive and computationally-intensive challenges raised by the advanced crop type classification method based on both spectral and time series information through machine learning approaches (especially for the deep neural network) for large geographic areas, we proposed a high-performance workflow to preprocess the data and speed up the model training and testing. Then we collected Landsat Surface Reflectance Data covering IL State in 2016 as model input information and collected Cropland Data Layer in 2016 as ground truth. We applied our model trained in Champaign County to the whole IL State, mainly exploring how the performance of the model change with the increasing radius of the study area, the different latitude, and the longitude. We analyzed the limitations of the model by interpreting the differences of the same crop phenology caused by climate factors (i.e. temperature and precipitation), and improved the performance of the model by modifying the deep neural network structure to eliminate those differences, which shows the potential of the model to be applied at large geographic areas.