Workshops

"Phenological based crop mapping using time series analysis of remote sensing images"

The potential of satellite imagery to map agricultural land cover is explored. The crop type change analysis shows that a remote sensing based crop mapping method is the only means to map the frequent change of major crop types. The traditional classification approach is first utilized to map crop types to test the classification capacity of existing algorithms. High accuracy is achieved with sufficient ground truth data for training, and crop maps of moderate quality can be timely produced. However, the large set of ground truth data required by this method results in high costs in data collection. It is difficult to reduce the cost because a trained classification algorithm is not transferable between different years or different regions. In this workshop, we will used a time series of Landsat images, preprocess and prepare it for phenology based classification approach. In this approach we extract the phenological metrics from annual vegetation index profiles and identify crop types based on these metrics using random forest classification and decision trees classification algorithms. Spectral signatures are associated with phenological stages instead of imaging dates are also employed to identify various crop types. The focus of this workshop will be on the advantages of using the phenology-based classification approach for producing the precise annual crop mapping and update of these maps for next year's when the size of the training set is limited by ground truth availability.

Presenter Information