Classification of bulk cargo types stored openly at ports using CNN
In this paper, we present a novel concept of tracking cargoes at open ports using remote sensing images and convolution neural network (CNN) to classify various dry bulk commodities. The dataset used is prepared using Sentinel-2 atmospherically corrected (Sentinel-2 L2A) images covering 12 spectral bands. There are total 4995 labeled and geo-referenced images for four different cargoes-bauxite, coal, limestone and logs. We provide benchmarks for this dataset using a CNN. The overall classification accuracy achieved was more than 90% for all cargo types. The dataset finds its applications in detecting and identifying cargoes on open ports.