Seagrass Detection in Coastal Water Through Deep Capsule Networks
Seagrass is an important factor to balance marine ecological systems, and there is a great interest in monitoring its distribution in different parts of the world. This paper presents a deep capsule network for classification of seagrass in high-resolution multispectral satellite images. We tested our method on three satellite images of the coastal areas in Florida and obtained better performances than those achieved by the traditional deep convolutional neural network (CNN) model. We also propose a few-shot deep learning strategy to transfer knowledge learned by the capsule network from one location to another for seagrass detection, in which the capsule network’s reconstruction capability is utilized to generate new artificial data for fine-tuning the model at new locations. Our experimental results show that the proposed model achieves superb performances in cross-validation on three satellite images collected in Florida as compared to support vector machine (SVM) and CNN.
Seagrass detection Convolutional neural network Capsule network Deep learning Remote sensing Transfer learning
Kazi Aminul Islam Daniel Pérez Victoria Hill Blake Schaeffer Richard Zimmerman Jiang Li
Department of Electrical and Computer Engineering,Old Dominion University,Norfolk,VA 23529,USA Department of Modeling,Simulation and Visualization Engineering,Old Dominion University,Norfolk,VA 2 Department of Ocean,Earth and Atmospheric Sciences,Old Dominion University,Norfolk,VA 23529,USA Office of Research and Development,U.S. Environmental Protection Agency,Durham,NC,USA
国际会议
广州
英文
320-331
2018-11-23（万方平台首次上网日期，不代表论文的发表时间）