An Internal Waves Detection Method Based on PCANet for Images Captured from UAV

As internal wave is a universal geophysical phenomenon in stratified fluids, study of internal wave features in the coastal ocean is one of the most important tasks in physical oceanography. Traditionally, various internal wave detection methods, such as acoustic, optical, electrical based techniques and SAR based technique have been proposed. However, those methods need expensive measuring devices and often face the difficulties of the installation when deployed in the ocean. With the development of machine learning recently, internal wave detection based on computer vision and machine learning becomes a hot topic. In this paper, a framework for internal waves detection based on PCANet which is a feature learning deep network is proposed. First, we collect simulated internal wave images and non-internal wave images, then we give a label to each image to indicate whether it includes internal waves or not. Finally, we train a discrimination model with PCANet and predict new images at the test stage. Experiment results demonstrated the feasibility of the technique for internal wave detection.