Ship Detection and Classification on Optical Remote Sensing Images Using Deep Learning

Ying Liu,Hong-Yuan Cui,Zheng Kuang,Guo-Qing Li,L Long,H Yang,Y Dai,X Li,Y Li

doi:10.1051/itmconf/20171205012

Abstract

Ship detection and classification is critical for national maritime security and national defense. Although some SAR (Synthetic Aperture Radar) image-based ship detection approaches have been proposed and used, they are not able to satisfy the requirement of real-world applications as the number of SAR sensors is limited, the resolution is low, and the revisit cycle is long. As massive optical remote sensing images of high resolution are available, ship detection and classification on theses images is becoming a promising technique, and has attracted great attention on applications including maritime security and traffic control. Some digital image processing methods have been proposed to detect ships in optical remote sensing images, but most of them face difficulty in terms of accuracy, performance and complexity. Recently, an autoencoder-based deep neural network with extreme learning machine was proposed, but it cannot meet the requirement of real-world applications as it only works with simple and small-scaled data sets. Therefore, in this paper, we propose a novel ship detection and classification approach which utilizes deep convolutional neural network (CNN) as the ship classifier. The performance of our proposed ship detection and classification approach was evaluated on a set of images downloaded from Google Earth at the resolution 0.5m. 99% detection accuracy and 95% classification accuracy were achieved. In model training, 75× speedup is achieved on 1 Nvidia Titanx GPU.

Highlights

Ship detection and classification in remote sensing images is of vital importance for maritime security and other applications, e.g., traffic surveillance, protection against illegal fisheries and sea pollution monitoring
The specific contributions of this paper are as follows: 1) CohenDaubechies-Feauveau 9/7 (CDF 9/7) wavelet coefficients were extracted from the raw images and ship candidates were extracted from the LL subband by conducting image enhancement, target-background segmentation and ship locating based on shape criteria; 2) a convolutional neural network (CNN) model was implemented for ship detection and 99% accuracy was achieved; 3) for ship classification, using the proposed model, 95% accuracy was achieved; 4) up to 75h speedup was achieved on a server with a GTX Titanx GPU
99% detection accuracy and 92% classification accuracy were achieved in dataset (1), which is comparable with some state-of-the-art algorithms, such as support vector machine (SVM)

Summary

Introduction

Ship detection and classification in remote sensing images is of vital importance for maritime security and other applications, e.g., traffic surveillance, protection against illegal fisheries and sea pollution monitoring. With the increasing volume of satellite image data, automatic ship detection and classification from remote sensing images is a crucial application for both military and civilian fields. Many valuable studies have been carried out in this field, but these typical algorithms are usually effective only for common image analysis, not for the task of ship detection and classification in remote sensing images which often contains vast data and many background noises. Deep learning, or deep neural network has shown great promise in many practical applications. Since the early work of ship detection and classification, it has been known that the variability and the richness of image data make it almost impossible to build an accurate detection and classification system entirely by hand

Methods

Results

Conclusion