A New Method for Detecting Ships in Low Size and Low Contrast Marine Images: Using Deep Stacked Extreme Learning Machines

Document Type : Research Article

Authors

1 IROST

2 Department of Electrical, Biomedical and Mechatronics Engineering, Qazvin Branch, Islamic Azad University, Qazvin, Iran.

3 Department of Electrical, Biomedical and Mechatronics Engineering, Qazvin Branch, Islamic Azad University, Qazvin, Iran

10.22060/eej.2020.18037.5341

Abstract

Detecting ships in marine images is an essential problem in maritime surveillance systems. Although several types of deep neural networks have almost ubiquitously used for this purpose, but the performance of such networks greatly drops when they are exposed to low size and low contrast images which have been captured by passive monitoring systems. On the other hand factors such as sea waves, cloud and wave sequence caused by the floats which all may be considered as clutter in sea images, also exacerbate this problem. In this paper a new method is proposed to improve the performance of deep neural networks in detecting ships in low size and low contrast marine images which has been based on the concept of deep stacked extreme learning machines. In proposed method the resultant features have more generality in modeling of marine images based on superposition of dedicated mapping functions of extreme learning machines. Furthermore they have the minimal overlap thanks to performing decorrelation process on features which are propagated between network layers. The performance of the proposed method is evaluated on several marine images which have been captured in sunny, rainy and hazy conditions. The obtained results are compared with some other state-of-the-art detection methods by using standard parameters. Increased F-measure of the proposed method (i.e. 3.5 percent compared to its closest alternative) in parallel with its better accuracy, recall and precision shows its effectiveness in detecting ships in low size and low contrast marine images.

Keywords

Main Subjects