SHIP DETECTION FROM SATELLITE IMAGES USING DEEP NEURAL NETWORK

SHIP DETECTION:

SHIP detection in space borne remote sensing images is of vital importance for maritime security and other applications, e.g., traffic surveillance, protection against illegal fisheries, oil discharge control, and sea pollution monitoring.

Vessel monitoring from satellite images provides a wide visual field and covers large sea area and thus achieves a continuous monitoring of vessels’ locations and movements.

 

It is also known that optical spaceborne images have higher resolution and more visualized contents than other remote sensing images, which is more suitable for ship detection or recognition in the aforementioned applications.

However, optical spaceborne images usually suffer from two main issues: 1) weather conditions like clouds, mists, and ocean waves result in more pseudotargets for ship detection, and 2) optical spaceborne images with higher resolution naturally lead to larger data quantity than other remote sensing images, and thus, optical spaceborne images are more difficult to be tackled for real-time applications.

 The ELM(Extreme Learning Machine), is adopted for feature fusion and classification, and thus, faster and better ship detection is achieved. Using these novel techniques, the proposed framework is more suitable for ship detection than the aforementioned approaches with the following advantages.

1) Faster detection. Compressed domain achieves much faster detection than pixel domain.

2) More reliable results. High-level feature representations are extracted by hierarchical deep architecture to ensure more accurate classification.

3) Better utilization of information.  Two DNNs are trained with multisubbands coefficients to make full use of the wavelet information.

  FOR 2015 IEEE PAPER ON SHIP DETECTION USING IMAGE PROCESSING

 Compressed-Domain Ship Detection on Spaceborne Optical Image Using Deep Neural Network and Extreme Learning Machine