| Marine traffic management and security, fishing monitoring agencies, marine pollution control and various other applications require the detection of ships via optical space-borne satellite images. The main advantage of using space-borne satellite optical images is that they have more visualized segments and higher resolution. Current techniques for ship detection by space-borne synthetic aperture radar and infrared images have two main drawbacks: sensitivity to weather conditions, and a large amount of data to be transmitted. Advances in technology can mitigate the effects of these drawbacks, but they face challenges such as balancing complexity and performance. The current project presents a methodology that addresses these two challenges by using Extreme Learning Machine (ELM) techniques and Deep Neural Networks (DNN), combined with wavelet feature extraction. Simulation results from JPEG2000 images validate the proposed method, and show that it can be used for efficient and accurate ship detection. |
