Detection of Marine Chemical Pollution Based on Image Processing and Machine Learning

The problem of oil spills is lethal to the ocean ecosystem. To solve the problem, one of the most important key steps is to detect the ocean surface and judge whether there are or not oil spills. Remote sensing provides the advantage of controlling and observing events remotely, and it can cover the areas that people cannot access, so we use it to build a database. Next, we choose to use Matlab for the pre-image processing and then use the neural network by Python to realize and there are five pre-processing methods: expanding the dynamic histogram range in the ‘Y’ channel (method 1), expanding the dynamic histogram range in three channels (method 2), contrast enhancement (method 3), expanding the dynamic histogram range and then contrast enhancement (method 4), and contrast enhancement, and then expanding the dynamic histogram range (method 5). Finally, we use a neural network to test accuracy, in comparison, method 1 is the best and we improve the accuracy from 72% to 82%.