Maritime radar oil spill detection method and system based on neural network and gmm clustering

Abstract

The application relates to the technical field of marine environment monitoring, and discloses a marine radar oil spill detection method and system based on a neural network and GMM clustering. The method comprises the following steps: acquiring a marine radar image and dividing the image into multiple image blocks; extracting an eight-dimensional feature vector from each image block and performing dimension reduction processing to obtain a four-dimensional feature vector; clustering the four-dimensional feature vectors to generate a region of interest; performing feature extraction and anomaly detection on non-background pixels in the region of interest based on a neural network, and marking abnormal points; after removing the abnormal points, clustering the remaining non-background pixels based on a Gaussian mixture model to obtain an oil film segmentation result; fusing the oil film segmentation result and the preprocessed image, converting the fusion result from a Cartesian coordinate system into a polar coordinate system, and outputting a final detection result. The scheme can significantly improve the detection accuracy and robustness, realizes stable and reliable oil spill identification under complex sea conditions, and provides support for marine environment monitoring.

Images