Using spectral and spatial information coupled with mathematical morphology for oil spill detection in multispectral imagery

How to detect, monitor and track the oil spill on the sea surface are always very important tasks, since it is harmful to the ecological environment. The remotely sensed image provides an effective and efficient way to monitor the sea area. This study focuses on the detection of oil slick on sea surface using multispectral imagery technology. Since the oil spill usually only occupies a few pixels compare to the sea area in the image scene, it can be considered as an anomaly. The widely used RX algorithm, designed for anomaly detection, did not produce satisfied results. Although it successfully detects the oil slick, but the light reflections from the small waves are also detected as false alarm. In order to discriminate oil slick from the other interferers, spatial features are introduced into anomaly detection. The spatial features of oil slick are very different from that of the small wave reflection. The former often forms a connected area, but the latter ones usually separated and spread in the image. Our proposed method is to extract different spatial features form the images and combine with spectral information, and then perform anomaly detection with both spectral and spatial features to improve the oil spill detection. In the experiment, we adopt SPOT multispectral images for performance analysis.