Sea-land Segmentation of Infrared Remote Sensing Image Based on Complex Background

With the development of infrared payload technology, the resolution is gradually improved, which makes it possible for infrared payload to detect marine ship targets all-weather. However, due to the influence of atmosphere, ocean thermal radiation, detector noise and other factors, ship targets appear as small targets with low SNR in infrared image. How to recognize the infrared ship in the complex background is a difficult problem. For the infrared ship target, the land and the sea need to be segmented first. For the situation in which the sea-land boundary is not obvious, for example, when the coastal area is covered by clouds or the water depth is shallow, it is difficult to use the traditional methods such as region growing method to realize the sea-land segmentation. For the sea-land segmentation of complex background mentioned above, this paper proposes a real-time positioning and matching method based on blocks combined with power spectrum statistics, regional growth and GIS database template, which realizes the sea-land segmentation of satellite infrared image with low SNR and low contrast.