Comparison of IR and Visual Stream in Maritime Zone Surveillance in Cases of Low and High Visibility Conditions

Fusion of IR and video can be useful in special cases, but is not necessary when visibility is relatively good (cloudy, light rain, sun, snowfall). When fusing images from different sources, resolution might be the biggest problem. Therefore, the input video signal must be downsampled to match the resolution of the IR sensor. This leads to problems in image registration. In this paper, examples from outdoor applications (maritime surveillance) are presented. An artificial neural network called YOLO was used to detect moving objects.