Turbulence mitigation methods for sea scenarios

Visual and infrared imagery is degraded by turbulence caused by atmospheric conditions. Because the degradation gets worse for longer distances, turbulence especially hampers long range observation. At sea this turbulence affects classification and identification of ships and other objects. State of the art software based processing algorithms assuming a static background assumption will fail in such conditions because of the non-static sea background. Therefore, we propose an adapted processing chain aiming to provide optimal turbulence correction for ships seen in the camera view. First we propose to use standard object detection and tracking methods for an indication of the location of the ship. Subsequently, image registration is performed within the ship's region of interest, covering only the ship of interest. After this region of interest registration, standard turbulence mitigation software can be applied to the region of interest. For ships with other movement than translation only we propose a two-step motion estimation using local optical flow. In this paper we show results of this processing chain for sea scenarios using our TNO turbulence mitigation method. Ship data is processed using the algorithm proposed above and the results are analyzed by both human observation and by image analysis. The improvement of the imagery is qualitatively shown by examining details which cannot be seen without processing and can be seen with processing. Quantitatively, the improvement is related to the energy per spatial frequency in the original and processed images and the signal to noise improvement. This provides a model for the improvement of the results, and is related to the improvement of the classification and identification range. The results show that with this novel approach the classification and identification range of ships is improved. © COPYRIGHT SPIE. Downloading of the abstract is permitted for personal use only.
The Society of Photo-Optical Instrumentation Engineers (SPIE)