実験用海上監視支援システムの試作と評価

抄録

Search and rescue or surveillance of spilled oil on the sea is performed visually and by radar in Japan. However, visual surveillance may be impaired by inclement weather and sea conditions. Present radar surveillance is greatly influenced by sea clutter.

We have developed an experimental surveillance system which can aid above missions by providing appropriate information of floating objects even in high-clutter environments and in inclement weather. The system consists of a high resolution infrared (IR) camera, a color video camera, a personal computer, data recorders and a GPS receiver. In this system, we place emphasis on functions to suppress unnecessary noise in observed images, to classify floating objects, to map and display the most likely objects so that watchmen can easily identify a true target.

Several image processing techniques like filtering and image fusion are surveyed to suppress noise. Statistical estimation of RGB color distribution in the image has been employed to classify the objects. An emergency flame is tried to help detect a target by the IR camera. These techniques were incorporated and a series of procedure was built to obtain images, suppress noise, detect and classify objects and display them with necessary information. The procedure was then included in the abovementioned system.

Flight experiments were conducted to evaluate the performance of the surveillance system. A life raft was left on the sea, and the aircraft with the surveillance system flew over it and gathered the image data. From the data, the operation of our image processing procedure was tested. The results showed that our procedure worked well and succeeded in deriving a right target from the images under different weather and sea conditions. Total processing time was about 1 second, which was enough for a real time operation. Problems left to be solved were also discussed.