Registration Of Airborne Remote Sensing Oil Spill Image Based On Harris And GA

With unique flexibility and rapid reaction capability, airborne remote sensing is extremely important in marine oil spill emergency response. However, sometimes the oil spill pollution belt on the sea surface is too long to cover in a picture. For example, the oil spill pollution accident happened in the Estuary of the Pearl River in December 2004, the oil spill pollution belt was 9 miles long and 200 meters broad. It was very hard for us to take the panoramic picture of the oil pollution in a photograph, so we needed a number of remote sensing images. Then we adopted image mosaic technology to obtain the panoramic picture of the oil spill pollution.According to the characteristic of airborne remote sensing images, in this paper, we selected some common used feature point operators (Moravec operator, Susan operator and Harris operator) to detect feature points of image. Then through testing, comparing and analyzing them, we found Harris operator was accurate, anti-noisy and its detecting results had no obvious change after rotating and shifting the images. So we adopted Harris operator to detect feature points of airborne remote sensing oil spill images.Genetic algorithm was adopted to matching feature points. Through designing its code method and fitness function, we matched feature points of two images successfully and the best accuracy can reach to 83.4%. By matched feature points of image, projective transformation model was acquired and ultimately the remote sensing oil spill image registration was completed.