Research On The Registration And Fusion Methods Of Dual-band Infrared Image Of Sea Surface

With the rapid development of China's economy,marine industries are expands rapidly.However,with the rapid development of marine transportation and sea tourism,the occurrence frequency of maritime traffic accidents has increased which caused great property loss and casualties.Maritime distress accidents occur mostly in the fog,strong wind and big waves.When the rescue helicopter is searching distress targets,infrared imaging system is all-weather and the most efficient.At present,single-band infrared imaging systems have been used on very few rescue helicopters.However,the image obtained by single-band infrared imaging system has limited available information which results in low recognition rate of distress targets.With the development of modern science,using dual-band infrared imaging system is more efficient.By means of image fusion and message complementarity,it can improve the abilities of target detection and anti-jamming.This article has mainly researched on the registration and fusion methods of dual-band infrared image of sea surface.Combining with the characteristics ofinfrared imaging technology,the dual-band infrared images were fused in order torealize the information complementation.The key step of image fusion is registration.It may result in fusion image blurring without accurate registration.However,under the background of this project,because of small distress targets,similar wave structure and few feature points,common registration methods,such as Scale Invariant Feature transform(SIFT)are not sufficient for good results.In this paper,an improvement image registration algorithm based on affine transform was proposed,according to the fixing characteristic of Mid-wave infrared system and Long-wave infrared system.On the basis of accurate registration,an image fusion method based on the wavelet transform was proposed.According to the studies,with the change in temperature,the gray feature of target relative to background luminance may different.In this circumstance,weighted average method may decrease the contrast of target.In the light of this limitation,an improvement fusion rule of low-frequency component was proposed.In this paper,local peak detection method was used to detect the suspected target area.Different fusion methods were established by assessing the brightness relationship between suspected targets area and background.This article used the evaluation indicators of fusion image quality to evaluate the fused images.Experiment results showed that the fused images have good effect which can improve the performance of targets detection and recognition.