Target Recognition In Synthesis Aperture Radar(SAR)Image

Since the first synthesis aperture radar (SAR) satellite Seasat was successfully implemented to observe the earth in 1978, SAR technology continued to develop rapidly. At present, US, Japan and European countries have been investing massive human and material resources in this field of research.SAR's basic principle is: when the satellite is in its orbit, it emits electromagnetic wave pulse signal to the identical object in different position on fixed time and simultaneously receives the echo signal, in some kind of significance, this extends the radar antenna length, thus greatly enhances the resolution. Moreover, because SAR also has merit of all-day, all-weather, and not effected by atmospheric dissemination and weather condition, strong penetrating power and so on, therefore it is extremely widespread in civil and the military aspect application. Its mainly used in air survey, aviation remote sensing, satellite sea observation, astronautics reconnaissance, picture match guidance and so on. It can discover hidden and camouflaged goal such as camouflaged underground missile launching silo, ground object in the fog covered area and so on. In the missile picture match guidance, the use of SAR can make the missile hit the hidden and camouflaged goal. SAR has also bean used in the deep space survey, for example, the survey of the Moon and Venus's geological structure. The research of synthesis aperture imagery radar technology in our country was started in middle 70's. After several generation of people's endeavors, SAR technology has obtained considerable progress.Based on the previous research of the world-wide SAR image segmentation and target recognition technologies, taking into account SAR images' complicated background and noisy feedback, this paper presented a solution for. First, we apply the logarithm transformation and Gauss filter in the image space field. Second, using partial characteristic of the mature wavelet transformation theory, calculate four stature belts to the picture transforms. We use the self-adapted ability partial window K-means algorithm to process the target images, and use Level Set algorithm to segment the results. Finally, the boundaries are represented by the chain code. The experimental results show that this method is efficient for segmenting SAR images.