Research Of SAR Image Segmentation Based On Fuzzy Entropy And ABC Algorithm

With the rapid development of Synthetic Aperture Radar (Synthetic Aperture Radar, SAR) technology, the SAR image plays a very important role in satellite remote sensing, military reconnaissance, marine monitoring and surveillance of agroforestry. SAR image segmentation is distilling the target parts of an image in order to study the image more accurately. Because of imaging characteristics of SAR image, there will be coherent speckle noise in SAR image, so it exists a big difference between SAR image and ordinary optical image, thus image segmentation is also different.In this thesis, a variety of image segmentation method at home and abroad are reviewed, threshold segmentation is most widely applied in image segmentation, and some intelligent optimization algorithms are used for image segmentation which have a very good optimization effect on image segmentation. Aiming at the existing problem of SAR image segmentation, it was found that the combination algorithm has many advantages and has a great prospect for image segmentation. As a result, this thesis will focus on researching SAR image segmentation based on threshold segmentation algorithm of maximum fuzzy entropy and improved ABC, the major work and research results are as follows:First of all, because one dimensional fuzzy entropy algorithm only considers the pixel gray features ignoring the space gray level characteristics of the neighborhood, a fuzzy entropy algorithm based on two-dimensional histogram (the two-dimensional fuzzy entropy) is introduced, the algorithm gains a two-dimensional histogram of gray- neighborhood average gray, and divides two-dimensional histogram into the object region and background region, and calculates fuzzy entropy and maximum threshold respectively to segment the image by maximum threshold. The algorithm proposed provides the basis for subsequent work.Then aiming at the phenomenon which the two-dimensional fuzzy entropy has a weak ability to resist noise this thesis improves the two-dimensional histogram to get the gray - gray gradient histogram which makes the two-dimensional fuzzy entropy resist noise stronger.At last, this thesis proposes ABC (Artificial Bee Colony) algorithm in order to improve the shortcoming that the two-dimensional fuzzy entropy algorithm is slower, it sees the two-dimensional fuzzy entropy as the fitness function of ABC algorithm, through finding the optimal threshold value of the largest fitness, the image can be segmented. In order to avoid ABC algorithm appearing the phenomenon of premature convergence, this thesis will improve ABC algorithm to put forward FEIABC algorithm. FEIABC algorithm improves the way of nectar source search and transition probability calculation so as to make the speed of algorithm faster. At the beginning of the image segmentation, the image is preprocessed by gray-scale morphological, and then uses the improved algorithm to segment the image to improve the noise resistance.