基于均值漂移滤波及谱分类的海面舰船红外目标分割

提出了一种有效的海面舰船红外目标分割方法.利用均值漂移方法的不连续保持性滤波特性,滤除海面的强杂波干扰,同时又不损失舰船目标的信息.根据滤波得到的区域构建区域邻接图,采用基于最大最小SST图划分算法对区域邻接图的节点进行划分.划分结果最终将图像分为天空背景、海面背景以及舰船目标3个部分.由于采用区域节点来表征图像,较之采用原始图像象素节点表示,其节点个数大大减少,从而使算法的计算效率得到很大提高.实验结果也表明提出两步算法具有优越的性能,能够在海面强杂波干扰的情况下有效提取舰船红外目标.

SHIP INFRARED OBJECT SEGMENTATION BASED ON MEAN SHIFT FILTERING AND GRAPH SPECTRAL CLUSTERING

A novel thresholding algorithm was presented to achieve an improved ship infrared object segmentation performance.The proposed algorithm uses discontinuity preserving smoothing algorithm based on mean shift procedure to filter the powerful noise without the loss of the ship object information.The regions produced by mean shift filtering can be represented by a planar weighted region adjacency graphs that incorporates topological information of the image structure and region connectivity.Under the graph representation,region merging algorithm based on SST-minimax was applied to partition the regions into different class,such as sky background,sea background and ship object.Due to the less nodes produced by the regions of filtered image than the original image,the region merging based on SST-minimax algorithms has much less computational complexity.A large number of examples are presented to show the superior performance of the proposed ship infrared object segmentation algorithm.