摘要
蚁群算法良好的离散性，并行性，正反馈性和鲁棒性，非常适合于图像分割。但基本蚁群算法蚂蚁的搜索是随机的，计算量大，不利于算法的收敛，为此，本文提出了设置初始聚类中心的设想，并以小窗口为对象实施算法，由此大大减小了计算量。另外基本蚁群算法中挥发系数固定，会导致算法可能过早收敛或停滞，针对这一不足，本文将其修改为随蚂蚁通过可行路径个数动态变化，使其收敛性和稳定性有了一定提高。实验证明了方法的有效性。
Ant colony algorithm, with good discretion, parallel, robustness and positive feedback, is suitable for image segmentation. But its search is random and has much computation for its convergence, Hence the ideal of setting primary clustering center is proposed in the paper, the algorithm is performed in a small window so as to reduce its computation. In addition, the evaporating efficient is constant in basic algorithm, this will lead to early convergence or stagnation. To improve it, the evaporating efficient is set to change with the number of convergence in order to keep good convergence and stability. Using this method can segment image＇s target fast and accurately. Experimental results show it＇s an effective approach for image segmentation.
关键词
蚁群算法
图像分割
收敛
聚类中心
Ant Colony Algorithm, Image Segmentation, Convergence, Clustering Center
