针对多尺度分割技术中的最优尺度选择问题，提出一种基于分形网络演化方法和改进模糊聚类遥感影像分割的方法。该方法利用分形网络演化方法对原始影像进行小尺度分割，并利用粒子群算法的全局搜索能力，从预分割的小尺度对象中确定最优初始聚类中心，在对小尺度对象聚类合并时，建立具有对象空间信息和对象间相关信息的目标函数，最终得到适应不同尺度地物的分割结果，降低了多尺度分割方法对尺度参数的过度依赖。实验结果表明，该方法可获得高质量的遥感影像分割结果。

Aiming at the optimal scale in multi-scale segmentation technology selection problem, a method is put forward based on fractal net evolution approach and improved fuzzy c-means of remote sensing image segmentation. In this method, the original image is segmented by small scale using fractal net evolution approach. The global search capability of the particle swarm method is used to determine the optimal initial clustering center from the pre-segmented small scale objects. When small scale objects are merged, the objective function of the object spatial information and the correlation information between objects is established. Ultimately, the segmentation results which can adapt to different scale features are obtained, and the excessive dependence on the scale parameters is reduced. Experimental results show that this method can obtain high quality segmentation results of remote sensing images.