基于自适应权值FNEA算法的高分辨率遥感图像分割

High- resolution remote sensing image segmentation based on weight adaptive fractal net evolution approach

在高分辨率遥感图像分割方法中,分形网络演化算法(fractal net evolution approach,FNEA)是一种经典的影像对象构造方法。但在计算影像对象之间的异质性时,使用根据经验选择的固定权值会导致该算法不能很好地适应不同属性的影像对象分割。针对这一问题,提出了一种改进的FNEA方法,根据不同影像对象的空间和光谱特征,自适应地计算空间判据权值和紧凑度判据权值,并将不同光谱分量对光谱判据的贡献引入到影像对象之间异质性的计算中。计算机仿真实验结果表明,该文提出的算法对不同属性的影像对象具有很好的适应性,与同类算法相比,图像分割结果得到了较好的改善。

In the methods for high -resolution remote sense image segmentation, the fractal net evolution approach （FNEA） is relatively mature among the object -oriented image segmentation algorithms. In calculating the heterogeneity of each pair of neighboring objects, the spatial criterion and weight of compactness are user - defined according to experience. In this paper, an improved method of adaptive FNEA algorithm was proposed by adaptively calculating the weights of spatial criterion and compactness according to the different properties of various kinds of objects. Moreover, the contributions of different spectroscopic components were introduced into calculating of the heterogeneity. Computer simulation demonstrates that the proposed algorithm has better adaptability to the image objects with different attributes. A comparison with some similar algorithms shows that the method proposed in this paper performs better for image segmentation.