基于近似域划分的可变离散精度粗逻辑网络及其遥感图像分类应用

Variable Discretization Precision Rough Logic Neural Network Based on Approximation Area Partition and Its Application to Remote Sensing Image Classification

为解决粗逻辑神经网络精度与网络规模复杂性和推广泛化能力之间的矛盾,该文提出了一种具有可变离散精度的粗逻辑神经网络设计方法。该方法通过近似域划分,将论域空间划分为确定性区域和可能性区域,由于可能性区域信息粒度过大是造成误分类的重要原因,只需对可能性区域离散区间进一步细化,即可达到提高粗逻辑网络的精度,同时抑制网络规模增长过快的目的。在长白山地区的遥感图像分类实验中,常规方法在离散等级为7时有最好性能,而该文方法以较小的网络代价和训练时间获得了逼近的分类结果。

A variable discretization precision rough logic neural network is proposed to solve contradiction between network precision and the size of network as well as generalization ability. Based on the approximation area partition, the universe discussed can be partitioned into certain area and possibility area. The important reason of misclassification is the granularity of the possibility area is too coarse. In this work, only possibility area is refined and the precision of the rough logic neural network is improved while the size of network is restrained. In the experiment of the remote sensing image classification about Changbai mountain area, the performance of conventional method is best when the discretization level is 7. The most approximated result is acquired, while less network cost and training time are expended, when this method is used.