关键词: 语义分割, 卷积神经网络, 模糊逻辑, 遥感图像, 多尺度特征

Abstract: Abstract: Affected by imaging distance, lighting, features, environment and other factors, objects in the same class may have differences, and different classes of objects in remote sensing images often show similar visual features. This leads to uncertainty in segmentation, i.e. intra-class heterogeneity and inter-class ambiguity. To solve this problem, a Fuzzy Multiscale Convolutional Neural Network (FMCNet) was proposed for remote sensing image segmentation. By extracting receptive fields of different scales, sizes and aspect ratios, the detailed information in remote sensing objects was fully represented. Fuzzy logic was used to effectively express the relationship between pixels and their adjacent pixels, thus overcoming the uncertainty problem in remote sensing image segmentation. The experimental results show that the overall accuracy (OA) of FMCNet on the ISPR Vaihingen dataset and Potsdam dataset are 85.3% and 86.3%, respectively, and it outperforms the existing state of the art semantic segmentation methods.

Key words: semantic segmentation, convolutional neural network, fuzzy logic, remote sensing image, multi-scale feature