PANet: parallel attention network for remote sensing image semantic segmentation

Abstract:
Remote sensing image processing technology plays an important role in crop planning, vegetation detection and agricultural land detection. However, unlike ordinary natural scene, the small objects in remote sensing images are difficult to detect because of their low resolution and small size, which leads to inaccurate boundary segmentation of adjacent areas. To address this dilemma, this paper introduces a parallel attention network (PANet) with an encoder-decoder structure. The multi-branch encoder uses a FPN-style structure to extract the features from the image step by step from bottom to top. In decoder, the proposed parallel attention module learns the correlation between the features of different scales, which are fused for the final prediction. In order to verify the effectiveness of the PANet, the training model is tested on 2020 CCF BDCI RS Dataset, and the mean intersection over union (mIoU) achieves 63.69%, and the overall accuracy achieves 89.94%.