Semantic segmentation is an important field for automatic processing of remote sensing image data. Existing algorithms based on Convolution Neural Network (CNN) have made rapid progress, especially the Fully Convolution Network (FCN). However, problems still exist when directly inputting remote sensing images to FCN because the segmentation result of FCN is not fine enough, and it lacks guidance for prior knowledge. To obtain more accurate segmentation results, this paper introduces edge information as prior knowledge into FCN to revise the segmentation results. Specifically, the Edge-FCN network is proposed in this paper, which uses the edge information detected by Holistically Nested Edge Detection (HED) network to correct the FCN segmentation results. The experiment results on ESAR dataset and GID dataset demonstrate the validity of Edge-FCN.

Automatic image registration of optical-to-Synthetic aperture radar (SAR) images is difficult because of the inconsistency of radiometric and geometric properties between the optical image and the SAR image. The intensity-based methods may require many calculations and be ineffective when there are geometric distortions between these two images. The feature-based methods have high requirements on features, and there are certain challenges in feature extraction and matching. A new automatic optical-to-SAR image registration framework is proposed in this paper. First, modified holistically nested edge detection is employed to detect the main contours in both the optical and SAR images. Second, a mesh grid strategy is presented to perform a coarse-to-fine registration. The coarse registration calculates the feature matching and summarizes the preliminary results for the fine registration process. Finally, moving direct linear transformation is introduced to perform a homography warp to alleviate parallax. The experimental results show the effectiveness and accuracy of our proposed method. However, due to the different geometric and radiometric properties of SAR and optical images, to automatically register these two types of images, one must overcome many difficulties. In particular, optical images and SAR images have different geometrical characteristics. Whereas geometric distortions such as foreshortening and layover exist in SAR images, perspective and shadow exist in optical images, which cause the differences between the two types of images. In addition, optical images and SAR images have different radiometric distortion, the SAR sensor is an active remote sensing system, but the optical sensor is a passive system [4]. A large quantity of speckle noise in SAR images renders it difficult to obtain common features from a SAR image and an optical image [5]. For these reasons, the registration of optical images and SAR images has more challenges than mono-sensor image registration.The existing optical-to-SAR registration methods are mainly divided into two types: intensity-based registration methods and feature-based registration methods. Intensity-based registration methods include mutual information (MI) [6], cross-cumulative residual entropy [7] and normalized cross-correlation (NCC) [8]. Although this kind of registration method can register multi-sensor images with intensity differences, it is insensitive to the local differences between the two images and it requires many calculations [9]. Therefore, some improved intensity-based registration methods combined edges and gradient have been proposed [10][11][12]. For example, Cheah et al. [10] proposed the adaptation of MI measure which incorporates the spatial information by combining intensity and gradient information. Chen et al. [13] implemented MI through joint histogram estimation using various interpolation algorithms to complete multi-sensor and multiresolution image registration. Saidi et al. [14] proposed a refined automatic co-registration method (RA...

Semantic segmentation is an important process of scene recognition with deep learning frameworks achieving state of the art results, thus gaining much attention from the remote sensing community. In this paper, an end-to-end conditional random fields generative adversarial segmentation network is proposed. Three key factors of this algorithm are as follows. First, the network combines generative adversarial network and Bayesian framework to realize the estimation from the prior probability to the posterior probability. Second, the skip connected encoder-decoder network is combined with CRF layer to implement end-to-end network training. Finally, the adversarial loss and the cross-entropy loss guide the training of the segmentation network through back propagation. The experimental results show that our proposed method outperformed FCN in terms of mIoU for 0.0342 and 0.11 on two data sets, respectively.

In recent years, methods based on neural network have achieved excellent performance for image segmentation. However, segmentation around the edge area is still unsatisfactory when dealing with complex boundaries. This paper proposes an edge prior semantic segmentation architecture based on Bayesian framework. The entire framework is composed of three network structures, a likelihood network and an edge prior network at the front, followed by a constraint network. The likelihood network produces a rough segmentation result, which is later optimized by edge prior information, including the edge map and the edge distance. For the constraint network, the modified domain transform method is proposed, in which the diffusion direction is revised through the newly defined distance map and some added constraint conditions. Experiments about the proposed approach and several contrastive methods show that our proposed method had good performance and outperformed FCN in terms of average accuracy for 0.0209 on ESAR data set.