Search

The recognition and categorization of ships is very significant for maritime security and defense board. The existing SAR image-oriented ship recognition techniques are not capable of real-time applications since the count of SAR sensors is sparse and the contrast of images is minimum. To cope up with certain issues, this article designs an efficient technique for ship detection and classification by introducing hybrid optimization-based deep learning technique. Here, segmentation is done using bounding box segmentation. The ships on the ocean surface are effectively detected using deep convolutional neural network, where the network is trained utilizing devised exponential mayfly optimization algorithm (EMO). The newly designed EMO is derived by the combination of the exponential weighted moving average (EWMA) and mayfly optimization algorithm (MA). The type of ships is classified effectively utilizing deep residual neural network (DRN) and the network is trained by introducing the proposed political exponential mayfly optimization algorithm (PoEMO). [ABSTRACT FROM AUTHOR]

Super-resolution (SR) based on deep learning has been playing an important role in improving the spatial resolution of remote sensing images. Although convolutional neural networks (CNNs) dominate the research of remote sensing image SR, most of them struggle to fully utilize the multilevel features in information transmission. Constantly expanding the network architecture sometimes also leads to an increase in feature redundancy and computational complexity. Moreover, CNN-based methods are unable to generate visually appealing images. To address the aforementioned issues, we propose a multilevel feature fusion attention SR method based on GAN called MFFAGAN. Specifically, we propose a novel enhanced mixed-attention block (EMAB), which enables the network to capture key feature information in both the channel and spatial domains. Meanwhile, in order to enhance the model's ability to extract various features at multiple levels more efficiently, we propose a multilevel feature fusion attention module (MFFAM). The output of each residual block is directly fed into the feature aggregation block and eventually combined with the attention branch. Thus, the network is capable of aggregating these information-rich residual features without any loss to produce more representative features. Experimental results show that our proposed MFFAGAN outperforms most state-of-the-art methods in both quantitative and qualitative metrics.

In remote sensing image (RSI) object detection, the oriented bounding box (OBB) can accurately locate objects with arbitrary orientation and obtain orientation information. The detection based on OBB is still a challenging task. In RSI, the distribution of objects is extremely uneven, which causes aggregation to occur. Some researchers believe that the characteristic of dense distribution is a reason for the difficulty of object detection. However, there are no in-depth experimental studies on this. This paper proposes an OBB-based dense object determination method, which determines the dense objects in datasets by two conditions consisting of interclass distance, intraclass distance, minimum distance between objects, and minimum edge length of objects. The experimental results of dense and non-dense object detection concludes that the characteristics of dense distribution in RSI do not easily cause the objects to be more difficult to detect. To make full use of the object features, we propose a second-stage detection head named RoIF-Net, in which we extract region of interest (RoI) from the input image and fuse it with the RoI extracted from feature maps to add detail features, and construct a feature induction module based on self-attention mechanism to achieve position regression and category classification. This structure can be used in any two-stage network to enhance detection capabilities. Using our method on three credible and challenging datasets, DOTA, DIOR-R, and UCAS-AOD, we obtained 81.80%, 68.49%, and 90.25% mAP, respectively, reaching SOTA based on OBB detection, proving the effectiveness and advancement of our method.

With advances in social development and economic growth, remote sensing technology has been attracted greater attention in monitoring the earth data using radar and optical sensors on satellite platforms for a wide range of applications in different fields such as coastal, hazard and natural resources. Satellite images could play a greater role in improving classification accuracy with high spatial resolution and rich spectral information for land cover classification. However, existing image fusion methods achieves low accuracy due to large-scale feature space. To focus on these issues, a deep learning network structure needs to classify different classes with high spatial resolution and rich spectral information to obtain higher accuracy. In this paper, a feature-based classification approach is proposed namely Multi-Fusion based Dense Transpose Convolutional layer in Fully Convolutional Network with Feature Alignment framework (MF-DTCFCN) to label and categorizes the label region in Remote Sensing Images (RSI). Initially, a multi-fusion feature framework is designed by adding a point-wise addition structure to handle large-scale feature space for high-resolution images. Secondly, the optimized features are pre-trained to classify the labels comprised of the most discriminative features in the pre-training network. The density of output label maps are improved by introducing dense transpose convolution in the network. Then combine the output to the feature alignment with point-wise addition is employed to balance the different features and similarities to achieve additional performance for classification. Here, the Land Use/land Cover (LULC) satellite image dataset namely, Sentinel-2 were used to classify the urban areas of Hyderabad city, India. Experimental results depict that the MF-DTCFCN approach outperforms an accurate improvement in classification accuracy than existing methods. [ABSTRACT FROM AUTHOR]

Remote Sensing Images (RSI) are captured by the satellites. The quality of the RSIs primarily depends on environmental conditions and image-capturing device capability. Rapid development in technology leads to the generation of High-Resolution (HR) images from satellites. However, these images are to be processed in a scientific way for the best results. A new Image Fusion (IF) technique with the help of wavelets, Deep Convolutional Generative Adversarial Networks (DCGAN), was designed to get super-resolution images for satellite images. Residual Convolution Neural Network (ResNet) increases the fused image accuracy by minimizing the vanishing gradient problem. Peak Signal to Noise Ratio (PSNR), Structural Similarity Index Method (SSIM), Feature Similarity Index Method (FSIM), and Universal Image Quality (UIQ) are taken as the metrics for comparing the results with other models. The experimental results are better than previous methods and minimize the spatial and spectral losses during the fusion. [ABSTRACT FROM AUTHOR]

This paper presents a novel computational model to detect airports in optical remote sensing images (RSI). It works in a hierarchical architecture with a coarse layer and a fine layer. At the coarse layer, a target-oriented saliency model is built by combing the cues of contrast and line density to rapidly localize the airport candidate areas. Furthermore, at the fine layer, a learned condition random field (CRF) model is applied to each candidate area to perform the fine detection of the airport target. The CRF model is learned based on sparse features of local patches in a multi-scale structure and it also takes the contextual information of target into consideration. Therefore, its detection is more accurate and is robust to target scale variation. Comprehensive evaluations on RSI database from the Google Earth and comparisons with state-of-the-art approaches demonstrate the effectiveness of the proposed model. [ABSTRACT FROM AUTHOR]