Convolutional neural networks (CNNs) have demonstrated their ability object detection of very high resolution remote sensing images. However, CNNs have obvious limitations for modeling geometric variations in remote sensing targets. In this paper, we introduced a CNN structure, namely deformable ConvNet, to address geometric modeling in object recognition. By adding offsets to the convolution layers, feature mapping of CNN can be applied to unfixed locations, enhancing CNNs’ visual appearance understanding. In our work, a deformable region-based fully convolutional networks (R-FCN) was constructed by substituting the regular convolution layer with a deformable convolution layer. To efficiently use this deformable convolutional neural network (ConvNet), a training mechanism is developed in our work. We first set the pre-trained R-FCN natural image model as the default network parameters in deformable R-FCN. Then, this deformable ConvNet was fine-tuned on very high resolution (VHR) remote sensing images. To remedy the increase in lines like false region proposals, we developed aspect ratio constrained non maximum suppression (arcNMS). The precision of deformable ConvNet for detecting objects was then improved. An end-to-end approach was then developed by combining deformable R-FCN, a smart fine-tuning strategy and aspect ratio constrained NMS. The developed method was better than a state-of-the-art benchmark in object detection without data augmentation.

Water-quality monitoring of inland lakes is essential for freshwater-resource protection. In situ water-quality measurements and ratings are accurate but high costs limit their usage. Water-quality monitoring using remote sensing has shown to be cost-effective. However, the nonoptically active parameters that mainly determine water-quality levels in China are difficult to estimate because of their weak optical characteristics and lack of explicit correlation between remote-sensing images and parameters. To address the problems, a convolutional neural network (CNN) with hierarchical structure was designed to represent the relationship between Landsat8 images and in situ water-quality levels. A transfer-learning strategy in the CNN model was introduced to deal with the lack of in situ measurement data. After the CNN model was trained by spatially and temporally matched Landsat8 images and in situ water-quality data that were collected from official websites, the surface quality of the whole water body could be classified. We tested the CNN model at the Erhai and Chaohu lakes in China, respectively. The experiment results demonstrate that the CNN model outperformed widely used machine-learning methods. The trained model at Erhai Lake can be used for the water-quality classification of Chaohu Lake. The introduced CNN model and the water-quality classification method could cover the whole lake with low costs. The proposed method has potential in inland-lake monitoring.

Change detection (CD) is an essential earth observation technique. It captures the dynamic information of land objects. With the rise of deep learning, convolutional neural networks (CNN) have shown great potential in CD. However, current CNN models introduce backbone architectures that lose detailed information during learning. Moreover, current CNN models are heavy in parameters, which prevents their deployment on edge devices such as UAVs. In this work, we tackle this issue by proposing RDP-Net: a region detail preserving network for CD. We propose an efficient training strategy that constructs the training tasks during the warmup period of CNN training and lets the CNN learn from easy to hard. The training strategy enables CNN to learn more powerful features with fewer FLOPs and achieve better performance. Next, we propose an effective edge loss that increases the penalty for errors on details and improves the network’s attention to details such as boundary regions and small areas. Furthermore, we provide a CNN model with a brand new backbone that achieves the state-of-the-art empirical performance in CD with only 1.70M parameters. We hope our RDP-Net would benefit the practical CD applications on compact devices and could inspire more people to bring change detection to a new level with the efficient training strategy. The code and models are publicly available at https://github.com/Chnja/RDPNet.

The launch of the Chinese Gaofen-3 (GF-3) satellite will provide enough synthetic aperture radar (SAR) images with different imaging modes for land cover classification and other potential usages in the next few years. This paper aims to propose an efficient and practical classification framework for a GF-3 polarimetric SAR (PolSAR) image. The proposed classification framework consists of four simple parts including polarimetric feature extraction and stacking, the initial classification via XGBoost, superpixels generation by statistical region merging (SRM) based on Pauli RGB image, and a post-processing step to determine the label of a superpixel by modified majority voting. Fast initial classification via XGBoost and the incorporation of spatial information via a post-processing step through superpixel-based modified majority voting would potentially make the method efficient in practical use. Preliminary experimental results on real GF-3 PolSAR images and the AIRSAR Flevoland data set validate the efficacy and efficiency of the proposed classification framework. The results demonstrate that the quality of GF-3 PolSAR data is adequate enough for classification purpose. The results also show that the incorporation of spatial information is important for overall performance improvement.

Convolutional neural networks (CNN) have achieved great success in the optical image processing field. Because of the excellent performance of CNN, more and more methods based on CNN are applied to polarimetric synthetic aperture radar (PolSAR) image classification. Most CNN-based PolSAR image classification methods can only classify one pixel each time. Because all the pixels of a PolSAR image are classified independently, the inherent interrelation of different land covers is ignored. We use a fixed-feature-size CNN (FFS-CNN) to classify all pixels in a patch simultaneously. The proposed method has several advantages. First, FFS-CNN can classify all the pixels in a small patch simultaneously. When classifying a whole PolSAR image, it is faster than common CNNs. Second, FFS-CNN is trained to learn the interrelation of different land covers in a patch, so it can use the interrelation of land covers to improve the classification results. The experiments of FFS-CNN are evaluated on a Chinese Gaofen-3 PolSAR image and other two real PolSAR images. Experiment results show that FFS-CNN is comparable with the state-of-the-art PolSAR image classification methods.

High-precision 3D laser scanning pavement data contains rich pavement scene information and certain components associations. Moreover, for pavement maintenance and management, there is an urgent need to develop automatic methods that can extract comprehensive information about different pavement indicators simultaneously. By analyzing the frequency and sparse characteristics of pavement distresses and performance indicators—including the cracks, road markings, rutting, potholes, textures—this paper proposes 3D pavement components decomposition model (3D-PCDM) which decomposes the 3D pavement profiles into sparse components x, low-frequency components f, and vibration components t. Designed high-pass filter was first employed to separate f, then, x and t are separated by total variation de-noising which based on sparse characteristics. Decomposed x can be used to characterize the location and depth information of sparse and sparse derived signals such as cracks, road marks, grooves, and potholes in profiles. Decomposed f can be used to determine the slow deformation of pavement. While decomposed t reflects the fluctuation of the pavement material particles. Experiments were conducted using actual pavement 3D data, the decomposed components can obtain by 3D-PCDM. The effectiveness and accuracy of the x are verified by actual cracks and road markings, the accuracy of extracted sparse components is over 92.75%.