A satellite image time series (SITS) contains a significant amount of temporal information. By analysing this type of data, the pattern of the changes in the object of concern can be explored. The natural change in the Earth’s surface is relatively slow and exhibits a pronounced pattern. Some natural events (for example, fires, floods, plant diseases, and insect pests) and human activities (for example, deforestation and urbanisation) will disturb this pattern and cause a relatively profound change on the Earth’s surface. These events are usually referred to as disturbances. However, disturbances in ecosystems are not easy to detect from SITS data, because SITS contain combined information on disturbances, phenological variations and noise in remote sensing data. In this paper, a novel framework is proposed for online disturbance detection from SITS. The framework is based on long short-term memory (LSTM) networks. First, LSTM networks are trained by historical SITS. The trained LSTM networks are then used to predict new time series data. Last, the predicted data are compared with real data, and the noticeable deviations reveal disturbances. Experimental results using 16-day compositions of the moderate resolution imaging spectroradiometer (MOD13Q1) illustrate the effectiveness and stability of the proposed approach for online disturbance detection.

Marine raft aquaculture (MFA) plays an important role in the marine economy and ecosystem. With the characteristics of covering a large area and being sparsely distributed in sea area, MFA monitoring suffers from the low efficiency of field survey and poor data of optical satellite imagery. Synthetic aperture radar (SAR) satellite imagery is currently considered to be an effective data source, while the state-of-the-art methods require manual parameter tuning under the guidance of professional experience. To preclude the limitation, this paper proposes a segmentation network combined with nonsubsampled contourlet transform (NSCT) to extract MFA areas using Sentinel-1 images. The proposed method is highlighted by several improvements based on the feature analysis of MFA. First, the NSCT was applied to enhance the contour and orientation features. Second, multiscale and asymmetric convolutions were introduced to fit the multisize and strip-like features more effectively. Third, both channel and spatial attention modules were adopted in the network architecture to overcome the problems of boundary fuzziness and area incompleteness. Experiments showed that the method can effectively extract marine raft culture areas. Although further research is needed to overcome the problem of interference caused by excessive waves, this paper provides a promising approach for periodical monitoring MFA in a large area with high efficiency and acceptable accuracy.

The recent fast development in computer vision and mobile sensor technology such as mobile LiDAR and RGB-D cameras is pushing the boundary of the technology to suit the need of real-life applications in the fields of Augmented Reality (AR), robotics, indoor GIS and self-driving. Camera localization is often a key and enabling technology among these applications. In this paper, we developed a novel camera localization workflow based on a highly accurate 3D prior map optimized by our RGB-D SLAM method in conjunction with a deep learning routine trained using consecutive video frames labeled with high precision camera pose. Furthermore, an AR registration method tightly coupled with a game engine is proposed, which incorporates the proposed localization algorithm and aligns the real Kinetic camera with a virtual camera of the game engine to facilitate AR application development in an integrated manner. The experimental results show that the localization accuracy can achieve an average error of 35 cm based on a fine-tuned prior 3D feature database at 3 cm accuracy compared against the ground-truth 3D LiDAR map. The influence of the localization accuracy on the visual effect of AR overlay is also demonstrated and the alignment of the real and virtual camera streamlines the implementation of AR fire emergency response demo in a Virtual Geographic Environment.

The three-dimensional (3D) information of buildings can describe the horizontal and vertical development of a city. The GaoFen-7 (GF-7) stereo-mapping satellite can provide multi-view and multi-spectral satellite images, which can clearly describe the fine spatial details within urban areas, while the feasibility of extracting building 3D information from GF-7 image remains understudied. This article establishes an automated method for extracting building footprints and height information from GF-7 satellite imagery. First, we propose a multi-stage attention U-Net (MSAU-Net) architecture for building footprint extraction from multi-spectral images. Then, we generate the point cloud from the multi-view image and construct normalized digital surface model (nDSM) to represent the height of off-terrain objects. Finally, the building height is extracted from the nDSM and combined with the results of building footprints to obtain building 3D information. We select Beijing as the study area to test the proposed method, and in order to verify the building extraction ability of MSAU-Net, we choose GF-7 self-annotated building dataset and a public dataset (WuHan University (WHU) Building Dataset) for model testing, while the accuracy is evaluated in detail through comparison with other models. The results are summarized as follows: (1) In terms of building footprint extraction, our method can achieve intersection-over-union indicators of 89.31% and 80.27% for the WHU Dataset and GF-7 self-annotated datasets, respectively; these values are higher than the results of other models. (2) The root mean square between the extracted building height and the reference building height is 5.41 m, and the mean absolute error is 3.39 m. In summary, our method could be useful for accurate and automatic 3D building information extraction from GF-7 satellite images, and have good application potential.

Marine floating raft aquaculture (FRA) monitoring is significant for marine ecological environment and food security assessment. Synthetic aperture radar-based monitoring is considered to be an effective means of FRA identification because of its capability for all-weather applications. Considering the poor generalization and extraction accuracy of traditional monitoring methods, a semantic segmentation model called D-ResUnet is proposed to extract FRA areas from Sentinel-1 images. The proposed model has a U-Net-like structure but combines the pre-trained ResNet34 as the encoder and adds dense residual units into the decoder. For this model, the final layer and cropping operation of the original U-Net model are removed to eliminate the model parameters. The mean and standard deviation of Precision, Recall, Intersection over Union (IoU), and F1 score are calculated under a five-fold training strategy to evaluate the model accuracy. The test experiments indicated that the proposed model performs well with the F1 of 92.6% and IoU of 86.24% in FRA extraction tasks. In particular, the ablation experiments and application experiments proved the effectiveness of the improvement strategy and the portability of the proposed D-ResUnet model, respectively. Compared with the other three state-of-the-art semantic segmentation models, the experiments demonstrate a clear accuracy advantage of the D-ResUnet model. For the FRA extraction task, this paper presents a promising approach that has refined extraction capability, high accuracy, and acceptable model complexity.

Speed detection of a moving object using an optical camera has always been an important subject to study in computer vision. This is one of the key components to address in many application areas, such as transportation systems, military and naval applications, and robotics. In this study, we implemented a speed detection system for multiple moving objects on the ground from a moving platform in the air. A detect-and-track approach is used for primary tracking of the objects. Faster R-CNN (region-based convolutional neural network) is applied to detect the objects, and a discriminative correlation filter with CSRT (channel and spatial reliability tracking) is used for tracking. Feature-based image alignment (FBIA) is done for each frame to get the proper object location. In addition, SSIM (structural similarity index measurement) is performed to check how similar the current frame is with respect to the object detection frame. This measurement is necessary because the platform is moving, and new objects may be captured in a new frame. We achieved a speed accuracy of 96.80% with our framework with respect to the real speed of the objects.