Superimposing Electronic Navigational Chart (ENC) data on marine radar images can enrich information for navigation. However, direct image superposition is affected by the performance of various instruments such as Global Navigation Satellite Systems (GNSS) and compasses and may undermine the effectiveness of the resulting information. We propose a data fusion algorithm based on deep learning to extract robust features from radar images. By deep learning in this context we mean employing a class of machine learning algorithms, including artificial neural networks, that use multiple layers to progressively extract higher level features from raw input. We first exploit the ability of deep learning to perform target detection for the identification of marine radar targets. Then, image processing is performed on the identified targets to determine reference points for consistent data fusion of ENC and marine radar information. Finally, a more intelligent fusion algorithm is built to merge the marine radar and electronic chart data according to the determined reference points. The proposed fusion is verified through simulations using ENC data and marine radar images from real ships in narrow waters over a continuous period. The results suggest a suitable performance for edge matching of the shoreline and real-time applicability. The fused image can provide comprehensive information to support navigation, thus enhancing important aspects such as safety.
This paper deals with the problem of considering both robust fault detection and accurate state estimation for a strapdown inertial navigation system and global navigation satellite system (SINS/GNSS)integrated navigation system for ships within a unified framework. Under the assumption that the process disturbance and measurement noise being 2 l-norm bounded, a cubature / i HH fault detection filter (FDF) considering both sensitivity and robustness to disturbance is proposed to detect faults with a postfilter and an observer gain. However, the estimated state value of the above algorithm is theoretically a suboptimal solution. To solve this limitation, an adaptive algorithm based on the measurement residual and a scaling factor is proposed to improve the accuracy of state estimation in the proposed cubature / i HH FDF. To guarantee stability when faults are extremely large, the framework uses a robust H-infinity cubature Kalman filter. Finally, a simulation is implemented to demonstrate the effectiveness of the proposed method, considering the threshold selection for different cases.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.