To improve the navigation safety of inland river ships and enrich the methods of environmental perception, this paper studies the recognition and depth estimation of inland river ships based on binocular stereo vision (BSV). In the stage of ship recognition, considering the computational pressure brought by the huge network parameters of the classic YOLOv4 model, the MobileNetV1 network was proposed as the feature extraction module of the YOLOv4 model. The results indicate that the mAP value of the MobileNetV1-YOLOv4 model reaches 89.25%, the weight size of the backbone network was only 47.6 M, which greatly reduced the amount of computation while ensuring the recognition accuracy. In the stage of depth estimation, this paper proposes a feature point detection and matching algorithm based on the ORB algorithm at sub-pixel level, that is, firstly, the FSRCNN algorithm was used to perform super-resolution reconstruction of the original image, to further increase the density of image feature points and detection accuracy, which was more conducive to the calculation of the image parallax value. The ships’ depth estimation results indicate that when the distance to the target is about 300 m, the depth estimation error is less than 3%, which meets the depth estimation needs of inland ships. The ship target recognition and depth estimation technology based on BSV proposed in this paper makes up for the shortcomings of the existing environmental perception methods, improves the navigation safety of ships to a certain extent, and greatly promotes the development of intelligent ships in the future.

Ship position prediction plays a key role in the early warning and safety of inland waters and maritime navigation. Ship pilots must have in-depth knowledge of the future position of their ship and target ship in a specific time period when maneuvering the ship to effectively avoid collisions. However, prediction accuracy and computing efficiency are crucial issues that need to be worked out at present. To solve these problems, in this paper, the deep long short-term memory network framework (LSTM) and genetic algorithm (GA) are introduced to predict the ship trajectory of inland water. Firstly, the collected actual automatic identification system (AIS) data are preprocessed and a series of typical trajectories are extracted from them; then, the LSTM network is used to predict the typical trajectories in real time. Considering that the hyperparameters of the LSTM network have difficulty obtaining the optimal solution manually, the GA is used to optimize hyperparameters of LSTM; finally, the GA-LSTM trajectory prediction model is constructed with the optimal network parameters and compared with the traditional support vector machine (SVM) model and LSTM model. The experimental results show that the GA-LSTM model effectively improves the accuracy and speed of trajectory prediction, with outstanding performance and good generalization, which possess certain reference values for the development of collision avoidance of unmanned ships.

This paper proposes to check the travel target of the dynamic background in the video surveillance with a fixed camera. A travel target detection method based on video picture acquisition and scene semantics for surveillance video was proposed. First, on the basis of combing the concepts and methods of picture recognition, the semantic information of the scene was fused to eliminate the interference factors in the unnecessary detection area. Secondly, a remote sensing picture visual feature representation method containing a semantic recognition method of remote sensing picture scenes and CSIFT features based on PLSA was presented. 10 types of typical remote sensing picture scenes are used for tests, and the visual vocabulary extraction method remains the same. The fixed visual vocabulary was 600, and the potential semantic subjects changes between 8∼50. The test results indicated that the highest average recognition rate was obtained when the latent semantic topics were 20. Inappropriate latent semantic topics will lead to a decline in recognition rates. The effectiveness of this method was fully verified.

No abstract