Recognition and detection of ports and ships based on image application and non local feature enhancement
Dongsheng Li, Yucheng Zhou · 2022 IEEE 10th Joint International Information Technology and Artificial Intelligence Conference (ITAIC) · 2022
Remote sensing image target detection has a wide application demand in the civil field. Using remote sensing image to supervise port facilities, ships and other targets is of great significance. With the rapid development of deep learning, target detection has made remarkable commercial application achievements in the general field. The target detection method of remote sensing image originates from the target detection method in the general field. Combined with the characteristics of remote sensing image, the application of deep learning target detection method in the field of remote sensing image is the main method of target detection in recent years. Firstly, the coarse-grained remote sensing target detection is gradually enhanced from the coarse-grained feature of RC fast to the fine-grained feature of remote sensing. Then, the non local feature enhancement module is used to obtain the features after non local feature enhancement. Finally, the scale of the features enhanced by nonlocal features is adjusted in turn, and the results are used as output. The result extracts the features of the horizontal candidate box area, learns the geometric features of rboi, and roughly adjusts the bbox to obtain the rotating candidate box, so as to make up for the lack of correction samples and reduce the impact of the reduction of the number of positive samples on the performance of the detector. Use the above methods to design and develop the software platform. It realizes the monitoring and display of port ships, silos and other objects.