Ship identification is of great significance in coastal area monitoring and port military reconnaissance. The characteristics of different types of ships in high-resolution images are highly similar, and the internal details of the target are clearer. In order to effectively use the target details in the high-resolution image for target recognition, based on the UAV images of Xingcheng wharf and bingjiawan in Huludao City, Liaoning Province, the UAV ship image data sets with coarse-grained and fine-grained classification modes are established respectively. Build a deep learning platform based on tensorflow, and use SSD algorithm to identify the ships in the data set. The recognition accuracy of ship image data set based on coarse-grained classification pattern is 90.87%; The accuracy of ship image data set based on fine-grained classification model is 94.63%. The results show that: SSD algorithm can realize ship recognition of UAV image and Image data set based on fine-grained classification pattern can effectively use the detailed features of high-resolution image targets, so as to improve the accuracy of ship recognition.
Research on UAV Image Ship Recognition Based on Fine-grained Classification Data Set
2021-11-12
4326126 byte
Conference paper
Electronic Resource
English
SFINet: An Oriented Fine-Grained Ship Identification Network Based on Remote Sensing Image
Springer Verlag | 2024
|