Ship target detection and fine-class recognition based on course-to-fine cascade neural networks
Shuxin Li, Hua Zhou · 2020
We proposed a new method to detect the ship targets in large scale remote sensing images, which can recognize the fine-class ship targets more than just detect the targets. The pipeline consists of cascade neural networks to reach higher detection and recognition performance. To detect better, we use different threshold in the cascade module, so that we can eliminate background interference better. We also use multi-direction boxes to detect the ship targets to get better performance. To recognize better, we firstly recognize ship category in a module, then we recognize the type in another module to get fine-class ship recognition results. Through experiments on the Google Earth remote sensing image data set collected by ourselves and HRSC2016 data set, it shows that the method is effective in ship target detection and fine-class recognition of remote sensing image, and the performance is improved compared with the previous algorithms.