中圖分類(lèi)號(hào)： TP389.1 文獻(xiàn)標(biāo)識(shí)碼： A DOI： 10.19358/j.issn.2096-5133.2021.03.007 引用格式： 余慧明，周志祥，彭?xiàng)睿? 一種基于改進(jìn)Mask R-CNN模型的遙感圖像目標(biāo)識(shí)別方法[J].信息技術(shù)與網(wǎng)絡(luò)安全，2021，40(3)：38-42，47.
A remote sensing image target recognition method based on improved Mask R-CNN model
Yu Huiming，Zhou Zhixiang，Peng Yang，Cui Zhibin
(Wuhan Xing Tu Xin Ke Co.，Ltd.，Platform Products Department，Wuhan 430073，China)
Abstract： As an important branch in the field of machine vision, target recognition technology has important applications in various fields. In view of the fact that the general target recognition model does not perform well in remote sensing images, the number of targets that need to be recognized is large, and the sizes are different, based on the classic target recognition Mask R-CNN model, a Mask R-CNN model fusing channel attention mechanism and data enhancement technology is proposed. When inputting image data, the data through the Random-Batch images operation to improve the accuracy of the model′s recognition of targets of different sizes is first enhanced; then, when extracting features, the FPN in the original Mask R-CNN model is improved to BiFPN, so that the extracted features can better reflect the original picture information; in the final Mask stage, the channel attention mechanism is added, so that the model can get more information. Experiments show that this model is used in remote sensing images. It has a good performance in the fine-grained recognition of special targets. For the same data set, its evaluation indicators are superior to other comparison algorithms in all aspects.
Key words : target recognition；Mask R-CNN；channel attention；data enhancement
