Remote sensing image target detection method

• Main Authors: LU JUN, LIM WOO-JUN, RAO ZIYU, DING LEI, WANG JIAQI, LIU XIANGYUN, GUO HAITAO, YU DONGXING, NIU YITING, JIN GUOWANG, ZHANG HENG, GONG ZHIHUI
• Format: Patent
• Language: Chinese; English
• Published: 13.06.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention relates to a remote sensing image target detection method, and belongs to the technical field of remote sensing image processing. According to the method, a target detection model comprising a positive and negative feature guide module and a global feature information complementation module is constructed, the positive and negative feature guide module is utilized to adaptively perceive the contribution degree of spatial positions and semantic information of beneficial features, and background noise information is removed from feature flow of the model; interference of noise features in a complex background is suppressed, and target features are highlighted and refined; a global feature information complementation module is utilized to compress deep abstract features through pooling operation, the ability of the model to resist feature displacement and rotation is improved, cavity convolution is utilized to increase the receptive field of the model under the condition that the spatial resolution