基于联合注意力机制的SAR图像飞机目标检测方法

• Format: Patent
• Language: Chinese
• Published: 15.07.2025
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

本发明具体涉及基于联合注意力机制的SAR图像飞机目标检测方法，包括：将待检测的SAR图像输入训练后的目标检测模型，输出对应的目标检测预测值；训练时：通过深度神经网络提取SAR图像不同层次的深度特征图，再对应输入至用于提取局部和全局联合注意力特征的金字塔网络的对应联合注意力层，同时将金字塔网络中上层联合注意力层的输出与相邻下层联合注意力层输入的深度特征图拼接作为该相邻下层联合注意力层的输入；然后分别基于联合注意力特征图进行预测，得到对应的预测框和分类预测概率；最后生成目标检测预测值并进行模型训练。本发明能够有效融合SAR图像的局部注意力特征和全局注意力特征，进而能够兼顾飞机目标本身的特征信息和周围位置信息。 The invention specifically relates to an SAR image aircraft target detection method based on a joint attention mechanism, and the method comprises the steps: inputting a to-be-detected SAR image into a trained target detection model, and outputting a corresponding target detection prediction value; during training, depth feature maps of different levels of an SAR image are extracted through a deep neural network, and then the depth feature maps are correspondingly input to corresponding joint attention layers of a pyramid network used for extracting local and global joint attention features. Meanwhile, splicing the output of the upper-layer joint attention layer in the pyramid network and the depth feature m