基于改进卷积神经网络的可见光航空图像解译网络研究OA

While aerial sensor technology improves detection capabilities,it also faces challenges such as data proliferation and modal fusion.Aiming at the problems of low target discrimination and strong background interference in visible light images in complex scenes,this research proposes a new interpretation framework.By integrating doubling convolution,multi-scale attention and non-local filtering modules in the fully convolutional network,feature extraction and semantic segmentation capabilities are significantly enhanced.Subsequently,selective search is used to generate candidate regions,combined with convolutional neural network classification,and the interpretation results are optimized through non-maximum suppression.Experiments show that this method significantly improves target discrimination and complex background recognition rate,providing effective technical support for remote sensing monitoring,national defense security and other fields.