Study of the application of image super resolution technology in intelligent coal mine
In order to solve the problems of distorted mine image structure and missing detail information caused by low light,low visibility,and large coal dust under monitoring in Jinjie Coal Mine,a super-resolution reconstruction algorithm based on channel graph convolution and residual aggregation was proposed.By utilizing the residual aggregation network to fully obtain the hierarchical features between residual blocks,constructing the intrinsic correlations between feature map channels through channel graph convolution,and integrating the features output from channel graph convolution and attention branches into the residual network through a multi-head attention transfer module,the output result not only fully utilized the hierarchical features between each residual block,but also took advantage of the potential correlations between high-level features.The experimental results showed that compared with existing classical super-resolution algorithms,the image reconstructed by this algorithm had clear structure and rich detail information.
mine imagesuper-resolution reconstructionchannel graph convolutionresidual aggregationmulti-head attentionhierarchical features
NETL
NSTL
万方数据
