Infrared image has the common problems of low resolution,blurry details and poor visual effect,making it difficult to directly apply them in PCB fault diagnosis system.To solve these problems,a hybrid residual dense network(HRDN)infrared image super resolution reconstruction model is proposed.Firstly,the CNN is used to extract shallow feature information of original low resolution(LR)image.Secondly,many hybrid residual dense blocks(HRDBs)with multi path structure are designed to extract more abundant deep feature information.Thirdly,global feature fusion and residual learning are used to adaptively learn global hierarchical features.Finally,deconvolution layers are integrated into the network to learn the up sampling filters and to complete the reconstruction.The experimental results show that the proposed algorithm can effectively improve the reconstructed infrared image resolution and the detail information,and enhance the visual effect.The PSNR and SSIM of the open databases and self collected databases are 42.17 dB/39.32 dB and 0.9503/0.9466 respectively,which are superior to the bicubic interpolation,SRCNN and ESPCN models listed in the paper.Reconstruction performance has been significantly improved.