Image Defogging Method for Transmission Channel Inspection by UAV Based on Deep Multi-patch Layered Network

Unmanned aerial vehicle (UAV) checking transmission channel in the process, when the camera to collect images easily affected by fog particles, the images quality, affect the accuracy of transmission channel safety hazard identification. However, in the real environment, haze particles are not evenly distributed. Therefore, an image de-fogging algorithm based on deep multi-patch layered network (DMPHN) was proposed for transmission channel images taken by UAV. DMPHN is a multi-level structure. A single level consists of encoder and decoder pairs of multi-scale residuals. Among the three levels, the feature information of the upper level of the input image rain is fused to achieve a coarse to fine end to end fog removal. Experiments show that DMPHN is effective under both objective and subjective evaluation. The average processing time of 1600*1200 high-resolution single image is 0.035 seconds.