Transmission Line Target Detection Method Based On Super-resolution Reconstruction

When performing target detection on transmission line images,image quality is an important prerequisite for ensuring accurate detection.However,when UAV power inspections are affected by factors such as the environment and hardware equipment,they are prone to jitter,motion blur,etc.,and small targets are in the image.It occupies a small number of pixels,which makes the captured image unable to meet the needs of further processing.To improve image quality and ensure target detection accuracy,super-resolution reconstruction algorithm is an effective way to improve.In this paper,researches on image super-resolution reconstruction technology and target detection methods are carried out.The main contents are as follows:This paper analyzes and studies commonly used image super-resolution reconstruction algorithms.For the deep recursive residual network super-resolution reconstruction algorithm(DRRN),there is a single feature extraction problem,and a recursive network super-resolution based on multi-scale feature fusion is proposed.The reconstruction algorithm uses a multi-feature extraction module to replace the original local residual structure to extract more shallow information;at the same time,a dense connection structure is added to the network to enhance feature propagation,reduce parameter calculations,and effectively alleviate the problem of gradient disappearance.The method proposed in this paper is applied to the image scene of the transmission line.The experimental results show that the image generated by the algorithm proposed in this paper has richer detailed information than the original algorithm,which can effectively improve the image quality,peak signal-to-noise ratio(PSNR)and structure The similarity(SSIM)value has a certain improvement compared with the DRRN algorithm.The image super-resolution reconstruction method is combined with YOLOv5 target detection to form a reconstruction-based target detection method.Comparative experiments are designed to perform target detection on the images before and after reconstruction.The experimental results show that the reconstructed transmission line image target detection The accuracy and m AP value are respectively increased by 1.8%and 21.9% compared with the corresponding values of low-resolution image target detection,which proves the effectiveness of the algorithm in this paper.