In the process of UAV inspection, jitter often occurs, which leads to motion blur of camera when shooting images. Image blur will cause image content distortion, and even lead to the loss of edge information of target objects. The lack of image detail information will make the subsequent image processing extremely difficult and reduce the accuracy of image target detection, which hinders the popularization and practicability of UAV inspection technology. In this paper, the generation mechanism of image motion blur is studied, and a scale cycle network based image deblurring method is proposed. In this method, the image deblurring is divided into three sub problems, and three scale networks are used to realize image deblurring step by step, and LSTM network is used to realize sharing and continuous optimization of network parameters of different scales. Experimental results show that the proposed method can effectively remove image blur and suppress image ringing effect. Compare to the traditional methods, more complex motion blur problems can be solved.