Restoration of motion blur image for detection robot by generative adversarial network

被引：0

作者：

Cui, Jifeng ^{[1
]}

Han, Jin ^{[1
]}

机构：

[1] College of Computer Science and Technology, Shandong University of science and technology, Qingdao,Shandong Province,266590, China

来源：

Journal of Network Intelligence | 2021年 / 6卷 / 02期

关键词：

Restoration - Generative adversarial networks - Image enhancement - Inspection - Robots;

D O I：

暂无

中图分类号：

学科分类号：

摘要：

The inspection robot is affected by its motion and the ground environment during its operation, and the collected images are prone to motion blur. The traditional algorithm has a large amount of calculation and a long processing time, which cannot meet the real-time needs of detection robots. For this problem, the present work is based on a Generative Adversarial Network. The generator uses a lightweight algorithm combining target detection FPN and depth separable convolution MobileNetV3. The discriminator uses a relative discriminator, and the loss function is composed of four mixed loss functions. Through the experimental verification on the Gopro data set and the actual collected data set, the processing speed of the lightweight algorithm given in the paper reaches 0.05s, which meets the real-time requirements of the inspection robot. It can effectively restore the motion blur images of the inspection robot and the restored images are more detailed and accurate. The lightweight algorithm requires fewer calculations, and the processing speed is faster. It can be directly run in the inspection robot, effectively improving its work efficiency, and has a certain reference value for image restoration of other embedded platforms. © 2021 Global Research Online. All rights reserved.

引用

页码：328 / 338