SlimRGBD: A Geographic Information Photography Noise Reduction System for Aerial Remote Sensing

Abstract

In the past ten years, civil drone technology has developed rapidly, and UAV (Unmanned Aerial Vehicle) has been widely used in various industries. Especially in the field of aerial remote sensing, the emergence of UAV technology has enabled the geographical information of remote areas that are not concerned to be quickly presented. However, UAV aerial photography is greatly affected by the weather. Pictures that use aerial drones for aerial photography in rainy weather will appear noise. In this paper, how to eliminate the noise of aerial image is to be talked, the multi-channel pruning technology is used to pruning the RnResNet network. Based on this, a new anti-convergence-convolution neural network noise reduction system for the operation of UAV airborne embedded equipment is proposed. The system is used to eliminate noise in the aerial image. This type of noise reducer has got rid of the current situation that the neural network noise reducer consumes too much power and is inefficient, and has certain advantages. INDEX TERMS SlimRGBD, ResNet, generative adversarial networks, image noise reduction, UAV, channel pruning, sparse training.

Compacting Deep Neural Networks for Internet of Things: Methods and Applications

“…The quality of images taken by drones is easily affected by weather and ambient light. In order to obtain a better quality of remote sensing images, Wu et al [185] designed a model namely SlimRGBD, which enables drones to automatically implement denoising operations when they took unclear photos. They trained a GAN model in advance to generate enough noisy images for the drone to learn the noise distribution.…”

Practical speed measurement for an intelligent vehicle based on double Radon transform in urban traffic scenarios

“…However, if the relative speed between the vehicle-mounted camera and the objects is very high, the acquired images are often subject to global motion blur, which significantly reduces the quality of the images and produces inaccurate information. Therefore, it is imperative to investigate the mechanism of motion blur to restore blurred images for the purpose of reducing the blur effect [6][7][8].…”