A Novel Attentive Generative Adversarial Network for Waterdrop Detection and Removal of Rubber Conveyor Belt Image

Xinxin Feng,Debao Zhou,Bin Li,Xiao Liu,Xianguo Li,Zongpeng Liu

doi:10.1155/2020/1037021

Abstract

The lens for monitoring the rubber conveyor belt is easy to adhere to a large number of water droplets, which seriously affects the image quality and then affects the effect of fault monitoring. In this paper, a new method for detecting and removing water droplets on rubber conveyor belts based on the attentive generative adversarial network is proposed to solve this problem. First, the water droplet image of the rubber conveyor belt is input into the generative network composed of a cyclic visual attentive network and an autoencoder with skip connections, and an image of removing water droplets and an attention map for detecting the position of the water droplet are generated. Then, the generated image of removing water droplets is evaluated by the attentive discriminant network to assess the local consistency of the water droplet recovery area. In order to better learn the water droplet regions and the surrounding structures during the training, the image morphology is added to the precise water droplet regions. A dewatered rubber conveyor belt image is generated by increasing the number of circular visual attention network layers and the number of skip connection layers of the autoencoder. Finally, a large number of comparative experiments prove the effectiveness of the water droplet image removal algorithm proposed in this paper, which outperforms of Convolutional Neural Network (CNN), Discriminative Sparse Coding (DSC), Layer Prior (LP), and Attention Generative Adversarial Network (ATTGAN).

Highlights

Rubber conveyor belts [1] have been widely used in coal, mining, port, and other fields, mainly for the transportation of bulk, granular, and powdery solid materials
Spraying water is required to reduce the coal dust, which makes the monitoring of water droplets in the lens more common. erefore, how to effectively remove the water droplets on the rubber conveyor belt monitoring image to ensure the sharpness of the image is an important issue to be solved
We propose a new method for detecting and removing water droplets from rubber conveyor belts based on the attentive generative adversarial network [9]

Summary

Introduction

Rubber conveyor belts [1] have been widely used in coal, mining, port, and other fields, mainly for the transportation of bulk, granular, and powdery solid materials. We propose a new method for detecting and removing water droplets from rubber conveyor belts based on the attentive generative adversarial network [9]. E generated dewater droplet rubber conveyor belt image is input into the attentive discriminator together with the true clear background image to judge the true and false area of the water drop, and the optimizer and loss function which are most suitable for the discriminator are designed. E TFRecord file stores binary data and label data (rubber conveyor belt with water droplets and no water droplet images) in the same folder, without compressing the data and quickly loading them into memory, improving network training efficiency. Data normalization mainly includes generating TFRecord format files and input normalization [11]. e TFRecord file stores binary data and label data (rubber conveyor belt with water droplets and no water droplet images) in the same folder, without compressing the data and quickly loading them into memory, improving network training efficiency. e data normalization classifies the input color image pixel values from [0, 255] to [− 1, 1], which match the pretraining model VGG16 [12] of the network to avoid the training loss explosion and accelerate the gradient descent to improve the convergence speed of the generative adversarial network (GAN) [10] model

Algorithm Implementation

Network Design

Experiments