CNN -based Augmentation Using Image Processing Techniques for Low Light Characteristics Images

The underwater images particularly in the shrimp pond are foggy and it is difficult to see shrimps through the water. For this analysis, data augmentation by using different image processing techniques is proposed which can increase the data use for better accuracy in CNN. This research was performe...

Full description

Saved in:

Bibliographic Details
Published in:	2023 IEEE Industrial Electronics and Applications Conference (IEACon) pp. 249 - 254
Main Authors:	Fauzi, Normasni Ad, Isa, Iza Sazanita, Hamzah, Siti Nur Mardhiah, Darus, Rosheila
Format:	Conference Proceeding
Language:	English
Published:	IEEE 06-11-2023
Subjects:	augmentation CNN Gaussian noise Gray-scale Histograms Image processing Industrial electronics low-lights Object detection Training underwater vision
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	The underwater images particularly in the shrimp pond are foggy and it is difficult to see shrimps through the water. For this analysis, data augmentation by using different image processing techniques is proposed which can increase the data use for better accuracy in CNN. This research was performed on shrimp dataset and underwater dataset to assess the classification accuracy of the proposed network. Therefore, data augmentation was conducted in order to evaluate the effect of image enhancement algorithms for underwater images in the CNN performance. In the data augmentation, image processing techniques used Grayscale Conversion and Gaussian noise whilst histogram equalization was applied in order to improve the contrast of images. The result shows a significant improvement as the performance of training accuracy is 99.1% and 92.8 % for classification of the underwater object. Therefore, this study proposed different image processing techniques as part of data augmentation method for underwater images detection. This study may evolve the intelligence method in deep learning to enhance technology advancement in aquaculture field and food security area.
DOI:	10.1109/IEACon57683.2023.10370357