Using UAV Image Histogram Kurtosis and Skewness to Automatically Differentiate Typhoon-Damaged Rice Field Regions from Undamaged Regions

Typhoons can cause extreme damages to rice fields causing revenue loss for farmers if not addressed early. However, timely damage assessment is difficult due to the scale of rice fields. In this study, we utilized images captured by a commercially available unmanned aerial vehicle (UAV) to create a...

Full description

Saved in:
Bibliographic Details
Main Authors: Eclarin, Niño, Fernandez, Proceso
Format: text
Published: Archīum Ateneo 2023
Subjects:
Online Access:https://archium.ateneo.edu/discs-faculty-pubs/400
https://doi.org/10.1063/5.0124485
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Ateneo De Manila University
id ph-ateneo-arc.discs-faculty-pubs-1400
record_format eprints
spelling ph-ateneo-arc.discs-faculty-pubs-14002024-02-20T04:53:04Z Using UAV Image Histogram Kurtosis and Skewness to Automatically Differentiate Typhoon-Damaged Rice Field Regions from Undamaged Regions Eclarin, Niño Fernandez, Proceso Typhoons can cause extreme damages to rice fields causing revenue loss for farmers if not addressed early. However, timely damage assessment is difficult due to the scale of rice fields. In this study, we utilized images captured by a commercially available unmanned aerial vehicle (UAV) to create a model that can identify rice plant lodging caused by typhoons. Local officials helped gather and establish ground truth data captured in a period of seven to ten days after a typhoon. Rectangular annotations of the 79 identified damaged portions were extracted. These, together with an equal number of randomly selected undamaged portions with matching areas under the same image comprised the image clippings dataset. The corresponding 8-feature numeric dataset was derived from the kurtosis and skewness of the image histograms in four color channels (red, blue, green and greyscale) produced after a 3x3 median filter was applied to every image clipping instance. Using the numeric dataset, several machine learning models were explored to classify the damaged and damaged clippings. Results from Tenfold Stratified Cross Validation showed the MLPClassifier has an accuracy of 79.25% and F-score of 80.24%, with the Support Vector Machine at 79.21% accuracy and 78.72% F-score, while the Random Forest and Naive Bayes Classifier performed almost similarly with a 74.75% and 76.08% accuracy and 74.69% and 74.62% F-score. This shows that a model can be created to distinguish damages caused by typhoons on images captured by commercial UAVs. 2023-05-16T07:00:00Z text https://archium.ateneo.edu/discs-faculty-pubs/400 https://doi.org/10.1063/5.0124485 Department of Information Systems & Computer Science Faculty Publications Archīum Ateneo Aircraft machine learning Computer Engineering Electrical and Computer Engineering Engineering
institution Ateneo De Manila University
building Ateneo De Manila University Library
continent Asia
country Philippines
Philippines
content_provider Ateneo De Manila University Library
collection archium.Ateneo Institutional Repository
topic Aircraft
machine learning
Computer Engineering
Electrical and Computer Engineering
Engineering
spellingShingle Aircraft
machine learning
Computer Engineering
Electrical and Computer Engineering
Engineering
Eclarin, Niño
Fernandez, Proceso
Using UAV Image Histogram Kurtosis and Skewness to Automatically Differentiate Typhoon-Damaged Rice Field Regions from Undamaged Regions
description Typhoons can cause extreme damages to rice fields causing revenue loss for farmers if not addressed early. However, timely damage assessment is difficult due to the scale of rice fields. In this study, we utilized images captured by a commercially available unmanned aerial vehicle (UAV) to create a model that can identify rice plant lodging caused by typhoons. Local officials helped gather and establish ground truth data captured in a period of seven to ten days after a typhoon. Rectangular annotations of the 79 identified damaged portions were extracted. These, together with an equal number of randomly selected undamaged portions with matching areas under the same image comprised the image clippings dataset. The corresponding 8-feature numeric dataset was derived from the kurtosis and skewness of the image histograms in four color channels (red, blue, green and greyscale) produced after a 3x3 median filter was applied to every image clipping instance. Using the numeric dataset, several machine learning models were explored to classify the damaged and damaged clippings. Results from Tenfold Stratified Cross Validation showed the MLPClassifier has an accuracy of 79.25% and F-score of 80.24%, with the Support Vector Machine at 79.21% accuracy and 78.72% F-score, while the Random Forest and Naive Bayes Classifier performed almost similarly with a 74.75% and 76.08% accuracy and 74.69% and 74.62% F-score. This shows that a model can be created to distinguish damages caused by typhoons on images captured by commercial UAVs.
format text
author Eclarin, Niño
Fernandez, Proceso
author_facet Eclarin, Niño
Fernandez, Proceso
author_sort Eclarin, Niño
title Using UAV Image Histogram Kurtosis and Skewness to Automatically Differentiate Typhoon-Damaged Rice Field Regions from Undamaged Regions
title_short Using UAV Image Histogram Kurtosis and Skewness to Automatically Differentiate Typhoon-Damaged Rice Field Regions from Undamaged Regions
title_full Using UAV Image Histogram Kurtosis and Skewness to Automatically Differentiate Typhoon-Damaged Rice Field Regions from Undamaged Regions
title_fullStr Using UAV Image Histogram Kurtosis and Skewness to Automatically Differentiate Typhoon-Damaged Rice Field Regions from Undamaged Regions
title_full_unstemmed Using UAV Image Histogram Kurtosis and Skewness to Automatically Differentiate Typhoon-Damaged Rice Field Regions from Undamaged Regions
title_sort using uav image histogram kurtosis and skewness to automatically differentiate typhoon-damaged rice field regions from undamaged regions
publisher Archīum Ateneo
publishDate 2023
url https://archium.ateneo.edu/discs-faculty-pubs/400
https://doi.org/10.1063/5.0124485
_version_ 1792202633509863424