Comparison of different classification methods in estimating sugar beet canopy cover fraction using drone images

Articles in Press

Document Type : Research Paper

Authors

1 abureihan university

2 PhD Std. Irr. & Dra. Eng. Water Sci. & Eng. Dept. IKIU

3 Water and Science Eng. Dept. IKIU

10.22059/jwim.2024.383445.1181

Abstract

Canopy cover fraction (CCF) is one of vital parameters to determine crop appearance and stress detection. Recent advancement in technologies and availability of digital camera with high quality provide suitable condition for monitoring and determining canopy cover fraction during whole growing season without disturbing. In this study sugar beet aerial photos taken from research field of science institute in Switzerland was used. A number of 481 images were taken in the visible spectrum band at an average height of 10 meters above the ground using a DJI MATRICE 100 drone on four different dates. To determine the canopy cover fraction, five supervised classification methods, including Mahalanobis distance (MahD), maximum likelihood (MaxLh), minimum distance (MinD), neural network (NN) and support vector machine (SVM) were evaluated. The results showed that SVM and MaxLh methods with an overall accuracy (OA) of 99% had the best results in image classification and CCF calculation. The comparison of the obtained results for all imaging dates showed that in terms of processing time, the MaxLh method with a relatively simple mechanism is the most appropriate method in estimating the sugar beet CCF and compared to the methods based on machine learning, like NN and SVM, it is faster and can be used as an alternative method with high accuracy and very close to machine learning methods.

Keywords

Main Subjects

Water and Irrigation Management

Articles in Press, Accepted Manuscript
Available Online from 16 December 2024

Files

Share

How to cite

Statistics