A Comparative Approach of Fuzzy Object Based Image Analysis and Machine Learning Techniques Which Are Applied to Crop Residue Cover Mapping by Using Sentinel-2 Satellite and UAV Imagery

Payam Najafi,Hossein Navid,Bakhtiar Feizizadeh

doi:10.3390/rs13050937

Abstract

Conservation tillage methods through leaving the crop residue cover (CRC) on the soil surface protect it from water and wind erosions. Hence, the percentage of the CRC on the soil surface is very critical for the evaluation of tillage intensity. The objective of this study was to develop a new methodology based on the semiautomated fuzzy object based image analysis (fuzzy OBIA) and compare its efficiency with two machine learning algorithms which include: support vector machine (SVM) and artificial neural network (ANN) for the evaluation of the previous CRC and tillage intensity. We also considered the spectral images from two remotely sensed platforms of the unmanned aerial vehicle (UAV) and Sentinel-2 satellite, respectively. The results indicated that fuzzy OBIA for multispectral Sentinel-2 image based on Gaussian membership function with overall accuracy and Cohen’s kappa of 0.920 and 0.874, respectively, surpassed machine learning algorithms and represented the useful results for the classification of tillage intensity. The results also indicated that overall accuracy and Cohen’s kappa for the classification of RGB images from the UAV using fuzzy OBIA method were 0.860 and 0.779, respectively. The semiautomated fuzzy OBIA clearly outperformed machine learning approaches in estimating the CRC and the classification of the tillage methods and also it has the potential to substitute or complement field techniques.

Highlights

Soils as the critical part of the agricultural systems are providing about 98.8% of human food [1]
Conservation tillage methods through leaving the previous crop residue cover (CRC) on the soil surface protect it from water and wind erosions
We considered a comparative approach for a novel fuzzy object based image analysis (OBIA) method and common pixel-based machine learning methods for the estimation of the CRC from previous cultivated winter wheat after the tillage and planting practices

Summary

Introduction

Soils as the critical part of the agricultural systems are providing about 98.8% of human food [1]. Global population growth rate amounts to around 1.1 percent per year, and it is expected to reach from 7.8 billion in 2020 to 9.8 billion by 2050 and 11.2 billion by 2100 [1] This will pressure on the soils to increase food production by 70% to achieve global food security [2]. The use of intensive tillage equipment for field preparation practices simultaneously to the development of mechanization in the agriculture that stabilized conventional tillage methods in the farming systems. Conventional tillage systems prepared a soft and free weed seed seedbed due to the intensive tillage practices that resulted in completely burying of the previous crop residue, these methods intensively have impacted the soil resources for decades [4,5]. Intensive tillage methods led to drop in soil organic

Objectives

Methods

Discussion

Conclusion