Detecting functional field units from satellite images in smallholder farming systems using a deep learning based computer vision approach: A case study from Bangladesh

Ruoyu Yang; Zia U Ahmed; Urs C Schulthess; Mustafa Kamal; Rahul Rai

doi:10.1016/j.rsase.2020.100413

Detecting functional field units from satellite images in smallholder farming systems using a deep learning based computer vision approach: A case study from Bangladesh

Remote Sens Appl. 2020 Nov:20:100413. doi: 10.1016/j.rsase.2020.100413.

Authors

Ruoyu Yang¹, Zia U Ahmed², Urs C Schulthess³, Mustafa Kamal⁴, Rahul Rai¹

Affiliations

¹ Department of Mechanical and Aerospace Engineering, 240 Bell Hall, University at Buffalo, Buffalo, NY, 14260-4400, USA.
² Research and Education in Energy, Environment and Water (RENEW) Institute, 112 Cook Hall, University at Buffalo, Buffalo, NY, 14260-1300, USA.
³ CIMMYT-Henan Collaborative Innovation Center, Henan Agricultural University, Zhengzhou, 450002, PR China.
⁴ CIMMYT-Bangladesh, House 10/B, Road 53, Gulshan 2, Dhaka, 1212, Bangladesh.

Abstract

Improving agricultural productivity of smallholder farms (which are typically less than 2 ha) is key to food security for millions of people in developing nations. Knowledge of the size and location of crop fields forms the basis for crop statistics, yield forecasting, resource allocation, economic planning, and for monitoring the effectiveness of development interventions and investments. We evaluated three different full convolutional neural network (F-CNN) models (U-Net, SegNet, and DenseNet) with deep neural architecture to detect functional field boundaries from the very high resolution (VHR) WorldView-3 satellite imagery from Southern Bangladesh. The precision of the three F-CNN was up to 0.8, and among the three F-CNN models, the highest precision, recalls, and F-1 score was obtained using a DenseNet model. This architecture provided the highest area under the receiver operating characteristic (ROC) curve (AUC) when tested with independent images. We also found that 4-channel images (blue, green, red, and near-infrared) provided small gains in performance when compared to 3-channel images (blue, green, and red). Our results indicate the potential of using CNN based computer vision techniques to detect field boundaries of small, irregularly shaped agricultural fields.

Keywords: CNN; Deep learning; Field boundaries; Smallholder farming.