Detecting uncut crop edge with convolutional neural networks

DOI

10.2991/ispc-19.2019.22How to use a DOI?

Keywords

advanced driver assistance systems, agricultural harvester, convolutional neural networks, semantic segmentation

Abstract

The paper proposes an approach to determining the boundary between cut and uncut crop, which is one of the most important tasks in creating a system of assistance to the combine operator. The approach is based on semantic segmentation of images that come from a camera installed in the combine cab. Semantic segmentation is performed using the ENet neural network, which is designed to work in real time. As a result of segmentation, 5 classes are recognized: cut crop, uncut crop, obstacles, harvester part, and background. The straight line of the boundary between cut and uncut crop is determined through linear regression. Testing of the performance of algorithm was carried out both on conventional CPU and on mobile ones. An implementation for mobile processors has been created that provides a performance of 3.61 FPS on an ARM Cortex-A53 processor. The algorithm's performance is sufficient to let the combine drive at 4 km/h.

Copyright

© 2019, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Denis Protasov
PY  - 2019/06
DA  - 2019/06
TI  - Detecting uncut crop edge with convolutional neural networks
BT  - Proceedings of the International Scientific and Practical Conference “Digital agriculture - development strategy” (ISPC 2019)
PB  - Atlantis Press
SP  - 98
EP  - 103
SN  - 1951-6851
UR  - https://doi.org/10.2991/ispc-19.2019.22
DO  - 10.2991/ispc-19.2019.22
ID  - Protasov2019/06
ER  -