Session 3 - Precision Agriculture: Towards Deep Learning for Weed Detection: Deep Convolutional Neural Networks Architectures for Plant Seedling Classification

Presentation Type

Oral

Student

Yes

Track

Precision Ag/Biological Sciences Application

Abstract

Traditional means of on-farm weed control has been known to use manual labor. This process is time consuming, costly and contributes to major yield losses. There are also environmental hazards with conventional or uniform application method of controlling weed infestation. To solve this using computer vision, researchers often use remote sensing weed maps, but this is ineffective due to problems such as solar and cloud cover in satellite imagery

In this study we leverage the automatic feature extraction capabilities of deep convolutional neural networks (DCNN) to classify plant seedlings. Theoretically, we demonstrate that DCNNs can successfully segment crops and weeds in various phenological growth stages, and identify limitations with these techniques that can further guide future research. In practice, this paper will be relevant to both researchers and producers of computer vision equipment, especially low-cost solution for ground-based site-specific weed control.

Key words: Deep learning, transfer learning, smart agriculture, weed detection, , plant seedling, classification

Start Date

2-11-2020 9:30 AM

End Date

2-11-2020 10:30 AM

This document is currently not available here.

Share

COinS
 
Feb 11th, 9:30 AM Feb 11th, 10:30 AM

Session 3 - Precision Agriculture: Towards Deep Learning for Weed Detection: Deep Convolutional Neural Networks Architectures for Plant Seedling Classification

Dakota Room 250 A/C

Traditional means of on-farm weed control has been known to use manual labor. This process is time consuming, costly and contributes to major yield losses. There are also environmental hazards with conventional or uniform application method of controlling weed infestation. To solve this using computer vision, researchers often use remote sensing weed maps, but this is ineffective due to problems such as solar and cloud cover in satellite imagery

In this study we leverage the automatic feature extraction capabilities of deep convolutional neural networks (DCNN) to classify plant seedlings. Theoretically, we demonstrate that DCNNs can successfully segment crops and weeds in various phenological growth stages, and identify limitations with these techniques that can further guide future research. In practice, this paper will be relevant to both researchers and producers of computer vision equipment, especially low-cost solution for ground-based site-specific weed control.

Key words: Deep learning, transfer learning, smart agriculture, weed detection, , plant seedling, classification