Deep learning techniques to classify agricultural crops through UAV imagery: a review

Area of Science:

• Agricultural Science
• Remote Sensing
• Computer Science

Background:

• Unmanned Aerial Vehicles (UAVs) are increasingly utilized in agriculture to boost productivity and reduce costs.
• UAV-based remote sensing provides valuable data for precision agriculture applications like crop classification.
• Deep Learning (DL) approaches, particularly Convolutional Neural Networks (CNNs), are state-of-the-art for image processing.

Purpose of the Study:

• To review recent CNN-based methods for crop/plant classification using UAV remote sensing imagery.
• To assist researchers and farmers in selecting appropriate algorithms based on crops and hardware.
• To identify challenges and potential solutions for improving DL-based crop classification from UAV data.

Main Methods:

• Review of existing literature on CNN applications in UAV-based remote sensing for crop classification.
• Analysis of different data fusion techniques combining UAV data and DL approaches.
• Identification of key challenges and proposed solutions in the field.

Main Results:

• CNNs demonstrate remarkable performance in image processing for crop classification.
• Fusing diverse UAV data with DL methods improves the accuracy of crop type classification.
• Several challenging issues in UAV-based crop classification have been identified.

Conclusions:

• CNNs are powerful tools for analyzing UAV remote sensing data for crop classification.
• Data fusion strategies enhance the accuracy of classifying different crop types.
• Addressing identified challenges can further improve DL algorithm performance in precision agriculture.