An Intelligent Health Assessment System for Crops Using Drone

Project Title

An Intelligent Health Assessment System for Crops Using Drone

Project Area of Specialization Artificial IntelligenceProject Summary

Agriculture is the process of cultivation of plants and livestock. In Pakistan, the contribution of agriculture is massive towards the economy since 48% of the labor is directly associated with it. Crop monitoring is essential to obtain a healthy and substantial-end product. We propose developing a standalone prototype for the health assessment of crops using computer vision and machine learning. For this purpose, we target the two important crops: Potato and Cotton. Potato is widely cultivated and largely used in industry to produce ready-to-use consumer products, while cotton plays a vital role in the garments industry and consequently contributes to the economy as exports. The developed prototype will detect the unhealthy/diseased plants in real-time with the help of an attached camera and onboard trained convolutional neural network (CNN). Thanks to recent advances in computer vision, several deep learning libraries as well as image datasets are available to train the CNN model of interest. The Nvidia Jetson Nano Kit will be used for real-time video processing. It is a single-board computer developed by Nvidia Corp. specifically to perform computationally expensive computer vision tasks in real-time as well as employing deep learning algorithms. A drone will carry the prototype and hover over the plants for real-time detection of unhealthy ones and their marking while covering a wider geographical area.

Project Objectives

The main objectives of the project are:

• To collect the labeled image datasets of the healthy and unhealthy leaves of cotton and potato plants.
• To develop the software model by training and optimizing the individual convolutional neural networks for potato and cotton crops using labeled image data on a laptop computer.
• To import the trained models on the hardware (single-board computer) and optimize it for real-time testing in the field.

Project Implementation Method

The implementation method has two parts:

Software Model development (Convolutional neural network)

• The labeled image datasets of cotton and potato crops will be collected having both healthy and diseased image samples.
• Image data will be pre-processed and normalized.
• The CNN models will be trained for the classification task. For this purpose, the available trained model such as Resnet 50, will be used and optimized for the task via transfer learning. In case of low accuracy, a customized model will be developed.
• The model will be evaluated on the test data and classification results will be recorded.
• The Python programming environment will be used with Keras library for deep learning model development.

• Hardware Development

• The Nvidia Jetson Nano Kit will be used as a single-board computer that is specifically designed to perform deep learning-based computer vision tasks in real-time.
• The operating system will be installed on the hardware and trained models will be imported to it.
• A camera will be interfaced to get the live video which will be fed to the trained model.
• The hardware will be tested on the field for real-time detection of unhealthy plants.
• Finally, the hardware will be carried by the drone to cover larger fields. The unhealthy plants will be localized and marked by the drone.

Benefits of the Project

The benefits of the project are summarized as follows:

• Large scale screening of plants
• Replaces the manual inspection method
• No need for a plant expert for disease detection
• The standalone and automatic health assessment system
• Real-time processing and classification of plants
• Intelligent and efficient system equipped with state of the art artificial intelligence capabilities
• Can be used with the robot for on-ground screening

Technical Details of Final Deliverable

Final Technical Deliverables of the project are:

Soft-form Deliverables:

• Image datasets of healthy and diseased plant leaves of cotton and potato.
• The trained CNN model for health classification of the two plants.

Hard-form Deliverables:

• Prototype hardware (Nvidia Jetson Nano Kit) with a camera, memory card, battery, an LCD display, and wireless network adapter.  
• A display device for remote monitoring of real-time processing
• A drone to carry the prototype for large scale plants screening.
• A flying intelligent health assessment system for cotton and potato crops.   

Final Deliverable of the Project HW/SW integrated systemCore Industry ITOther Industries Agriculture Core Technology Artificial Intelligence(AI)Other Technologies Robotics, OthersSustainable Development Goals Decent Work and Economic Growth, Industry, Innovation and InfrastructureRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	79700
Jetson Nvidia nano Kit with IMX219-77 camera package	Equipment	1	29000	29000
INUI Power Bank	Equipment	1	5500	5500
Jetson Nano UPS Power Module for 5V	Equipment	1	4000	4000
Quadcopter	Equipment	1	30000	30000
Rechargeable Battery	Equipment	4	300	1200
Delivery Charges, Travelling, Printing	Miscellaneous	1	10000	10000