Smart drone scouting app for precision agriculture using whole field based deep learning techniques

Project Title

Smart drone scouting app for precision agriculture using whole field based deep learning techniques

Project Area of Specialization Computer ScienceProject Summary

Over the past few decades, the global population is growing tremendously. This raise demanded agriculture to meet these trends with proper balancing of energy, fertilizer, labor, and cost. Precision agriculture(PA) is considered to be one of the great solutions for handling such issues. The involvement of mobile and drone technologies enhances the impact of precision agriculture and provides more valuable results. The technological development in the agricultural area of Pakistan currently limited itself to the terms cultivation, soil fertility, terrestrial environment, etc. the role of advanced technology especially Drone, Deep learning and remote sensing based on color indices needs to be addressed. Moreover, the utilization of such technologies with full benefits is very difficult for the Pakistani farmers to adopt with their limited educational background and knowledge.  To overcome the above-mentioned gaps we propose a smart scouting approach that will facilitate the farmers’ predicting the crop health, reduce labor costs, and increase agricultural profits. For the successful implementation of the proposed solution, we used various tools (Python, PHP, My SQL, and Android Studio), techniques (vegetation indices, AI, and CNN), and technologies (Drone, and DL). Real images will be taken from drone. 70% of the data sets will be used for training the model whereas 30% are used for testing purposes. The basic modules of the project will be Mobile Application Manager, Image Acquisition, Smart Scouting, and Crop Health Prediction. This project will not only facilitate the farmers and agriculture-associated personnels but also help the researchers in their further research.

Project Objectives

The objectives of the proposed project are:
•    A literature review will be done in this project which helps to identify the suitable tools, techniques, and methodologies essential for the project.
•    A Mobile Application will be developed which provides drone scouting information to the user. Through this application, users can manage and control the route of a drone easily and efficiently.  This application will also help in the collection of study datasets which later be stored in the database. 
•    Customized vegetation indices algorithm will be developed which facilitates the identification of the healthy and unhealthy parts of the crop. This algorithm will generate datasets that further be used to train the proposed model.  
•    A crop health prediction model will be created by using deep learning techniques (i.e. CNN). In preprocessing phase the whole field will be decomposed into fixed zones and each zone is handled individually. The model will facilitate the scouting section to move according to the whole field's predicted maps. 
•    A drone scouting algorithm will be created which utilizes the proposed predicted model and predicted the whole field map on the bases of local field information.  
•    More than 1000 data sets will be taken through drones out of the 70% will be used for training and the remaining 30% will be used for testing.
•    The effectiveness of the final experimental results will be tested and verified by the proposed mobile application connected with the cloud database.

Project Implementation Method

The proposed smart scouting application will provide an effective and efficient mechanism for routing the drone on the field. The block diagram of the proposed idea is presented in Figure 1. The whole idea is decomposed into four main modules. The mobile application module provides an interface to the user where they can route the drone easily and efficiently. The system facilitates users by providing them the predicted maps through which they can see the whole field information in a single click. The crop health prediction module will train the predicted model by using the vegetation indices, AI, and CNN techniques.  In the smart scouting module, an AI-based routing algorithm will be used to generate and display information to the user.     Finally, experimental results are verified using a web-based mobile application that is connected to a database server (i.e. My SQL ).

Figure 1: Block Daigram Of Proposed Project

Benefits of the Project

The basic benefits provided by the proposed project are:
•    The literature review done in this project facilitates the researchers and practitioners to further investigate this area in different scenarios and contexts. 
•    The proposed idea will enhance the crops’ productivity by reducing drone battery consumption, labor costs, and increasing agricultural profits 
•    It will also help in monitoring the crops with more accuracy and less complexity. 
•    The proposed smart scouting mechanism avoids surveying 100% of a field which will ultimately save time and energy.
•    The ease and simplicity of interfaces provided by the software will enhance the farmer’s interaction and knowledge.  
•    The datasets generated by the proposed software using machine learning techniques will be used for further research.

Technical Details of Final Deliverable

In the end, we will deliver the code and the documentation in the form of a final project report containing all the proposed algorithms pseudo and implementation code.

Final Deliverable of the Project Software SystemCore Industry ITOther Industries Agriculture Core Technology Artificial Intelligence(AI)Other Technologies Cloud InfrastructureSustainable Development Goals Industry, Innovation and Infrastructure, Responsible Consumption and Production, Life on LandRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	80000
Drone	Equipment	1	55000	55000
Mobile phone	Equipment	1	15000	15000
Documentation	Miscellaneous	1	10000	10000