Designing a Plant Diagnostic System

Introduction

Farming is known as a practice that is often unpredictable as nature controls everything about it. From the weather to the soil fertility, a lot of dynamic factors need to be checked upon and updated to ensure the success of the farm. This is where machine learning IoTs can come to use — to predict these factors in the future by analyzing data and patterns that help us identify the mistakes and flaws of a plant during its growth.

One example of a state-of-the-art product on agriculture IoT with ML is Taranis. Taranis is an AI platform that utilizes machine learning programs as well as IoT components, like sensors, to analyze farms with images taken by drones or satellites. Its algorithms then process the data collected to generate possible actions that may be taken to improve the state of agriculture in a specific field. Another example is the See & Spray developed by Blue River Technology. It also uses machine learning programs and sensors, this time to differentiate between weeds and crops. As a result, the spray only releases chemical herbicides on the weeds, which is extremely useful for regulating unwanted plants found on farms.

Despite the great effectiveness of such technologies, they still face multiple challenges when exposed to today’s technological society. One problem is, as many other data technologies also encounter, the risk of exposed data security and privacy. The data collected may be sensitive in certain cases and is very likely to be vulnerable to being breached unauthorizedly. But more importantly, these ML IoTs may have limited scalability, meaning that they have a restricted amount of resources they can use to process their data. Additionally, its complexity may also become an issue in some cases. These machine learning models especially can give developers a difficult time in interpreting the code and diagnosing issues if necessary; it may also lead to inaccuracy in data as a result of overfitting. In general, the complexity makes it more of a challenge to organize and manage its lifecycles and hardware components.

The aim is to create a program that can analyze images of plants from public datasets to identify if a plant is unhealthy or harmed. I would need to create the software of the program itself using machine learning. Hopefully, I can implement it with hardware to create tangible devices such as drones or monitor systems using sensors. If that is not possible, I can also just use simulation to test how it might function in real life; that can also be a good start to my project to make sure it works before actually applying and creating it in the real world.

For the research itself, I would like to focus on researching the necessary components of a managing system for a farm. I would look into what type of sensors and data collectors are required for the necessary inputs that will be used in processing. I will also look into agriculture itself and study what type of practices and factors would form the best farming conditions, and view those as targets for me to aim to reach in my programs. One other thing important to be studied is the embedded systems needed and the method to create connectivity, as different elements that affect agriculture closely work together. Overall, the objective of this project is to create a system that can help produce the best conditions for agriculture on a farm.

Research

One research paper that highlighted the use of such sensors in an agricultural environment is “A Review of Wireless Sensor Technologies and Applications in Agriculture and Food Industry: State of the Art and Current Trends” by Luis Ruiz-Garcia, Loredana Lunadei, Pilar Barreiro, and Jose Ignacio Robla. The paper discusses the wireless sensor technologies used in the Agri-Food center, in which tasks that monitor the soil, crops, and environment are completed with the help of Wireless Sensor Networks (WSN) that ensure connection between the devices. One example of an application made possible with WSN was a greenhouse that contained sensors that collect data such as photosynthetic radiation, electrical conductivity, leaf moisture level, etc. to regulate the amount of water and fertilizer used on the plants.

A research paper that takes it to another level and integrates these agriculture applications with machine learning is “Machine Learning for IoT-based Smart Farming” made by A. Ramesh Kumar, K.B. Archana, and P. Medhinya. The research presents a model that analyzes the variables essential to plant growth (such as pH level and temperature) evaluated using certain sensors and implements machine learning to predict the future conditions of the soil itself. This research is interesting as it shows how we can utilize the potential predictions of certain factors in the environment to make early preparations to optimize the best farming conditions. This displays how machine learning can be useful when implemented in IoT systems as such.

Focus

It is essential we tackle this by tracking the plant's health and growth. For this, we can use sensors that check for aspects of a plant’s appearance – such as color, structure, pattern, texture, etc. Sensors are needed in the technologies in this field to collect information about plants in order to create the best conditions for farming. One sensor that can be used is the color sensor, in which we can conclude a plant’s health by the color it is. Since the amount of chlorophyll in a plant affects its appearance in that sense, analyzing the color is utterly important. Another example is the photoelectric sensor, which can help read out a plant’s structure and shape, which is also highly impacted by the condition of its health. Additionally, if we want to take this project deeper and analyze plants’ health beyond just its visual appearance, we can use thermal imaging sensors to check plant’s temperature or turbidimetric biosensors to examine the turbidity of the water on the surface or inside the plants, which can be useful for identifying any unwanted bacteria.

Wearable plant sensors have already been developed that collect real-time data on that. One way how machine learning can be used to advance this process is to identify healthy and diseased plants using a collection of image datasets. This has already been done so by Dr. V. Devarajan and Dr. R. Gunasundari in their research of “Determining Plant Health Using Machine Learning,” in which they used 54,306 pictures from a public dataset to train the ML model to diagnose diseases in different species of plants.

Machine learning can be used to analyze a public dataset containing various unique images of different crops to generate a prediction. Using that prediction, we can then input information and images of the actual crop we want to diagnose. The program will then compare and contrast the images in the database to the ones of our crops to generate a full diagnosis of the plant’s health. This can help identify certain diseases or even just improper growth. ML programs, with the vast dataset they have, can also generate certain suggestions or advice, such as which fertilizers to use or how to prevent certain diseases from spreading. Not only that but ML can also be used to make assessments of how well a farm is doing, including checking the health conditions of the plants using the information datasets they store and train their models on. ML could potentially enhance certain evaluations and predictions combined with the data they receive from the sensors or other datasets to advance the accuracy of their models, which is helpful in optimizing maximum yield in real time.

Work Cited

• Abdussalam Elhanashi, et al. “Integration of Deep Learning into the IoT: A Survey of Techniques and Challenges for Real-World Applications.” Electronics, vol. 12, no. 24, Multidisciplinary Digital Publishing Institute, Dec. 2023, pp. 4925–25, https://doi.org/10.3390/electronics12244925. Accessed 8 Mar. 2024.

• Basit, Abdul. “The Dangers of Overfitting in Machine Learning.” Linkedin, 2023, www.linkedin.com/pulse/dangers-overfitting-machine-learning-abdul-basit/. Accessed 8 Mar. 2024.

• Blue River Technology. “Our Products - Welcome | Blue River Technology.” Bluerivertechnology.com, 2020, bluerivertechnology.com/our-products/. Accessed 8 Mar. 2024.

• Devarajan, V., and R. Gunasundari. Determining Plant Health Using Machine Learning, 2021. Accessed 17 Mar. 2024.

• Guliy, Olga I., et al. “Optical Sensors for Bacterial Detection.” Sensors, vol. 23, no. 23, Multidisciplinary Digital Publishing Institute, Nov. 2023, pp. 9391–91, https://doi.org/10.3390/s23239391. Accessed 15 Apr. 2024.

• Kumar, A. Ramesh. “Machine Learning for IoT-Based Smart Farming.” ResearchGate, Auricle Technologies Pvt., Ltd., Nov. 2023, www.researchgate.net/publication/375325008_Machine_learning_for_IoT-based_smart_farming. Accessed 17 Mar. 2024.

• Loubna Bouhachlaf, et al. “Monitoring Soil Elements for Irrigation Management Using Internet of Things (IoT) Sensors.” World Water Policy, vol. 9, no. 4, Wiley, Nov. 2023, pp. 756–66, https://doi.org/10.1002/wwp2.12151. Accessed 17 Mar. 2024.

• Mathur, Geetika. “Top 10+ IoT Research Topics for 2024 [with Source Code].” Knowledgehut.com, 22 December 2023, www.knowledgehut.com/blog/others/iot-research-topics. Accessed 8 Mar. 2024.

• Md Reazul Islam, et al. “Machine Learning Enabled IoT System for Soil Nutrients Monitoring and Crop Recommendation.” Journal of Agriculture and Food Research, vol. 14, Elsevier BV, Dec. 2023, pp. 100880–80, https://doi.org/10.1016/j.jafr.2023.100880. Accessed 17 Mar. 2024.

• “Overview of Photoelectric Sensors | OMRON Industrial Automation.” Omron.com, 2024, www.ia.omron.com/support/guide/43/introduction.html. Accessed 1 Apr. 2024.

• Prem Rajak, et al. “Internet of Things and Smart Sensors in Agriculture: Scopes and Challenges.” Journal of Agriculture and Food Research, vol. 14, Elsevier BV, Dec. 2023, pp. 100776–76, https://doi.org/10.1016/j.jafr.2023.100776. Accessed 17 Mar. 2024.

• “Region 2 - Home.” Usda.gov, 2024, www.fs.usda.gov/detail/r2/home/?cid=stelprdb5388915#:~:text=Chlorophyll,and%20blue%20and%20appears%20green. Accessed 1 Apr. 2024.

• Ruiz-Garcia, Luis, et al. “A Review of Wireless Sensor Technologies and Applications in Agriculture and Food Industry: State of the Art and Current Trends.” Sensors, vol. 9, no. 6, Multidisciplinary Digital Publishing Institute, June 2009, pp. 4728–50, https://doi.org/10.3390/s90604728. Accessed 17 Mar. 2024.

• SIMON IoT. “Sensors in Agriculture: What Are They All About? | Simon IoT.” SIMON IoT, 16 Aug. 2023, www.simoniot.com/sensors-in-agriculture/. Accessed 17 Mar. 2024.

• “Taranis - Crop & Farm Management Software | Precision Agriculture Technology.” Taranis, 4 Mar. 2024, www.taranis.com/. Accessed 8 Mar. 2024.

• “‘Wearable’ Sensors for Plants: World Economic Forum Emerging Technologies Report.” Frontiers Science News, 2023, www.frontiersin.org/news/2023/06/26/wearable-sensors-for-plants-world-economic-forum-emerging-technologies-report/. Accessed 17 Mar. 2024.

• “What Is a Color Sensor? | Sensor Basics: Introductory Guide to Sensors | KEYENCE.” Keyence.com, 2024, www.keyence.com/ss/products/sensor/sensorbasics/color/info/. Accessed 1 Apr. 2024.

• World Economic Forum. “Strategic Intelligence | World Economic Forum.” Stategic Intelligence, 2024, intelligence.weforum.org/topics/a1G680000008gwPEAQ#:~:text=The%20next%20frontier%20is%20wearable,%2C%20moisture%2C%20and%20nutrient%20levels. Accessed 17 Mar. 2024.

• Vishal Meshram, et al. “Machine Learning in Agriculture Domain: A State-of-Art Survey.” Artificial Intelligence in the Life Sciences, vol. 1, Elsevier BV, Dec. 2021, pp. 100010–10, https://doi.org/10.1016/j.ailsci.2021.100010. Accessed 8 Mar. 2024.