Smart AgriSense: AI‑Powered Hybrid System for Crop Health

Monitoring

P. Devika, S. Anushobini and V. Harini

Department of Computer Science and Engineering, Nandha Engineering College (Autonomous), Erode, Tamil Nadu, India

Keywords: Precision Agriculture, Crop Disease Detection, AI, PH‑EC Sensors, Convolutional Neural Networks (CNNs),

Fertilizer Optimization, Smart Farming, Sustainable Agriculture, Machine Learning, Bluetooth‑Enabled

Farming.

Abstract: Precision agriculture is essential for enhancing crop yields and sustainability while reducing resource waste.

This study presents Smart AgriSense, a hybrid system powered by AI that merges data from pH and electrical

conductivity (EC) sensors with Convolutional Neural Networks (CNNs) for the early detection of crop

diseases and improved fertilizer recommendations. Unlike traditional image-based AI models or expensive

IoT solutions, this method leverages affordable soil sensors to identify abnormalities before any visible

symptoms emerge, facilitating timely intervention. The system analyzes real-time data from sensors and plant

images to diagnose diseases and suggest accurate treatments, thereby improving decision-making for farmers.

A mobile app with Bluetooth functionality ensures easy access to data, even in rural regions with limited

internet access. By incorporating AI-driven insights, the proposed system minimizes pesticide usage,

optimizes fertilizer applications, and encourages sustainable farming practices, making cutting-edge

agricultural technology more accessible and economical for small-scale farmers.

1 INTRODUCTION

Agricultural sciences play an important role in the

global food crisis, as to have the basemant of good

cultivation, the soil must be preserved by timely

intervention from disease. Soil is one of the most

important factors influencing the growth of plant,

nutrient uptake and soil fertility, such as soil

moisture, pH and the electric conductivity (EC).

Differences in these metrics may lead to reduced

efficiency, soil degradation and increased

susceptibility to plant diseases. Moreover, climate

change has enhanced the incidence and spread of

plant diseases, which are caused by bacteria, fungi

and viruses and pose a serious threat to food security

and agricultural productivity. (F. F. Hossain et al.,

2022) Those impacting factors calls for the need of

integrating technologies like Internet of Things and

artificial intelligence into the methods of farming. We

propose an IoT based precision agriculture system,

which includes real-time soil condition monitoring

along with AI powered plant disease detection. Using

LoRaWAN supported wireless sensors, we monitor

soil moisture, pH and EC in order to provide

continuous data to improve irrigation efficiency and

optimize the management of nutrients. (B. Althaph et

al.,2024). So, this technique saves water at the same

time, aiding in prevention of soil salinization and

preserving soil health. Additionally, we use a

Machine Learning based Crop Disease Detection

system which processes plant images to detect

diseases at an earlier stage using Convolutional

Neural Networks (CNNs) based on transfer learning.

Timely identification of diseases allows farmers to

initiate preventative measures that prevent crop

damage and enhances overall yield. The outlined

work in the paper gives the novelty of how IoT,

WSNS, and AI-based disease detection can be

integrated into our proposed system to introduce

sustainable agricultural development. LoRaWAN

technology is long-range low-power data

communication technology which makes it suitable

for large scale farming. Farmers leverage the real-

time data collected to make informed decisions,

automate the irrigation process and take quick disease

control measures. By doing so, it enables resources to

be used judiciously, ensures sustainable agricultural

practices, and equips the farmers with smart tools to

Devika, P., Anushobini, S. and Harini, V.

Smart AgriSense: AI-Powered Hybrid System for Crop Health Monitoring.

DOI: 10.5220/0013882000004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 2, pages

309-315

ISBN: 978-989-758-777-1

309

improve productivity and grow food securely. Our

project seeks to innovate tomorrow's agriculture and

pave the way for a future where farming, both data-

driven and precise, effective as it is environment-

friendly, is the norm (M. K. Roy et al., 2023).

2 RELATED WORKS

2.1 Progress in AI and IoT for

Accurate Identification of Crop

Diseases

On the application of Artificial Intelligence &

Internet of Things in detection of crops diseases; A

state of the art survey, Houda Orchi, Mohamed Sadik,

Mohammed Khaldoun It assesses the efficiency and

technical challenges in utilizing machine learning,

deep learning, and hyperspectral imaging

approaches. The study highlights the significance of

using automated detection methods to enhance

agricultural productivity and reduce losses. It also

discusses recent constraints and proposes potential

future study directions to improve the Precision,

efficiency, and sustainability of disease

identification, providing help to both farmers and the

agricultural market (Houda Orchi et al.,2021).

2.2 Improving Smart Farming through

IoT and Wireless Sensor Networks

Internet of Things and Wireless Sensor Networks for

Smart Agriculture Applications by Md. Najmul

Mowla, Neazmul Mowla, A. F. M. Shahen Shah,

Khaled M. Rabie, and Thokozani Shongwe examines

how IoT and Wireless Sensor Networks (WSNs) can

improve smart agriculture. It covers various uses,

including automated irrigation systems, soil moisture

measurement, optimization of fertilizer usage, pest

and disease control, and energy savings. The research

also evaluates different wireless communication

protocols like ZigBee, WiFi, SigFox, and LoRaWAN

for the real-time collection of agricultural data.

Furthermore, it points out the challenges associated

with the integration of IoT and WSN, such as issues

with scalability, security, and energy consumption,

while suggesting future pathways to enhance the

efficiency and sustainability of contemporary

farming practices. (Mowla et al., 2023).

2.3 UAS-Supported IoT and

LoRaWAN for Accurate Soil

Moisture Monitoring

Soil Moisture Monitoring through UAS-Assisted

Internet of Things LoRaWAN Wireless Underground

Sensors by Fahim Ferdous Hossain, Russ Messenger,

G. Levi Captain, Sabit Ekin, Jamey D. Jacob, Saleh

Taghvaeian, and John F. O’Hara presents an

innovative method for monitoring soil moisture

through the use of buried IoT sensors that

communicate via LoRaWAN technology. A gateway

mounted on a UAS gathers information from these

underground sensors, negating the necessity for on-

site base stations and reducing disruptions to

agricultural operations. Field trials validate the

system’s efficiency regarding communication range,

energy use, and scalability. The research emphasizes

its promise for precision agriculture, facilitating

improved water conservation and more sustainable

farming methods (S. S. Chakole et al., 2022).

2.4 FarmGuide Using AI Techniques

Crop Disease Prediction by Using Machine Learning

by G. Kavya Siri, B. Madhavi, A. Bhavani, A.

Lakshmi Sowjanya and A.V.S. Sudhakar Rao offers

an A.I.-based agricultural decision- support system

designed to help farmers improve crop selection,

optimize fertilizer use and diagnose plant diseases.

The developed system is a web application that

encompasses million-dollar ideas including machine

learning and deep learning approaches like XGBoost

and Random Forest, as well as Convolutional Neural

Networks. (B. Althaph et al.,2024) Crop

recommendation tool provided suitable crops based

on soil quality information; Fertilizer

recommendation tool forecast nutrient deficiency and

recommend solutions, Disease detection module

detected the disease after analyzing leaves the

initiative employs AI with the objective of improving

the efficiency, sustainability, and productivity of

farmers, allowing for better decision-making for

them. (Nanda et al.,2023).

2.5 Crop Disease Identification

Utilizing Deep Learning with CNN

and Mobile AI

Machine Learning Technique for Crop Disease

Prediction Through Crop Leaf Image by S. Nandhini

and K. Ashokkumar investigates the application of

deep learning for the early detection of crop diseases.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

310

The researchers employed a Convolutional Neural

Network (CNN) trained on 64,412 images sourced

from the Plant Village dataset, which encompasses 16

crop species and 25 different diseases. The developed

model achieved an impressive accuracy of 99.35%,

demonstrating its efficacy in plant disease

classification. The paper emphasizes the potential of

using smartphones for disease detection, utilizing

high-resolution cameras and AI for instantaneous

analysis. (S. Nandhini and K. Ashokkumar 2022).

3 METHODOLOGY

The Smart AgriSense system combines affordable pH

and electrical conductivity (EC) sensors with AI-

enhanced image analysis to identify crop diseases and

suggest optimized fertilizer use. The method involves

three primary steps: monitoring soil health using

sensors, detecting diseases with CNN, and providing

AI-based fertilizer recommendations. Figure 1 shows

the Phases of System Development and Schedule for

Smart AgriSense.

Figure 1: Phases of System Development and Schedule for

Smart AgriSense.

3.1 Monitoring Soil Health with

Sensors

Farmers place pH and EC sensors into the soil close

to plant roots to obtain real-time data on soil

conditions. Healthy plants exhibit consistent pH and

EC levels, while those affected by disease display

irregularities due to altered nutrient absorption. The

correlation between electrical conductivity (EC) and

ion concentration (C) in soil can be expressed as:

𝐸𝐶 = 𝑘𝐶 (1)

where k is the constant of proportionality. Any

variations in EC and pH beyond set limits indicate a

possible disease occurrence. The data is wirelessly

transmitted via Bluetooth to a mobile application,

ensuring farmers in remote areas can access it.

3.2 Disease Detection Using CNN

When abnormal soil conditions are identified, the

farmer captures an image of the plant using a

smartphone. The Convolutional Neural Network

(CNN) analyzes the image and identifies important

features necessary for classifying the disease. The

CNN model employs convolutional filtering and

pooling techniques to detect patterns, mathematically

represented as follows:

𝑓(𝑥) = ∑ (𝑤𝑖 ∗ 𝑥𝑖) + 𝑏 (2)

where wi signifies the weights in the CNN, xi

represents the pixel values, and b is the bias term. The

CNN then compares the extracted features against a

pre-trained dataset of diseases and provides a

classification along with confidence scores.

3.3 Fertilizer Recommendations

After determining the type of disease, the AI model

computes the ideal NPK (Nitrogen, Phosphorus,

Potassium) ratio based on current soil readings. The

suggested fertilizer dosage (F) is calculated using:

𝐹 = (𝑁𝑑 − 𝑁𝑐) + (𝑃𝑑 − 𝑃𝑐) + (𝐾𝑑 − 𝐾𝑐) / 3 (3)

where:

{Nd, Pd, Kd}: Required nutrient levels for target

growth | {Nc, Pc, Kc}: Current nutrient levels in

soil.

Providing advice on organic or chemical

fertilizers, it helps you tailor a detailed guideline and

track application through a mobile app, to be exact.

This integrated AI-based method enables early

detection of diseases, optimizes the use of resources,

lowers expenses, and makes precision agriculture

practicable for small- scale farmer.

Smart AgriSense: AI-Powered Hybrid System for Crop Health Monitoring

311

4 PROPOSED SYSTEM

Smart AgriSense, an innovative and AI-based

agriculture solution is a low-cost pH and electrical

conductivity (EC) sensor integrated with a

Convolutional Neural Network (CNN) for image

processing. Thus, enabling early detection of soil

irregularities as well as crop diseases to differentiate it

from the traditional soil monitoring, where it’s are

solely relied on manual inspections. By tracking real-

time fluctuations in soil conditions, it can predict

problems before they become visible. It begins with

the sensors that take measurements of the soil pH and

EC levels to detect any imbalances that can affect

plant growth. The system takes and processes images

of the plants using a deep learning CNN model if the

irregularities are detected. This model has been trained

on a diverse dataset and is able to analyze plant images

against known disease patterns to accurately detect

potential infections or nutrient deficiencies. After a

problem is discovered, the system recommends the

most relevant and useful NPK (Nitrogen, Phosphorous

and Potassium) fertilizer ratio, used in accordance to

the soil’s nutrient profile. Smart AgriSense provides

customized recommendations based on real-time soil

data, as opposed to the old-school method of applying

fertilizers without knowing soil conditions, thus

ensuring nutrient rush management and reduced

environmental impact. Smart AgriSense can work

without an internet connection, which is a most

important feature for farmers who works in remote

areas. This allows for seamless functioning as

Bluetooth is used to send data, so no requirement of

Wi-Fi. Moreover, a user-friendly mobile-based

application facilitates timely alerts, pictorial reports

and actionable suggestions in an easily readable

format that can be comprehended quickly by farmers.

SmartAgriSense uses the power of AI and IoT to

promote better crop health, prevent fertiliser wastage

and increase productivity in agriculture. It is cost-

effective, adaptable, and environmentally friendly,

making it an incredible resource for small-scale

farmers that help promote sustainable farming

practices while enabling higher yields with

significantly lower environmental damage.

5 EXPERIMENTAL RESULT

Table 1 represents the crop disease detection. Under

actual agricultural conditions, we performed the

evaluation of a crop disease detection system based

on ph and EC along with an artificial intelligence (AI)

fertilizer recommendation framework. The key aim

was to determine if low-cost pH and EC sensors are

successful in detecting early signs of plant diseases,

along with suggesting precise fertilizers based on

real-time soil health status. The collected images

capture two separate states of the plant, including one

with disease symptoms and one displaying healthy

crops.

Table 1: Performance Comparison of Crop Disease

Detection.

Parameter

Healthy

Crop

(Millet)

Diseased Crop

(Peanut)

Observations & AI

Recommendations

Soil pH

6.5 - 7.0

(Optimal)

5.2 - 5.8

(Acidic)

Unhealthy soil

linked to disease;

AI suggests lime

treatment.

Electrical

Conductivit

y (EC)

0.3 - 0.5

dS/m

(Normal)

1.2 - 1.8 dS/m

(High)

High EC indicates

salt

accumulation; AI

recommends

proper irrigation.

Leaf

Condition

Green,

healthy

leaves

Black spots,

yellowing

AI classifies

fungal/bacterial

infection.

Growth

Rate

Normal

growth

pattern

Stunted growth

Imbalance in

nutrients

affect growth.

Fertilizer

Recommen

-dation

Maintain

current

levels

Increase

nitrogen, add

organic matter

AI suggests

tailored

NPK balance for

recovery.

Yield

Impact

(Predicted)

High Yield

Potential

30-40%

Reduction (if

untreated)

AI-based early

intervention can

prevent yield loss.

5.1 Soil Health Monitoring and

Analysis

Soil Samples were collected from different sites

during the testing period and PH and EC readings

were recorded. The peanut plant (top photo) showed

obvious symptoms of being diseased dark spots on its

leaves and stunted growth indicative of either a

fungal or bacterial ailment. Soil properties in the

vicinity of these plants displayed deviation from the

pH range, and substantial deviation of EC value was

noted, indicating an imbalance in the uptake of

nutrients.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

312

In contrast, the millet growth (second image) was

strong and thriving, indicating that its soil parameters

were likely in the optimum range. The differences in

the soil conditions of healthy and diseased plants

support the hypothesis that variations in these

variables, particularly pH and EC, can serve as an

early warning of diseases and nutrient deficiencies in

plants. Figure 2 shows below:

Figure 2: Sample Detection of Healthy Millet Crop.

5.2 AI-Based Disease Detection

A CNN based model was established with crops

healthy and diseased images. The experimental

validation results were found that the AI was able to

detect the disease in peanut plants and classify them

into categories like bacterial leaf spot, fungal

infections, or nutrient deficiency. As real-world data

was collected from sensors and images in fine-tuning

the model, its accuracy in detecting diseases

improved significantly. In addition, to provide the

model flexibility to adapt to newly emerging disease

variants, continuous learning techniques were

implemented. The system also provides real-time

notifications and recommendations to farmers based

on this data, which enables timely action and loss

mitigation. Figure 3 shows the peanut crop.

Figure 3: Sample Detection of Diseased Peanut Crop.

5.3 AI-Driven Fertilizer

Recommendation

Once it diagnosed the diseases and nutrient

deficiencies in the soil, the AI system made

recommendations for correcting them, including the

best way to apply fertilizer. The AI model predicted

the required NPK (Nitrogen, Phosphorus, Potassium)

ratio according to the parameters of the soil so that

farmers use only the required amount of fertilizer,

preventing any excess use that would cause soil

degradation.

In the field of millet where the plants were

healthy, the AI model confirmed that nutrient levels

were sufficient, affirming the value of precision

monitoring. But for sick peanut plants, the AI

suggested using organic nitrogen-potassium-based

fertilizers and the requisite fungicides to limit the

spread of infection. Figure 4 shows the crop roots.

Figure 4: Sample Detection of Crop Roots.

5.4 Final Observation and Impact

Figure 5: Analysis of Crop Health and Environmental

Trends Utilizing AI Technology.

Smart AgriSense: AI-Powered Hybrid System for Crop Health Monitoring

313

Detection of Plant Diseases: It proved to be highly

competent in finding plant diseases before visible

symptoms occurred, so that farmers could take

preventive measures. Applying the Right Amount of

Fertilizer: AI recommendations helped avoid over-

fertilization, reducing costs for farmers. Low Cost &

Scalability: Due to the low costing of the pH and EC

sensors (around ₹500-₹1000 each) the solution could

be viable for the smallholder farmers. Improved

Crop Health & Yield: Early disease detection and

targeted treatment of the soil have led to overall

productivity of their crops. The results of the

experiment confirm the effectiveness of this

innovative approach, establishing it as a cost-

efficient, scalable, and pragmatic solution for

contemporary precision agriculture.

Figure 5 shows

the crop health and environment.

6 CONCLUSIONS

This project successfully displays how AI- powered

system can help in detecting crop diseases and

suggest for fertilizer in modern agriculture. Using

machine learning and image analysis techniques, this

system effectively identifies diseases at their early

onset stages to allow farmers to take necessary

actions immediately. The experimental results show

that all unhealthy plants present unique symptoms

like leaf discoloration, fungal spots, and reduced

growth which the AI Model was applied and

classified with great accuracy. Additionally, the

fertilizer recommendation system, powered by

artificial intelligence, provides personalized

recommendations for improving soil nutrients,

leading to healthier and more productive crops.

Gentilcore says that the only way forward in

agriculture relies on the acceptance of technology to

improve efficiency and sustainability. This system

minimizes reliance on chemical pesticides and

promotes environmental-friendly agricultural

approaches by resolving issues such as the spread of

disease and depletion of nutrients in the soil. Future

work should be directed towards a broader dataset

which may provide better precision, and real-time IoT

based monitoring along with continuous feedback. In

summary, it serves as a prerequisite for smart; data-

driven farming solutions that will assist farmers in

achieving GFS.

REFERENCES

“Machine Learning Technique for Crop Disease Prediction

Through Crop Leaf Image” by S. Nandhini and K.

Ashokkumar in Naturals Publishing in January 2022.

2024.10499178.

B. Gupta, S. Bomble, O. Gaikar, S. Chalekar, S. R. Vispute

and K. Rajeswari, "Convolutional Neural Networks for

Detection of Crop Diseases and Weed," 2022 6th

International Conference on Computing, Communicati

on, Control and Automation (ICCUBEA, Pune, India,

2022, pp. 1- 5, doi: 10.1109/ICCUBEA54992.2022.1

0010772.

B. Mondal, M. Bhushan, I. Dawar, M. Rana, A. Negi and

S. Layek, Crop Disease Prediction using Machine

Learning and Deep Learning: An Exploratory Study,

2023 International Conference on Sustainable

Computing and Smart Systems (ICSCSS), Coimbatore,

India, 2023, pp. 278-283, doi:

10.1109/ICSCSS57650.2023.10169612.

B. Althaph, N. P. Challa, B. Naseeba and N. K. Rao, Crop

Disease Analysis and Detection using GoogleNet

Model, 2024 11th International Conference on

Computing for Sustainable Global Development

(INDIACom), New Delhi, India,2024, pp. 592-596,

doi:10.23919/

F. F. Hossain, R. Messenger, G. L. Captain, S. Ekin, J. D.

Jacob, S. Tag- varian, J. F. O’Hara, Soil moisture

monitoring through uas-assisted internet of things

lorawan wireless underground sensors, IEEE Access

10(2022)

Houda Orchi & Mohamed Sadik & Mohammed Khaldoun,

2021."On Using Artificial Intelligence and the Internet

of Things for Crop Disease Detection: A Contemporary

Survey," Agriculture, MDPI.

K. V. Paresh, N. Srinivasan, P. Gautham Raj, J. Varma and

M. Kanthi, "An Efficient, Low-cost Plant Disease

Detection System using IoT," 2022 International

Conference on Futuristic Technologies (INCOFT),

Belgaum, India, 2022, pp. 1- 6, doi: 10.1109/INCOFT

55651.2022.10094361.

M. K. Roy, P. Manna, R. Roy, P. Roy, S. Rakshit and M.

K. H. Mondal, "Disease Detection in Paddy Crop

using Machine Learning Techniques," 2023

International Conference on Sustainable Computing

and Data Communication Systems (ICSCDS), Erode,

India, 2023, pp. 287- 292, doi: 10.1109/ICSCDS56580

.2023.10104908.

Mowla, M. N., Mowla, N., Shah, A. F. M. S., Rabie, K. M.,

& Shongwe, T. (2023). Internet of Things and Wireless

Sensor Networks for Smart Agriculture Applications: A

Survey. IEEE Access, 11, 145813-145852.

Nanda, S. Nayak, A. K. Patra, A. Nanda, S. Pani and B. S.

Nanda, "Crop Disease Prediction by Using Machine

Learning," 2023 International Conference in Advances

in Power, Signal, and Information Technology

(APSIT), Bhubaneswar, India, 2023, pp. 706-711, doi:

10.1109/APSIT 58554.2023.10 201690.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

314

R. M. Prakash, M. Vimala, K. Ramalakshmi, M. B.

Prakash, A. Krishnamoorthi and R. S. S. Kumari, "Crop

Disease Detection and Classification with Transfer

learning and hyper-parameters optimized

Convolutional neural network," 2022 Third

International Conference on Intelligent Computing

Instrumentation and Control Technologies (ICICICT),

Kannur, India, 2022, pp.1608-1613, doi:

10.1109/ICICICT54557.2022.9917901.

S. S. Chakole, S. Khera and N. A. Ukani, Analysis of Plant

Leaf Disease Based on Sensor and Machine Learning

Technique, 2022 IEEE International Conference on

Current Development in Engineering and Technology

(CCET), Bhopal, India, 2022, pp. 1-4, doi:

10.1109/CCET56606.2022.10080352.

S. Kashyap, T. Thaware, S. R. Sahu and M. R. K, "Multi-

Crop Leaf Disease Detection using Deep Learning

Methods," 2022 IEEE 19th India Council International

Conference (INDICON), Kochi, India, 2022, pp. 1-6,

doi: 10.1109/INDICON56171.2022.10040099.

S. Yadav, M. Rajendaran, A. Nagpal, A. Karthik, K.

Aravinda and B. T. Geetha, "Exploring the Potential of

Machine Learning in Enhancing Agricultural Practices

and Food Production," 2024 IEEE 13th International

Conference on Communication Systems and Network

Technologies (CSNT), Jabalpur, India, 2024, pp. 663-

669.

Singh and A. K. Bhamboo, Crop Recommendation and

Disease Detection Using Deep Neural Networks, 2022

IEEE 6th Conference on Information and Communic-

ation Technology (CICT), Gwalior, India, 2022, pp. 1-

5, doi: 10.1109/CICT56698.2022.9997839.

Systematic study on deep learning-based plant disease

detection or classification" by C. Karan. Sunil,C. D.

Jaidhar and Nagamma Patil published in Mili

Publications in June 2023.

Smart AgriSense: AI-Powered Hybrid System for Crop Health Monitoring

315