Good [morning/afternoon], everyone.

My name is Markanday Persaud, and I am here to

present my proposal for a Smart Pest Detection System aimed at controlling Gandhi

bug infestations in Guyana’s rice fields. As you may know, rice cultivation is a crucial

part of Guyana’s economy, supporting food security and generating nearly $200 million

annually. Guyana is one of the Caribbean’s largest rice producers, exporting over

500,000 tonnes each year. However, the industry faces a persistent threat from the

Gandhi bug, which damages rice during critical growth stages, reducing yield and

quality which according to the Guyana Rice Development Board (GRDB), losses from

paddy bug infestations in 2024 are estimated at around $4 billion. Traditional methods

for managing this pest, such as manual inspections and broad pesticide use, are costly,

reactive, and often insufficient. Moreover, the excessive use of pesticides leads to

environmental degradation and pest resistance, highlighting the need for a more

effective solution.

To address this challenge, I am proposing a Smart Pest Detection System that

combines machine learning—specifically, Convolutional Neural Networks (CNNs)—with

thermal imaging for real-time pest monitoring and early intervention. This approach

allows for more targeted pesticide use, which lowers costs and reduces environmental

impact, aligning well with the sustainable farming goals promoted by the Guyana Rice

Development Board.

Now, let’s take a look at the problem statement. Currently, the lack of an efficient

detection system means farmers struggle to detect Gandhi bug infestations in time,

leading to delayed interventions and significant crop losses. This project aims to
develop a system that can accurately monitor and control Gandhi bug infestations,

enabling early detection and reducing these losses.

To accomplish this, I will start by gathering and analyzing data from rice fields across

Guyana, focusing on the occurrence, distribution, and impact of Gandhi bug

infestations. Understanding the scale and frequency of these infestations will be

essential for designing an effective detection system. Next, I will develop a CNN-based

thermal mapping system that uses data collected from thermal cameras to identify and

monitor Gandhi bug activity in real-time.

The CNN-based system will be integrated into a live monitoring setup, where it can

analyze video feeds or images from thermal cameras and display real-time maps and

alerts for farmers. Finally, I will conduct field trials to compare the performance of this

system against traditional pest management methods, assessing its impact on pesticide

usage, crop yield, and quality.

Let’s take a closer look at how this system works. Thermal imaging cameras detect

subtle heat differences in the field, identifying pests like the Gandhi bug by their warmer

heat signatures compared to the surrounding plants.

Literature Review
Slide 1

Convolutional Neural Networks (CNNs) are a type of deep learning model particularly suited to

image recognition and classification tasks. CNNs have gained significant traction in agriculture,

especially for pest detection, due to their ability to accurately recognize objects within complex

image data by learning distinguishing features, including patterns, shapes, and textures. Here’s

a deeper look at how CNNs work and why they are effective in agricultural applications like pest

detection:

Key Components of CNNs

1. Convolutional Layers:

○ These layers apply a series of filters (or kernels) to the input image, which helps

detect specific features, such as edges, colors, and textures, by scanning the

image and recognizing patterns.

○ Each filter learns a different feature, and as the CNN goes deeper, it captures

increasingly complex aspects, enabling the network to differentiate between

objects, such as pests and background noise.

2. Pooling Layers:

○ Pooling layers reduce the spatial dimensions of the data, consolidating

information while retaining the most critical features.

○ By reducing dimensionality, pooling helps the CNN generalize better and reduces

computational demands, making it efficient even for high-resolution agricultural

imagery.

3. Fully Connected Layers:

features to classify the image or make predictions.

○ In agricultural pest detection, these layers can help identify the type of pest

based on the features extracted from the convolutional layers.

4. Activation Functions:

○ Activation functions introduce non-linearity, allowing the CNN to handle complex

patterns that cannot be represented by linear transformations.

○ ReLU (Rectified Linear Unit) is a common activation function, enhancing the

network’s ability to capture intricate patterns found in pest imagery.

How CNNs Work for Pest Detection

CNNs excel at feature extraction, meaning they can learn to recognize characteristics unique to

specific pests, such as shape, color, or movement patterns.For pests like the Gandhi bug, CNNs

can differentiate the bug's appearance from other objects in the field (like rice grains) by

learning its distinct features. CNNs process data hierarchically, learning simple features first and

building up to complex representations. This approach improves detection accuracy, even when

images contain environmental noise or varying light conditions.CNNs can be integrated into

real-time monitoring systems, providing farmers with immediate feedback on pest activity, which

is crucial for timely intervention in agricultural settings.With sufficient training data, CNNs can be

adapted to detect various pests or anomalies in other crops, making them a versatile tool in

agriculture.

Slide 2
Early Detection & Targeted Intervention: Identifies Gandhi bugs early, allowing timely,

location-specific pest control to reduce crop loss.

Cost Efficiency: Cuts pesticide and labor costs by focusing treatments on hotspots only,

leading to lower production expenses.

Enhanced Crop Quality & Yield: Maintains crop health and quality, improving both yield and

market value.

Environmental Sustainability: Reduces pesticide use, lowering soil and water pollution while

protecting beneficial insects.

Reduced Labor Dependency: Minimizes need for manual inspections and enables remote

monitoring, saving time and effort.

Real-Time Monitoring & Data Collection: Continuous surveillance with real-time alerts,

generating valuable pest pattern data for predictive insights.

Scalability & Flexibility: Adaptable to various field sizes, crop types, and conditions, with

potential to detect other pests.

Long-Term Economic Benefits: Reduces crop loss risk, supporting economic stability and

security for farming communities.

Slide 3
Slide: Real-World Example of CNN in Agriculture

Example: Use of CNN for Pest Detection in Rice Farming (India)

● Scenario: Rice farming in India faces significant challenges from pests such as the rice

stem borer and brown plant hopper. These pests can destroy up to 30% of a rice crop

if not managed effectively.

● Solution: Researchers and farmers in India integrated CNNs into pest management

systems, using drones and cameras to capture high-resolution images of rice fields.

These images were then analyzed using CNNs to detect the presence of pests.

● How CNN Works:

○ CNNs were trained on thousands of images of rice fields, both with and without

pest infestations, to learn the unique features and patterns associated with pest

activity.

○ The system automatically detects pests and flags areas of the field that require

attention.

● Results:

○ Early Detection: The CNN system was able to identify pest activity days or even

weeks earlier than traditional methods.

○ Increased Yield: Early intervention prevented significant crop damage, leading

to a 20-25% increase in yield compared to traditional pest control methods.

○ Reduction in Pesticide Use: The system enabled targeted pesticide application,

reducing pesticide usage by 30-40% and minimizing environmental impact.

● Impact:

○ Cost Savings: Farmers saved on both pesticide and labor costs.

biodiversity, benefiting the ecosystem.

● Why It’s Effective: The CNN system, being accurate and automated, provides farmers

with the ability to detect and manage pest outbreaks efficiently, without relying on broad

pesticide application or manual field scouting.