A Camera-Based AI Fruit Picker with Leaf Health Monitoring to Reduce Post-Harvest
Losses in Small-Scale Farming
I. Introduction
Fruit spoilage and declining leaf health remain pressing challenges in agricultural
production, often leading to reduced yield quality and significant economic losses for
farmers (FAO, 2021; Kumar et al., 2020). Globally, post-harvest losses are estimated to
account for 30–50% of fruit production, with small-scale farmers in developing countries
being the most affected (Parfitt et al., 2010). In tropical regions, where environmental
conditions such as high temperature and humidity accelerate deterioration, the problem
becomes even more severe. For many farmers, improper handling during pre- and post-
harvest stages—such as premature or delayed picking, poor storage conditions, and
inadequate transportation—results in fruits becoming overripe, bruised, or spoiled before
they reach consumers (Bekele, 2018).
Climacteric fruits such as mangoes, bananas, and apples are particularly vulnerable because
they continue to ripen after harvest and have a short time frame between peak ripeness and
over-ripeness (Wills et al., 2016). Temperature fluctuations, mechanical damage, and
microbial infections can further accelerate spoilage. Alongside fruit quality, another critical
yet sometimes overlooked factor is leaf health. Leaves are responsible for photosynthesis,
nutrient transport, and overall plant vitality, and their deterioration—caused by pests,
nutrient deficiencies, or diseases—can drastically reduce fruit yield and quality (Singh &
Sharma, 2018). Early detection of unhealthy leaves can help prevent disease spread and
optimize crop management.
Traditionally, farmers rely on manual observation—assessing ripeness by fruit color, aroma,
and firmness, and checking leaves for visible damage. However, these methods are
subjective, time-consuming, and prone to errors, especially in larger orchards or when
multiple fruits ripen at different rates. As a result, farmers may harvest too early or too late,
leading to reduced quality and increased waste. Furthermore, detecting early signs of leaf
disease without specialized tools is difficult, often allowing problems to progress unnoticed
until significant damage has occurred.
Advances in artificial intelligence (AI), computer vision, and sensor technology now offer a
promising solution to these challenges. Camera-based AI systems can be trained to analyze
visual cues such as color gradients, texture patterns, and shape changes to determine fruit
ripeness with high accuracy (Koirala et al., 2019). These same systems can also detect leaf
discoloration, deformation, or spotting—symptoms often linked to nutrient deficiencies or
diseases (Picon et al., 2019). Integrating AI-driven ripeness detection with leaf health
monitoring into a single portable fruit picker device could revolutionize small-scale farming
practices.
By combining mechanical harvesting with intelligent sensing, such a device could reduce
human error, save time, and minimize post-harvest losses. It would enable farmers to
harvest fruits at their optimal ripeness, ensure leaf health is continuously monitored, and
improve overall productivity without requiring costly laboratory testing or extensive manual
labor. This research seeks to develop such a device, aiming to provide small-scale farmers
with an affordable, user-friendly, and effective tool for maximizing both yield quality and
quantity while supporting sustainable agricultural practices.
By adding this AI camera technology to a portable fruit picker, farmers can quickly find and
harvest fruits that are at their perfect ripeness. At the same time, the device can check the
condition of the leaves, helping farmers identify early signs of diseases or damage. This is
important because preventing leaf problems early can improve the overall health of the
plant and the quality of its fruits (Picon et al., 2019). A device like this can help reduce waste,
improve fruit quality, and increase income for small-scale farmers while also protecting the
health of their plants.
Statement of the Problem
This study aims to develop a fruit picker device that automatically detects the ripeness of
climacteric fruits using a camera-based AI system, while also monitoring leaf health. The
goal is to help small-scale farmers and home gardeners reduce food waste and improve
harvesting efficiency by enabling timely and safe picking of ripe fruits.
Specifically, the study seeks to answer the following questions:
a. How accurate is the camera-based AI system in detecting ripeness in various
climacteric fruits such as mango, banana, and apple?
b. Can the AI model distinguish between healthy and unhealthy leaves?
c. Will the device be able to safely pick ripe fruits without damaging them?
d. Is the fruit picker device reliable and practical for small-scale farm or backyard
use?
e. What are the advantages and limitations of using this fruit picker device
compared to manual picking?
Objective
This study aims to design a fruit picker device that uses a camera-based Artificial
Intelligence (AI) sensor. This sensor works by analyzing images of fruits and leaves to
determine if the fruit is ripe and if the plant is healthy. When the AI detects that a fruit is
ripe based on its color, texture, and shape, the device will automatically pick the fruit. The AI
can also identify signs of leaf disease or stress. This system can help small-scale farmers or
backyard growers save time, reduce fruit spoilage, and improve harvest quality.
Alternative Hypothesis
The AI fruit picker with leaf health monitoring significantly enhances harvesting accuracy,
reduces fruit damage, and supports better fruit quality compared to manual harvesting.
Null Hypothesis
The camera-based AI fruit ripeness detection and picking device does not significantly
improve the accuracy and efficiency of detecting fruit ripeness compared to manual
inspection methods.
Expected Outcome
The researchers expect the fruit picker device to accurately detect when fruits are ripe and
assess whether plant leaves are healthy using the integrated AI camera. The device should
be able to pick fruits without causing physical damage, work effectively in a small-scale
environment (like backyard farms), and demonstrate improved results over manual picking
in terms of speed and precision. Overall, the device is expected to contribute to reducing
fruit spoilage and increasing farmer productivity.
Engineering Goals
This study aims to develop an innovative fruit picker equipped with a camera-based AI
sensor capable of identifying ripe fruits and assessing the health of plant leaves. To create a
low-cost, efficient, and user-friendly device that supports sustainable agriculture by
improving harvest timing, reducing labor, and minimizing post-harvest losses. The
engineering process involves designing a reliable picking mechanism that gently harvests
ripe fruits, ensuring minimal damage, while integrating a compact AI system trained to
recognize fruit ripeness based on visual characteristics such as color, size, and texture.
Additionally, the system will include a leaf health monitoring feature that uses image
processing to detect early signs of diseases or stress.
II. Methodology
This study will use an experimental research method to test the effectiveness of the fruit-
picking device equipped with a camera-based ripeness detection system and an automatic
cutting mechanism. The device will use a camera-based sensor with AI to detect if a fruit is
ripe and if a leaf is healthy. The system is designed to help reduce fruit waste by making
sure only ripe and healthy fruits are picked.
The AI camera will be programmed using a trained model like YOLOv4, which has already
been proven to detect plant diseases and fruit ripeness accurately. The trained model will
help the device decide:
• If the fruit is ready to harvest
• If the leaf nearby is healthy or showing signs of disease
a. Materials
MATERIAL QUANTITY
1-1/4” OD x .049” Wall Round Aluminum 1
Tubing
1-1/8” OD x .038” Wall Round Aluminum 1
Tubing
1” OD x .038” Wall Round Aluminum Tubing 1
B Split Collar Lock 1-1/8” OD to 1-1/4” OD 1
B Split Collar Lock 1” OD to 1-1/8” OD 1
249x114mm Transparent Acrylic Dome 2
Cover
208x94mm Transparent Acrylic Dome 2
Cover
Raspberry Pi 4 (4GB) 1
Raspberry Pi Camera Module V3 (High 1
Quality)
SparkFun 16x2 SerLCD RGB Backlight 1
(Qwiic)
Arduino Mega 2560 Rev 3 2
High Speed Metal Gear Coreless Waterproof 3
Digital Servo DS3225 25KG Servo
Linkage Rods 2
25mm Mini Circular Saw Blade 1
37mm 12V DC Gear Motor 1
IRLZ44 MOSFET 1
1N4007 Diode 1
MOSFET Heatsink 1
Jungla 3S2P 12V 12800mAh battery 18650 1
Li-ion with BMS
DC-DC step-up/step-down converter 1
Push Button 1
100k Ohm Resistor 1
10k Ohm Resistor 2
Kingbright WP7113SGD Green LED 1
Kingbright WP7113SGD Red LED 1
b. Planning
researchers will begin by creating a detailed sketch of the device, indicating the placement
of the camera, servo motors, cutting mechanism, and display screen. After completing the
design, they will research and select materials that are durable, affordable, and compatible
with the device components. These materials will be bought from both local and online
shops to ensure quality and cost efficiency. And after acquiring all materials, the
researchers will prepare to assemble the device following the schematic as a guide.
c. Construction of the Device
Following the design sketch, the researchers will construct the aluminum tubing frame and
securely mount the transparent acrylic domes. The camera module will be installed inside
one dome to capture images of the fruit, while the display screen will be mounted externally
for easy visibility. The Arduino Mega 2560 boards will be connected to the camera for
image acquisition and processing, and to the display for showing the ripeness status
(“Unripe,” “Ripe,” or “Overripe”). Servo motors will be installed at pivot points to allow the
domes to open and close automatically. The circular saw blade will be attached to the DC
gear motor, which will be controlled by a MOSFET for precise cutting of fruit stems. The
researchers will carefully connect all wiring, insulate electrical components to prevent short
circuits, and test each sensor and motor individually prior to full integration. They will
consult electronics and programming experts to optimize device functionality and safety.
Once assembly is complete, the researchers will program the Arduino boards to capture
images, classify ripeness using a pre-trained algorithm, display the ripeness status on the
screen, activate LEDs to indicate fruit condition, and control the servo motors and cutting
mechanism.
d. Process of Function
When powered on, the device will use the camera to capture images of fruit inside the
dome. The Arduino will process these images and classify the fruit’s ripeness status as
unripe, ripe, or overripe based on color and texture.
• If the fruit is ripe, the green LED will light up. The device will automatically close the
dome, position the saw blade using the servos, and activate the motor to cut the fruit
stem.
• If the fruit is unripe or overripe, the red LED will light up, and the cutting mechanism
will remain inactive.
• The ripeness status will be clearly displayed on the OLED or LCD screen for
monitoring.
• A manual push button will allow users to override automatic operation if necessary.
After cutting, the dome and saw blade will return to their original positions, preparing the
device for the next fruit.
e. Testing
The device will undergo multiple trials to test its accuracy and efficiency in ripeness
detection and harvesting. The researchers will evaluate:
• The accuracy of ripeness classification displayed on the screen.
• The responsiveness and reliability of the automatic cutting system.
• The mechanical stability and durability of the device components.
• The ease of manual operation through the push button.
Feedback from experts and potential users will be gathered to assess the device’s
performance and identify opportunities for improvement.
References
Bekele, A. (2018). Post-harvest losses and food waste in developing countries: Causes and
solutions. Journal of Agricultural Science, 10(3), 45–56.
Food and Agriculture Organization (FAO). (2021). The State of Food and Agriculture 2021:
Making agrifood systems more resilient to shocks and stresses. Rome, Italy: FAO.
Koirala, A., Walsh, K. B., McCarthy, C., & Manandhar, S. (2019). Deep learning for real-time
fruit detection and ripeness estimation using a robotic system. Computers and
Electronics in Agriculture, 162, 431–442.
https://doi.org/10.1016/j.compag.2019.04.003
Kumar, P., Singh, R., & Sharma, V. (2020). Advances in post-harvest technologies to reduce
losses in fruits and vegetables. Food Processing & Technology, 11(2), 115–128.
Parfitt, J., Barthel, M., & Macnaughton, S. (2010). Food waste within food supply chains:
Quantification and potential for change to 2050. Philosophical Transactions of the
Royal Society B: Biological Sciences, 365(1554), 3065–3081.
https://doi.org/10.1098/rstb.2010.0126
Picon, A., Alvarez-Gila, A., Seitz, M., Ortiz-Barredo, A., Echazarra, J., Johannes, A., &
Villalobos, F. (2019). Deep learning-based plant disease detection using smartphone
images. Computers and Electronics in Agriculture, 163, 104863.
https://doi.org/10.1016/j.compag.2019.104863
Singh, V., & Sharma, R. (2018). Leaf health monitoring: Techniques and applications in
agriculture. International Journal of Agricultural Technology, 14(4), 589–599.
Wills, R. B. H., McGlasson, W. B., Graham, D., & Joyce, D. C. (2016). Postharvest: An
introduction to the physiology and handling of fruit, vegetables, and ornamentals (6th
ed.). University of New South Wales Press.
