Development of Mechatronics Approach for Philippine Table Egg-

Grading Integrated with Rotten Egg Sorter Utilizing Binary Image

Processing

First Author: Maris Yray Adviser: Engr. Regien Mary Abano

Email: [email protected] Email: [email protected]
Phone : 09451494952 Phone: 09912166391

Second Author: Glenysse Gale Maligro

Email: [email protected]
Phone: 09565969082

Abstract –

This study presents an approach for developing an egg sorting machine that utilizes
binary image processing. The application of machine vision image processing, specifically binary
image conversion followed by pixel counting, is crucial for accurately assessing egg quality and
size. Two tests were conducted to evaluate the machine's performance: one comparing its
sorting accuracy and time with existing manual methods, and another examining the similarity
of readings produced by the machine. Results from the dynamic test revealed an impressive size
sorting accuracy of 93.2%, quality accuracy of 100%, and a processing capacity of approximately
1,183 eggs per hour for a sample of 100 table eggs. In contrast, manual sorting by human
experts resulted a size accuracy 49.1%, quality accuracy of 49% and for manual weighing
instrument yields a size accuracy of 89.4% and quality accuracy of 57.5%. Manual sorting
yielded significantly lower accuracy rates. The test of similarity of readings indicated low relative
variability given by the range of CV 0.12- 3.19% , demonstrating that the developed automated
table egg sorter can provide precise readings. Overall, this study successfully achieved its
objectives by employing binary image processing in the egg sorting machine, which significantly
improved accuracy compared to manual methods. Integrating deep learning algorithms into the

Making every drive

better™
machine offers a potential avenue for further enhancements, allowing it to adapt and improve
accuracy in handling diverse egg varieties and sorting tasks.

Keywords –binary image, image processing, machine vision, pixel counting, egg sorting machine

I. Introduction

Eggs are a highly sought-after poultry product in the Philippines, with a significant
increase in production observed from April to June 2022. This surge in demand has sparked
interest in technological innovations to enhance egg quality and production efficiency, as
manual sorting and packing processes prove to be costly and prone to inaccuracies. These
innovations aim to address the growing demand for high-quality eggs in the market. [1]
One prevalent method for sorting eggs in poultry farms involves the use of the candling
method to assess their quality. Candling allows for the observation of germ development
and can detect issues such as bloody whites, blood spots, or meat spots, which are critical
for ensuring the safety and quality of eggs. However, there are significant challenges
associated with this method. It requires a darkened room and relies on the observer's visual
acuity, making it susceptible to human error. Moreover, maintaining consistent and ideal
lighting conditions for candling can be difficult in practical farm settings. [2]
From a study conducted by Mahesa, Putrada, & Abdurohman in 2019, it was noted that
the candling method could be affected by the individual's eye vision, which may vary among
personnel tasked with this job. Additionally, achieving optimal lighting conditions
consistently can be challenging, leading to some rotten eggs slipping through the sorting
process and reaching consumers. This presents a clear need for a more reliable and
objective egg sorting method to ensure the highest quality eggs are delivered to consumers
and to streamline the process for poultry farms.

Making every drive

better™
 The literature review further underscores the importance of egg quality testing. It
reveals that egg quality has a direct impact on the price of eggs and highlights the need for
improvements in technology and equipment for producing, sorting, and selling egg
products. These notable studies have explored innovative approaches to enhance egg
quality assessment and sorting:

 Egg quality detection system utilizing fuzzy logic method: The Egg quality detection
system using fuzzy logic presents an innovative approach for efficient egg quality
assessment. It incorporates light and weight sensors into a microcontroller, accurately
detecting healthy and rotten eggs. Fuzzy Logic and Firebase methods enhance
categorization accuracy, resulting in a remarkable 95% accuracy rate with only one
misclassified egg among 20 tested. This research offers a practical and reliable method
for measuring egg quality, making it a valuable addition to the paper. It replaces
traditional manual egg quality assessment with a modern, sensor-based approach. [3]

 Automated single-line table egg sorting machine: This study presents an automated
egg sorting machine that combines machine vision and mechatronics principles.
Constructed with cost-effective materials, it utilizes a 12V DC geared motor for conveyor
operation and a DC servo motor for precise sorting. EGGSoTiC machine vision software
yields consistent results with low coefficients of variation for projected area and
estimated weight, demonstrating a 91% sorting accuracy in dynamic testing with 100
table egg samples, equivalent to an estimated hourly capacity of 1,426 eggs. While
minor weight estimation inaccuracies exist, they are close to the bias of 0.93 g, making
this study a valuable insight into a cost-effective and innovative egg sorting solution. [4]

 Design and Performance of an Automatic Egg Sorting System Based on Computer

Vision: Alikhanov et al. (2019) present an automatic egg sorting system that

Making every drive

better™
 distinguishes itself by sorting eggs based on size and irregular shapes, replacing
traditional weighing mechanisms with a smart camera, controller, and electro-
pneumatic actuator. Advanced image processing methods assess geometric properties,
facilitating egg weight calculation through regression. The system's flexibility allows for
easy category modifications, and evaluation at different conveyor speeds (2 and 3
eggs/second) using 760 eggs reveals high accuracy, reaching 94.6% at 2 eggs/second
and 90.3% at 3 eggs/second. This technical vision system significantly improves egg
quality determination and sorting efficiency, enhancing productivity, accuracy, and
parameter recording. [1]

By incorporating these studies into the research paper, researchers can gain valuable
insights into advanced technologies and cost-effective solutions that address the challenges
associated with egg quality assessment and sorting in the poultry industry. By leveraging
innovations such as binary image processing, machine vision software, and advanced image
processing methods, such a system has the potential to benefit both producers and
consumers in the poultry industry by enhancing efficiency, accuracy, and productivity in egg
sorting processes.

Making every drive

better™
 II. Innovation Discussion

The idea is to develop a method approach for an egg sorting machine utilizing binary image
processing, composed of key components: the egg hopper for feeding eggs onto the conveyor,
the vision chamber for image acquisition and scanning eggs for rot or health using LED lamps or
the candling method, and the sorting area for categorizing eggs by pixel counting to enable size-
based classification. Two main processes are involved in this binary image processing method:
sorting eggs as rotten or healthy and grading eggs into size categories, including pewee, small,
medium, large, extra-large, and jumbo. The main goal of this study is to assess the sorting
accuracy and capacity of the method implemented in the designed machine. Additionally, the
researchers aim to make a comparative analysis between this method and the currently existing
approaches employed in egg sorting machines. The main goal of this experiment is to assess the
sorting accuracy and capacity of the method employed in the designed machine and compare
these results with existing methods used in egg sorting machines. Key parameters used to
assess the success of this study include the actual capacity per minute, the average accuracy of
detection, and the repeatability of detection. To conduct dynamic accuracy evaluation of the
machine, a sample of one hundred (100) eggs is utilized in the study, which undergoes three
different experimental setups: (1) through the operation of the designed sorting machine, (2)
visual evaluation by a human expert or "classifier," and (3) manual loading into a weighing
instrument for sorting.
The study incorporates a process innovation by replacing the traditional candling method
with a novel approach of pixel counting applied to measure egg sizes through binary image
processing. Instead of relying on manual methods and utilizing RGB index values, the proposed
method directly employs binary image processing within the machine vision system for egg
sorting. Notably, this egg sorting process is seamlessly integrated into a single machine,

Making every drive

better™
featuring an egg hopper for input, the capability to sort rotten eggs, and the ability to classify
eggs by size.

Method Process Idea

Binary image processing is used in an egg sorting machine to convert the input image into a
binary format, allowing to distinguish between rotten and healthy eggs by depicting black and
white pixels. Pixel counting then evaluates the distribution of black pixels, identifying clusters to
distinguish between rotten and healthy eggs, and classifies eggs based on size by determining
the number of pixels they occupy.

A. Sorting Rotten Eggs using Grayscale to Binary Conversion

Figure 1.1 Grayscale to Binary Format

Based on Figure 1.1, the conversion of the grayscale image into binary significantly
reduces the range of gray shades to just two values: black and white. The image is improved to
facilitate subsequent analysis, resulting in gray-images where each pixel corresponds to a
varying shade of gray, ranging from black (0-low intensity) to white (1-high intensity). Below is
the step-by-step outline of the process for efficiently sorting rotten eggs.

Making every drive

better™
 (a) Thresholding and Segmentation

In converting grayscale image to binary format from figure 3.25, a threshold value of 95
was set to differentiate between rotten and non-rotten eggs. Pixels with intensity values below
95 to be assigned the value 0 (black), indicating potential rot, while pixels with intensity values
above or equal to 95 will be assigned the value 1 (white), indicating non-rotten.

(b) Feature Extraction

The proportion of black pixels (lower intensities) in the binary image can be computed.
Rotten eggs are likely to have a higher proportion of black pixels, indicating lower intensities,
compared to non-rotten eggs.

(c) Classification

By comparing the computed feature (proportion of black pixels) against a predefined

threshold or using machine learning algorithms, the eggs can be classified into "rotten" or "non-
rotten" categories.

B. Eggs Size Classification Grayscale to Pixel Counting

Figure 1.2 Egg size Pixel Counting Counting

Making every drive

better™
 To categorize eggs into different sizes based on the grayscale image, the pixel count
within each egg will be examined. Figure 1.1 illustrates the image processing outcomes,
demonstrating the production of cropped images using the object detection segmentation
method. To accurately segment the eggs and analyze their sizes, a threshold value was
established. Below, is the outline process for efficiently classifying eggs according to 6 sizes.

(a) Thresholding and Segmentation

With thresholding and segmentation technique ,the method successfully distinguishes

the eggs from the background, creating a clear boundary that allows for precise pixel counting.

(b) Pixel Counting and Size Categories

The threshold value serves as a crucial cutoff point, distinguishing the pixels associated
with the eggs from the image's background or noise. For instance, in the case of the pewee egg
size classification, where the weight ranges from 40 to 49 grams, the obtained pixel counts was
39,000 for 40 grams and 46,899 for 49 grams. When an egg's pixel count falls within this range,
it is classified as pewee. The same process applies to the rest of the egg size classifications.

Table 1.1 Weight Range per Classification

SIZE WEIGHT RANGE ( in grams/egg)

Jumbo 70 and above

Extra-Large 65 – 69

Large 60 – 64

Medium 55 – 59

Small 50 – 54

Pewee 40 – 49

Making every drive

better™
 Summary of the Process:

Figure 1. 3 Enhanced RGB Color Image Figure 1. 4 Grayscale to Binary Conversion

Figure 1. 5 Egg Size Pixel Counting

Egg Sorting Machine Idea

Making every drive

better™
Figure 1.6 3D Model of Egg Sorting Machine
 From figure 1.6, the egg sorting machine is an innovative invention that utilizes binary
image processing algorithms to automate the process of sorting eggs. Its primary objective is to
efficiently categorize eggs based on their quality (whether they are rotten or healthy) and size.
The machine comprises several key components, including the conveyor, egg hopper, vision
chamber, and sorting area. The egg hopper, shown in the figure above, serves as the initial entry
point for eggs into the sorting process. It carefully transports the eggs onto the conveyor,
ensuring a controlled and continuous flow throughout the entire sorting process. And the
sorting area, where eggs are separated into rotten or healthy categories and their sizes are
determined through evaluation by the vision chamber. The conveyor system is carefully
designed to minimize any damage to the eggs during transport, maintaining the integrity of the
product.

Making every drive

better™
 Note:

During the egg size classification

process, the Raspberry Pi (raspi)
communicates with the stepper
motor to initiate movements either
in a clockwise or counterclockwise
direction from the center, which
corresponds to the Large Bin
classification.

CW – Clockwise
CCW – Counterclockwise
MIS – Micro-step

Figure 1.8 Flowchart of the Programming System

Making every drive

better™
 Figure 1.8 illustrates a streamlined flowchart of the egg sorting machine system
utilizing binary image processing. The process initiates by capturing the raw image for RGB color
enhancement. Subsequently, the image undergoes preprocessing, converting it first to
grayscale and then to a binary image. Following this, the system performs Rotten Egg Detection.
If the egg is found to be healthy, it proceeds to size classification based on pixel counting. On
the other hand, if the egg is detected as rotten, it is immediately sorted into the reject
classification. Once both processes are completed, the sorting mechanism receives a signal
from the relay to categorize eggs according to their sizes.

Making every drive

better™
Figure 1.9 Fabricated Egg Sorting Machine

Making every drive

better™
III. Conclusion

The following conclusions can be drawn from the results of the study.

1. The designed machine effectively sorted eggs into different sizes and detected rejected
eggs based on predetermined ranges, showcasing its potential for automation and
accuracy in the sorting process.
2. The machine has an accuracy of 93.2% and achieved a sorting rate of 3 seconds per egg.
This demonstrated the system's efficiency and achieved the primary objectives set by
the study.
3. The rejection detection capability of the machine, with a perfect accuracy rate of 100%,
emphasizes its potential to identify and eliminate undesirable eggs from the batch,
which is crucial for maintaining product quality and safety.
4. The comparison of three sorting methods—the machine, manual sorting by human
expert, and manual sorting using a weighing instrument—revealed the superiority of the
machine in terms of accuracy and efficiency.
5. The assessment of repeatability and consistency in the machine's readings underscored
its reliability, which is vital for maintaining accurate and consistent sorting results across
multiple trials.
6. The correlation between mean pixel readings and egg weight, established through
regression analysis with an R-squared value of 0.98, highlights the feasibility of indirect
egg weight estimation using machine vision. However, the limitations of this regression-
based egg weight prediction were revealed through the RMSE of 2.18g, which
highlighted the impact of the narrow weight ranges of specific egg classifications. This
finding underscores the need for refining the prediction model for improved accuracy.
7. The utilization of machine vision in egg sorting aligns with previous studies that also
harnessed similar technology to achieve high levels of accuracy. The research

Making every drive

better™
 contributes to the growing knowledge of automated sorting techniques in the food
industry.
8. The study contributes to bridging a gap in the field by emphasizing the importance of
precise egg sorting in the food industry, particularly in achieving accurate size
classification and reject detection altogether in a single machine. The developed
machine offers a practical solution that can revolutionize egg-sorting processes,
ultimately benefiting both producers and consumers. The successful application of the
machine vision-based sorting method using the pixel counting approach opens avenues
for further research and innovation in automated quality control within the poultry and
food processing industries.

Overall, the study demonstrates the significant advancements that can be achieved by
integrating binary image conversion and pixel counting into machine vision systems. By
addressing specific challenges and optimizing the technology, the research lays the groundwork
for future advancements in efficient and accurate egg-sorting techniques, thereby contributing
to the larger field of automation and quality control in food production.

Making every drive

better™
IV. References

[1] J. Alikhanov, S. M. Penchev, T. D. Georgieva, A. Moldazhanov, Z. Shynybay, E. Stefanov and P. I.

Daskalov , "Design and Performance of an Automatic Egg Sorting System Based on Computer Vision,"
TEM Journal, vol. 8, no. 4, pp. 1319-1325, 2019.

[2] "Candling Eggs," University of Illinois "Incubation and Embryology", [Online]. Available: Available:
https://www.fao.org/poultry-production-products/production/en/..

[3] I. Mahesa, A. G. Putrada and M. Abdurohman, "Egg Quality Detection System Using Fuzzy Logic
Method," ResearchGate, vol. 4, no. 3, pp. 207-217, 2019.

[4] E. P. Quilloy, D. C. Suministrado and P. M. Bato, "Single-line automated sorter using mechatronics and
machine vision system for Philippine table eggs," Academics Journal, vol. 13, pp. 918-926, 2018.

[5] I. Mahesa, M. Abdurohman and A. G. Putrada, "Egg Quality Detection System Using Fuzzy Logic
Method," ResearchGate, vol. 4, no. 3, pp. 207-216, Augus 2019.

Making every drive

better™