The document describes a proposed system for detecting grain adulteration using deep neural networks. The objectives are to generate training data by labeling grain images, extract shape features to identify adulteration, train a model using online GPU resources, and understand the impacts of adulteration. The proposed system uses image processing techniques like brightness equalization and edge detection to preprocess grain images before segmenting and classifying them using a convolutional neural network model. This automated approach aims to overcome limitations of existing manual inspection methods.
1. )
SRI SIDDHARTHA INSTITUTE OF TECHNOLOGY
MARALURU, TUMAKURU-572105
(A Constituent college of Sri Siddhartha Academy of Higher Education, Deemed to be University)
DEPARTMENT OF INFORMATION SCIENCE AND ENGINEERING
PHASE-1 PROJECT SEMINAR
On
“DEEP NEURAL NETWORK -BASED GRAIN ADULTERATION DETECTION”
Presented by: Under the guidance of:
ADARSH R (19IS001) Mr Pradeep M
POOVI M S (19IS044) Assistant Professor
SRUJANA K(19IS066) Dept of ISE
3. Abstract
Introduction
Literature survey
Existing system
Proposed system
Objectives
System architecture
Methodology
System Requirements
References
4. Food is the basic need which provides nutrients to our body. Quality of food is the most necessary thing to be
checked as the adulteration of food grains is increasing on a greater pace.
Existing quality assessment is done manually which is tedious and include human errors (unknowingly or
intentionally).
This will affect the farmer who is relying on the farm for his daily bread, they don’t get fair price after their
years together efforts in farm.
Manual Assessment is also encouraging adulteration of food grains thus deceiving the customers by mixing
bad quality grains or materials which resemble the grains and gaining high margin profit.
5. INTRODUCTION
India has a very long history of grain adulteration and consumption for producing low cost grains.
Automation is utmost necessary when human inspection is misleading the society and selling cheap quality
food.
This project illustrates the grading system which process the grain image digitally and classify the grains into
Good, Average and Bad class by taking predefined values into consideration with the help of Digital Image
Processing and the grains and foreign materials are easily distinguished.
This can be achieved by training the system with a set of images and using these results, the system will test
the input image. It can automate the process which is reliable, factual and time saving ,life saving from the
toxic disease which is going to caused by the adulteration.
6. LITERATURE SURVEY
5/20
Sl.
No.
Title Year Description Advantages Disadvantages
1. Deep Neural
Network-Based
Sorghum
Adulteration
Detection in
Baijiu Brewing
2022,
IEEE
In this paper, they proposed
a method that uses sorghum
images as input and
combines image processing
and deep neural networks to
identify grain varieties and
calculate the adulteration
ratio.
The colour based segmentation of
adulterated region and calculate
using the pixel ratio.
The grain segmentation performance
was tuned by manually aligning the
Canny detector thresholds, blur filter
kernel size and other parameters,
which is in-efficient and holds low
generalizability.
7. LITERATURE SURVEY
5/20
Sl.
N
o.
Title Year Description Advantages Disadvantages
2. Automated recognition
and classification of
adulteration levels from
bulk paddy grain
samples
2018,
IEEE
The proposed
methodology consists of
five stages, namely image
acquisition, features
extraction, feature
selection, recognition of
adulterant paddy varieties
and classification of
adulteration levels
(%).Machine beings
identify paddy varieties
from their bulk sample by
their color and textual
features with OpenCV
methods.
Performance will be high
compared to tensor flow
If the lighting is different in the
testing environments it predicts
without efficiency and
Finds the adulteration with the
colour feature.
8. LITERATURE SURVEY
5/20
Sl.
No.
Title Year Description Advantages Disadvantages
3. Fruit quality detection
using opencv/python
2020,
IEEE
Contribution: This paper
presents an automatic fruit
quality detection system for
sorting and grading of
fruits and defected fruit.
Different types of adulteration
will be detected using the shape
and pattern of adulteration.
Even if the quality of fruit is good,
the accuracy turns out to be low
because it considers only outer
appearances.
4. Image Segmentation
K-Means Clustering
Algorithm for Fruit
Disease Detection
Image Processing.
2020,
IEEE
Clustering and fruit image
segmentation algorithms
are implemented for
identifying the fruit
diseases.
The detection time will be very
fast compared to other
algorithms.
The dataset used here are static
images and k means clustering cannot
handle noisy data and outliers which
turns out to give less accurate output.
9. LITERATURE SURVEY
5/20
Sl.
No
.
Title Year Description Advantages Disadvantages
5. Fruit Disease
Classification and
Identification using
Image Processing.
2019,
IEEE
An image processing
approach is proposed for
apple fruit disease
identification and
categorization using
different color, texture and
shape feature
combination.
Disease will be detected using
different shape and pattern of
disease.
The SVM gives lesser accuracy than
CNN.
6. Development of
Machine Learning
based Fruit Detection
and Grading system.
2020,
IEEE
The proposed system
captures the fruit placed
on conveyor belt then the
captured image is
compared with the trained
data set using CNN which
extracts the features of the
fruits like texture, color,
and size.
The accuracy is high because of
different fruit where different
shapes for each class
Adam optimizer gives high
accuracy in object detection
Even if the quality of fruit is good,
the accuracy turns out to be low
because it considers only outer
appearances.
10. The proposed methodology consists of five stages, namely image acquisition, features extraction, feature
selection, recognition of adulterant varieties and classification of adulteration levels (%).
Machine begins to identify dal varieties from their bulk sample by their color and textual features with
OpenCV methods.
The color feature extraction starts with the separation of the color channels of the work images. In this,
the images in the RGB color model are separated into R, G, and B components respectively.
EXISTING SYSTEM
11. PROPOSED SYSTEM
A neural network model is developed for classification. will developed a knowledge based nearest mean
classifier for classification of Bulk grain objects like (dal) raw.
Machine vision systems are successfully used for recognition of dal using Convolutional Neural
Network(CNN) .
A method for the classification and gradation of different grains (for a single grain kernel) such as Dal is
described .we using the feature of extracting the shape and structure find the adulteration to get more
efficient working model..
12. OBJECTIVES
Data generation for training with the image name(label_img)
To know the shape difference of grain adulteration.
Labelling the grain with the help of Makesense.ai.
Splitting the data and training with the online GPU google Collaboratory.
To know other effects of grain adulteration on society health.
14. METHODOLOGY
Brightness equalization:-When taking pictures to acquire images, the uneven illumination caused shadows in the
background, which will affect the edge detection effect and also lead to differences in the histogram distribution of
color space between segmented images, reducing the data uniformity. Hence, it is necessary to identify and subtract the
background, and adjust the image to equalize the brightness.
Edge detection:- Convert the image to grayscale, use low-pass filtering methods such as Gaussian blur to remove the
image noise. Take morphological operations like erode and dilate to separate adjoint sorghum grains. Then find the
outside contours of each connected component using the Canny edge detector .
Image segmentation:- Approximate the contours with polygons using Douglas–Peucker algorithm . Calculate the
minimal up-right bounding rectangle of each polygon, manually set thresholds for width and height to filter noise
points. To preserve the size information of the grains, we used square with fixed side length (140 pixels) to segment
each granule. The square has the same center as the bounding rectangle, and its side length is larger than the maximum
value among the long sides of the rectangles.
15. HARDWARE AND SOFTWARE REQUIREMENTS
HARDWARE REQUIREMENTS
System : intel i3/i5 2.4 GHz.
Hard Disk : 500 GB
Ram : 4/8 GB
SOFTWARE REQUIREMENTS
Operating system : Windows XP/ Windows 7.
Software Tool : Open CV or Python
Coding Language : Python
Toolbox : Image processing toolbox.
16. REFERENCES
[1] H. Liu and B. Sun, “Effect of fermentation processing on the flavor of Baijiu,” J. Agricult. Grain Chem.,
vol. 66, no. 22, pp. 5425–5432, 2018, doi: 10.1021/acs.jafc.8b00692.
[2] X.-W. Zheng and B.-Z. Han, “Baijiu, Chinese liquor: History, classification and manufacture,” J. Ethnic
Grains, vol. 3, no. 1, pp. 19–25, 2016, doi: 10.1016/j.jef.2016.03.001.
[3] H. Liu, A. O. Zonghua, M. Wang, X. Liu, J. Chen, and H. Zhou, “Research progress in Baijiu-making
Sorghum,” Liquor Making Sci. Technol., to be published.
[4] M. Guo, Y. Bao, Y. Huang, and Y. Huang, “Fermentation characteristics of Moutai-flavor Baijiu with
different sorghum varieties,” China Brewing, to be published.
[5] H. Eksi-Kocak, O. Mentes-Yilmaz, and I. H. Boyaci, “Detection of green pea adulteration in pistachio nut
granules by using Raman hyperspectral imaging,” Eur. Grain Res. Technol., vol. 242, no. 2, pp. 271–277, Feb.
2016, doi: 10.1007/s00217-015-2538-3.
17. [6] S. Verdú, F. Vásquez, R. Grau, E. Ivorra, A. J. Sánchez, and J. M. Barat, “Detection of adulterations with different
grains in wheat products based on the hyperspectral image technique: The specific cases of flour and bread,” Grain
Control, vol. 62, pp. 373–380, Apr. 2016, doi: 10.1016/j.graincont.2015.11.002.
[7] B. S. Anami, N. N. Malvade, and S. Palaiah, “Automated recognition and classification of adulteration levels from
bulk paddy grain samples,” Inf. Process. Agricult., vol. 6, no. 1, pp. 47–60, Mar. 2019, doi: 10.1016/j.inpa.2018.09.001.
[8] P. Vermeulen, M. Suman, J. A. Fernández Pierna, and V. Baeten, “Discrimination between durum and common wheat
kernels using near infrared hyperspectral imaging,” J. Cereal Sci., vol. 84, pp. 74–82, Nov. 2018, doi:
10.1016/j.jcs.2018.10.001
REFERENCES