IRJET- Assessing Food Volume and Nutritious Values from Food Images using Decision Tree Approach

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 12 | Dec 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 1458
Assessing Food Volume and Nutritious Values from Food Images using
Decision Tree Approach
Gopiga T1, Arunpriya C2
2Assistant Professor
1,2Department of Computer Science, PSGR Krishnammal College for Women, Coimbatore, Tamil Nadu, India
----------------------------------------------------------------------***---------------------------------------------------------------------
Abstract – Obesity and being overweight have become
growing concerns due to their association withmany diseases,
such as type II diabetes, several types of cancer and heart
disease. Thus, obesity treatments have been the focus of a
large number of recent studies. Because of these studies,
researchers have found that the treatment of obesity and
being overweight requiresconstantmonitoringofthepatient’s
diet. One of the important steps in the success of healthy dietis
measuring food intake each day. One of the challenges in
obesity management studies is measuring daily food
consumption for obese patients. Countless recent studies have
suggested that using technology like smart phones may
enhance the under-reporting issue in dietary intake
consumption. In this thesis, we propose a Food Recognition
System (FRS) for calories and nutrient values assumption. The
user employs the built-in camera of the smart phone to take a
picture of any food before and after eating. The system then
processes and classifies thephotographstodiscoverthekindof
food, portion size and then uses the knowledge to estimatethe
quantity of calories within the food using decision tree. An
essential step in the system as it is used to estimate and
calculate the food volume and amountofcaloriesintheimage.
Keyword: Image Processing, Food Obesity, Monitoring and
decision tree
1. INTRODUCTION
High calorie intake has proved harmful worldwide, as it has
led to many diseases.However,dieticianshavemistakenthat
a standard intake of number of calories is essential to
maintain the right balance of calorie content in human body.
There has been an unprecedented rise worldwide in over
weightiness and obesity, as well as in diseases they cause.
Thus, monitoring daily eating routines in order to avoid
excess calorie intake has become an important issue in
maintaining a good quality of life. Studies have shown that
several diseases are linked to excessive calorie intake in
humans. Pan found that breast, colon, and prostate cancers
are caused by high calorie intake. High calorie intake was
found to be only second to tobaccoindirectlycausingcancer.
Moreover, a previous study found that a proper diet
involving lower calorie intake helped the residents of
Okinawa to increase their life expectancy and lower the risk
of age-associated diseases. High calorie values in foodthat is
nutritionally poor led to systemic inflammationandreduced
insulin sensitivity, as well as a cluster of metabolic
abnormalities, including obesity, hypertension, and glucose
intolerance. a review of the literature showed that reducing
the calorie intake lessened the risk of cancer in humans. To
assist people in tracking their calorie intake, efficient and
convenient mobile applications have been developed that
alert users to the number of calories they consume. Such
mobile applications are becoming increasingly popular.
Mobile applications have the capacity to provide the easy
collection of personal health-related data and timely
behavioral cues. Additionally, research has focused on the
benefits of mobile and 2 Internet technologies in reaching
diverse populations, such as rural communities, in order to
reduce health disparities and promote health interventions.
Obesity treatment needs the patient to note the amount of
the daily food intake, but in most cases, it is not simple for
the patients to measure or control their daily intake due to
the lack of nutrition or self-control. Therefore, by using an
automatic food intake monitoring system, we can assist the
patient and provide an effective tool for the obesity
treatment. Nowadays, new technologies such as computers
and smart phones are involved in the medical treatment of
different types of diseases, and obesity is considered as one
of the common disease. From the last Calorie intake must
rely mostly on the weight of the individual, his or her daily
activity, age and gender. Eachpersonshoulddailyconsumea
certain amount of calories. If the amount of calories
expenditure is increased, it will lead to weight gain and,
therefore, the risk of obesity. Other facts for any food
categories or any food item thus, all nutrient facts tables
should include the number ofcaloriesplus.Inoursystem, we
will use the nutrient fact tables and the amount of calories
for each type of food as a basic criterion, and this will allow
us to find the amount of calories in a food image. In fact, the
FRS relies on the already established nutrient facttablesasa
reference to estimate the number of calories from any
selected food photo. These data are stored in the system
database in tables. Image processing is a form of signal
processing in which an image, picture or frame from a video
is processed to produce another image or a set of
information, parameters or specific data obtained from the
characteristics of the initial image to be analyzed. In our
case, we used image processing to analyses the two-
dimensional signals inside the data oftheimagetodefinethe
contours of the food in the image, perform a segmentation
and a measurement of the objects to obtain an
approximation of the real-life size of the portions,andfinally
allow the nutritional facts calculations with the information
obtained from the image processing procedure. An image or
a set of images is used as input of the FRS (re-image)plusthe
objects present inside each photo(thethumb).Theseimages
are analyzed by signal processing, and the outcome of the

image processing is another image, images, or a set of
features extracted from the original images. The pre-
processing action comprises color, texture featuresand,size
and shape. Many tests and trials should be done to achieve
the desired result image processing requires. In this paper
an automatic dietary monitoringofcanteencustomersthatis
based on histograms feature techniques for automatic food
recognition and leftover estimation in a canteen scenario.
Although the canteen scenario includes some apparent
simpliﬁcation such as controlled image acquisition
conditions, known weekly menu etc., the problem of food
recognition and leftover estimation. Nowadays, the major
problem faced by people in food sector is due to the
enormous variations in the tray and plate composition. The
visual appearance of the same dish may greatly change
depending on how it is placed on the plate. The system is
able to identify and recognizethefoodcategoryandestimate
the amount of food and carbohydrates. Moreover, we can
classify by decision tree and identify different calories of
food.
2. LITERATURE REVIEW
S. Saha and D. Nandi [6], twenty well known data mining
classification methods are applied on ten UCI machine
learning medical datasets, and the performance of various
classification methods are empirically compared while
varying the number categorical and numeric attributes. The
types of attributes and the number of instances in datasets.
Out of this twenty classification methods Bayes Net., Naïve
Bayes, Classification via Regression, Logistic Regressionand
Random Forest Classification methods are thebest.NB-Tree,
Classification via Regression and Bayes Net methods are the
best for categorical attributedatasets.Outoftheseabovefive
rule based classification methods PART and Decision Tree
methods are the best.
Kiran Ambhore [7], Obesity is the major cause ofoverweight
this leads to the type II diabetes, heart disease and cancer. A
food is said to be healthy after measuring its calories and its
nutrition and it is a challenging task. Smartphone plays a
vital role in today’s technological world using this technique
will enhance the issue in intake of dietary consumption. A
food image recognition system was being devolped for
measuring the nutrition and calorie values. After taking the
food image the colour, shape, size and texture features are
extracted and it is given to the K-nearest neighbour (KNN)
for recognizing the food and then the nutrition tablehelps to
measure the calorie values.
I. M. N. A. Yasa, [8], C5.0 algorithm is used to create a
decision tree based on the experts, while fuzzy logic used to
categorize the type of hypertensive disease that suffered by
the patient and increase the level of accuracy of the
diagnosing system. The accuracy of the combination
between c5.0 algorithm and fuzzy logic is about 97.19%
treats an evaluation of the analyzed results of classification
algorithms selected for better prediction based on the
characteristics of data from thedata setwithliverdisease, by
Naïve Bayes, Decision Tree, Multilayer Perceptronandk-NN
used in a previous study, which developed our data set, and
Those algorithms were compared in several kinds of
evaluating criteria like precision, recall, sensitivity,
specificity, and so on. Through the experiments, we could
know that in view of precision, in the performance of
prediction test, NaïveBayesiscomparativelypreferablethan
others, but in other criteria such as Recall and Sensitivity,
Logistic and Random Forest took precedence over other
algorithms as considering the algorithmic characteristics to
liver patient data set.
Daniele Lima da Cruz1 Márcia Nacif [9] New nutritional
guidance tools have been used by various health
professionals. Among these are health-oriented cell phone
applications with several resources that can help the user
maintain a healthy diet. In this context, this study had as
main objective to build a nutritional assessment application
for mobile phones with Android system. An app called Diet
Help was developed, with seven screens: “User Profile”,
“Food Record”, “Summary of the day”, “Diet history”, “Food
calories”, “Healthy Eating” and “Delete User”. The main
programming languages used were Android Java, Javascript
and SQlite for database. Diet Help allows calculation of body
mass index, nutritional status according to age and sex,
informs recommended daily energy intake, the amount of
calories and nutrients consumed in each meal and if
consumption (intake) is in accordance with the
recommended levels for the user age and sex. It also allows
users to find the amount of calories and nutrients per 100g
of any food and provides information on healthy eating
based on the food guide
3. PROPOSED SYSTEM
Food is one of the most important requirements of every
living being on earth. The human beingsrequiretheirfoodto
be fresh, pure and of standard quality. The standards
imposed and automation carried out in food processing
industry takes care of food quality. Now a day, peopleacross
the universe are becoming more sensitive to their diet.
Unbalanced diet may cause many problems like weightgain,
obesity, diabetes, etc. Hence, different systems were
developed so as to analyze food images to calculate calorie
and nutrition level. This system proposes an effectivewayto
measure and manage daily food intake of patients and
dietitians. The system will take the images of food and using
image processing, segmentation and classification it
calculates the nutrition and calorie content in the food. The
proposed system will certainly improve and facilitate the
current calorie measurement techniques. In this paper,food
portion recognition system uses for measuring the calorie
and nutrition values. The user will just take a picture of the
food image then recognize the image to detect the type of
food portion and classify the image using decision tree. To
classify the image perform segmentation, food portion
recognition using skull striping and classification using
support vector machine to calculate the calorie along with
the type of energy in accurate way.

3.1 Pre-Processing
Once the picture of the food is captured by the user, the
image is transformed and prepared for the next step, which
we will explain in the following subsection. In the beginning,
a simple conversion must be performed on the image to
change the image size into a standard format for precise
results for system segmentation.Thus,thesizeof eachimage
will be compared with standard size categories. We have
defined one size category as a standard, whichare970×720
pixels for simplicity. Larger images will be reduced to this
size before accomplishment of any image-processing
technique.
3.2. Segmentation
Segmentation is a process of extracting and representing
information from an image is to group pixels together into
regions of similarity Segmentation subdivides an image into
its constituent regions. According to a set of predefined
criteria similar features are being segmented. Clustering
pixels into salient image region is the main aim of image
segmentation. Segmentation could be used for object
recognition, occlusionboundaryestimationwithinmotionor
stereo systems, image compression, image editing, or image
database look-up. The segmentation phase starts
immediately after analyzing the pre-processing step. This
part will operate with four different features: color, texture,
shape and size, on which we are mainly concentratinginthis
project. These parts also include the calculation in pixels of
the thumb and its size in pixels. The extracted size will be
used in transforming the pixel size of food portionstoactual,
real-life size. In addition, the color feature will be extracted
by using the color histogram, while the size feature will be
extracted by including the pixels in the Region of Interest
(ROI) for each food portion. Moreover, this will give us the
shape feature which will be used in our calculating method.
Standardized datasets are needed for the assessment of
segmentation algorithms. The segmentation is often the key
step in interpreting the image, if an image has been
preprocessed appropriately to remove noise and artifacts.
The process in which regions or features sharing similar
characteristics are identified together is called image
segmentation. Statistical classification, thresholding, edge
detection, region detection, or any combination of these
techniques may be used in image segmentation. The most
segmentation techniques are either region-based or edge
based; the output of the segmentation step is usually a set of
classified elements. In image processing,Segmentationisthe
process of partitioning a digital imageintoseveral segments.
The goal of segmentation is to simplify an image into
something that is more meaningful and easier to analyze.
3.3 Texture feature extraction and Food PortionVolume
Measurement
It is extremely appropriate to estimatethemeanandalsothe
variance of the energy of the filtered image. To calculate the
surface area for a food portion, weproposetosuperimposea
grid of squares onto the image segment so that each square
contains an equal number of pixels and, therefore, equal
area. First comparison with other methods, the grid will
easily match with irregular shapes, which is important for
food images because most of the food portions will be
irregular. Naturally, there will be some estimation error,but
this error can be reduced by making the grid finer. Second,
depending on the processing capabilitiesoftheuser’smobile
device and the expected system response time from the
user’s perspective, we can adjust the granularity of the grid
to balance between the two factors.
Fig -1: A captured food images to find calorie
3.4. Classification using decision tree
The processing of finding a set of model thatdistinguishdata
class for the purposes of being able to use the model to
predict the class of objects whose class label is unknown is
called classification. Classification istheprocesstoconstruct
a model based on the training set and uses it to classify new
data or test set. It is a supervised learning as observations;
measurements are accompanied by known class labels in a
large amount of training set and new data is classified based
on training set. Classification is one of main activities in the
distillation of knowledge by learning supervised approach.
High calorie intake has proved harmful worldwide, as it has
led to many diseases.However,dieticianshavemistakenthat
a standard intake of number of calories is essential to
maintain the right balance of calorie content in human body.
In this thesis, we consider the category of tools that use
image processing to recognize single and multiple mixed
food objects, namely decision tree. We propose a method for
the fully automatic and user-friendly calibration of the sizes
of food portions. This calibration is required to estimate the
total number of calories in food portions. In this work, to
compute the number of calories in the food object, we go
beyond the finger-based calorie calibration method that has
been used in the past, measuring the distance between the
user and food object has been done automatically.

A decision tree is a predictive modeling technique from the
fields of machine learning and statistics that builds a simple
tree-like structure to classify the data according their
categories. A decision tree is like a tree structure, in which
each internal node denotes a test on an attribute. Decision
tree model worked as step by step then food addiction is
used as data source. Data Training is used to determine
entropy an information gain. Then, select max value gain.
Choice node by gain maximum. TargetisClass.DecisionTree
based on Decision Tree Induction.
Step 1: Tree is constructed in a top-down recursive divide-
and-conquer manner.
Step 2: At start, all the training examples are at the root.
Step 3: Attributes is categorical (ifcontinuous-value,they are
discretized in advance).
Step 4: examples are partitioned recursively based on
selected attributes.
Step 5: test attributes are selected on the basis of a heuristic
or statistical measure.
Conditions for stopping partitioning when all samples or a
given node belong to the same class. There are no remaining
attributes for further partitioning-majority voting is
employed for classifying the leaf, and there are no samples
left. Entropy is a measure of uncertainty associated with a
random variable.
4. EXPERIMENTAL RESULT
The estimation of food volume through an image is a major
challenge in dietary intake assessment applications. In this
section, we will examine the methodology of obtaining the
volume of any food image by utilizing the area size that has
been extracted from the photo after the image analysis and
the shape recognition process. As soon as the photos of the
selected food are captured, the application starts to analyse
the pixels of both thumb and meal from the first photo (top
view).
Fig 2 -Pre-Processing image
The main resulting value from this calculation is the area
size (height and length) in pixels, which will beusedwiththe
other dimensions of the food item from the second side of
photo (photo) to generate the width (volume). Finding the
volume of the photo leads us to easily calculate the amount
of the calories in the calories selected food via a special
algorithm that depends on the nutritional tables stored
inside our application. In the next subsections, we will
elucidate in detail how we calculate volume in irregular and
regular shapes.
The main objective of the FRS is to estimate the amount of
nutrition values and calories for any food type from an
image. Thus, caloric estimation is the main, final stage for
our system. In fact, the importance of using the already
stored nutrient fact tables will also appear in this stage. In
general, the system starts to calculate
Fig 3 - Calculating calories
the calories by comparing the inputs from the image (mass)
with the inputs from the nutrient tables (mass measured by
gram and calorie amount, which is measured by calories),
which are already stored in the application’s database. We
propose a novel method to measure and estimate the
amount of calories in any image.
After calculating the size, we started applying the proposed
method to compute the mass of the food inside the image.
Thus, we used the known variables, which are volume and
food density, to calculate the mass. We also measured and
weighed all the selected food manually to ensurethevalidity
of the proposed method at the same time. Thus, finding the
mass of the selected food allowed us to calculate the amount
of the calories and this is simply by applying our proposed
equation that allowed us to compare the amount of calories
and mass in the standard tables with the estimated results.

Table - 1: Volume of different types of simple food
5. CONCLUSION
Now-a-days, people are getting more health conscious and
tend to keep a check on nutritional gain from the packed
food products they consume. Main problem in human
nutrition is the assessment of usual food intake. The recent
proposal of e-Health personalized intervention has been
given. Performing dietary assessment using digital photosof
food became popular. The system then processes and
classifies the images to detect the type of food and portion
size, then uses the information to estimate the number of
calories in the food using decision tree. Our system provides
better result.
REFERENCES
1. R. Almaghrabi, G. Villalobos, P. Pouladzadeh, and S.
Shirmohammadi, "A Novel Method for Measuring
Nutrition Intake Based on Food Image," in Proc.
IEEE International Instrumentation and
MeasurementTechnologyConference,Graz,Austria,
2012, pp. 366– 370.
2. Y. Yue, W. Jia J. D. Fernstrom, R. J. Sclabassi, M. H.
Fernstrom, N.Yao, M. Sun, “ Food volume estimation
using a circular reference in image-based dietary
studies,” In Proceedings of 35th Northeast
Biomedical Engineering Conference, New York, NY,
USA 2010.
3. Puri, M., Zhu, Z., Yu, Q., Divakaran, A., & H.S.
Sawhney, "Recognition and volume estimation of
food intake using a mobile device", in Proc. WACV,
2009, pp. 1-8.
4. Z.N. Zhang, Y.F. Yue, W.Y. Jia, et al., “Food Volume
Estimation from a SingleImageUsingVirtual Reality
Technology,” 2011 IEEE 37th Annual Northeast
Bioengineering Conference (NEBEC), 2011, Troy.
5. Y. Yang, Y. Yue, Z. Wei, J. Robert, W. Jia, and M. Sun,
"Food volume calculation in different imaging
scenarios," in 2011 IEEE 37th Annual Northeast
BioengineeringConference(NEBEC),April 2011,pp.
1-2.
6. S. Saha and D. Nandi, “Data classification based on
decision tree, rule generation, bayes and statistical
methods: an empirical comparison”, International
Journal of Computer Applications, vol. 129, no.7,
(2015), pp. 36-41.
7. Kiran Ambhore, “Measuring Calories and Nutrition
from Food Image”,IJARCCE, june 2016
8. I. M. N. A. Yasa, I. K. G. D. Putra and N. M. I. M.
Mandenni, “Hypertension expert system with c5.0
algorithm and fuzzy logic”, Telkomnika Indonesian
Journal of Electrical Engineering, vol. 12, no. 7,
(2014), pp. 5669-5677
9. Daniele Lima da Cruz1 Márcia Nacif, “Nutritional
assessment application for mobile phones with
Android system”., Demetra; 2015; 10(4); 977-988
10. Beasley, J., "The pros and cons of using pdas for
dietary self-monitoring," Journal of the American
Dietetic Association, vol. 107, no. 5, 739, 2007
11. J. Nishimura and T. Kuroda, Eating habits
monitoring using wireless wearable in-ear
microphone, International Symposium on Wireless
Pervasive Computing, 2008, pp.130- 133.
12. Martin, C.K., Kaya, S., and Gunturk, B.K.,
"Quantification of Food Intake Using Food Image
Analysis," Conference Proceedings IEEE English
Medical Biology Society, pp. 6869– 6872, 2009.
13. Zulkifli, S.N., & Yu, S.M.: The food frequency method
for dietary assessment. Journal of the American
Dietetic Association, 1992, 92(6):681-685.
14. Fox, T.A., Heimendinger, J., & Block, G. The use of
telephone surveys as a methodforobtainingdietary
information: A review. Journal of the American
Dietetic Association. 1992; 92:729-732
15. Kristal AR, Shattuck AL, Williams AE. Food
frequency questionnaires for diet intervention
research. In: Proceedings of the 17th National
Nutrient Databank Conference, June 7-10, 1992;
Baltimore, Md. Washington, DC: International Life
Sciences Institute Press; 1994:110-125.

IRJET- Assessing Food Volume and Nutritious Values from Food Images using Decision Tree Approach

More Related Content

Similar to IRJET- Assessing Food Volume and Nutritious Values from Food Images using Decision Tree Approach (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- Assessing Food Volume and Nutritious Values from Food Images using Decision Tree Approach