© January 2021| IJIRT | Volume 7 Issue 8 | ISSN: 2349-6002
IJIRT 150658 INTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN TECHNOLOGY 191
Real-Time object color Identification
Atharva Borawake
1
, Nilima Kulkarni
2
, Anshul Ghorse
3
1,2,3
Department of Computer Science and Engineering, MIT School of Engineering, MIT Arts Design and
Technology University, Pune,412201, India
Abstract - The computer vision field is a rapidly growing
field devoted to analyzing and understanding digital
images. We can create computer vision projects through
OpenCV. In OpenCV image processing processes such as
image filtering, simple geometric photo transformation,
color space transition, histograms, etc. are covered.
Picture and real-time object color identification focus on
OpenCV color identification through using the RGB
model as well as the K-Nearest Neighbors Classification
algorithm trained on r, g, b pixel values. Color
identification in the image can be done through the RGB
value of the target pixel as input and then calculates the
distance, and the nearest color is chosen. From this
method, we can identify 800 plus different colors from
our datasets including the RGB value of each color. We
conduct extraction of features in real-time color
identification of objects to extract their RGB color
Histogram attributes from training images and trained
classification algorithm via RGB Color Histogram
attributes. The KNN classifier analyzes the webcam
frames and performs feature extraction and then shows
the color.
Index Terms - Color Histogram, Feature extraction, K-
Nearest Neighbor (KNN)
1.INTRODUCTION
Software applications & devices try to imitate human
eyes. Using libraries like OpenCV we can develop
computer vision applications. In OpenCV image
processing operations such as image color analysis,
color space conversion, histograms, etc. are covered.
Digital image identification is a program that helps
you to instantly acquire the identity of color by simply
clicking on it. Then, from each color, we will measure
the distance and determine the shorter one.
In Color identification of the real-time object, we can
pan the camera towards the target object to identify the
color. For this, we trained the KNN classifier on the
training dataset which contains color images and can
be updated with new colors. Color identification is a
technique used in different applications like
Photoshop and several other animation software as a
color dropper feature. For various AI-based systems or
robots color recognition of an object is required. This
paper represents an approach based on the K Nearest
Neighbor (KNN) classifier and feature extraction to
detect the real-time color of an object. The extracted
features are being used to get the Color Histogram of
training images. RGB Color Histogram attributes are
used to train the KNN classification algorithm.
Trained K-Nearest Neighbor (KNN) is classified to
scan live webcam scene by scene to perform extraction
of features from each frame, and then the color is
identified by trained K-Nearest Neighbor (KNN)
classifier.
Color identification is an area of research and interest
for more than five decades. Some of the research
surveys are given below.
Manuel G. et.al., 2020 used the method of the extreme
color channel, compared with the other two within
RGB. Observations: The target color channel value is
compared to the other two-color channels in the
equations. The color with the greatest channel value is
the target color.
Jayme et.al., 2016 used the method of three-color
models RGB, HSV, and CIE Lab to determine the final
result. Observations: Color transformation is applied
using RGB-related equations. The best result was
selected after applying transformation 3 times.
I.Al-Bahadlya et.al., 2005 used the method of the
intensity of the color determined using two-color
models RGB and CMY. Observations: Color pixels
are taken as inputs and recorded as RGB then
converted to CMY intensities.
Trupti et.al., 2013 used the method of Image
Segmentation Using XYZ Color Plane. Observations:
RGB values are converted into XYZ color space value
Then thresholding is applied to anyone's plane.
Shamik et.al., 2002 used the method of Image
Segmentation Using Features from the HSV Color
Space. Observations: The RGB value of a pixel is first