2021 3rd International Conference on Advances in Computing, Communication Control and Networking (ICACCCN)

Object Detection using OpenCV and Python

2021 3rd International Conference on Advances in Computing, Communication Control and Networking (ICAC3N) | 978-1-6654-3811-7/21/$31.00 ©2021 IEEE | DOI: 10.1109/ICAC3N53548.2021.9725638

Ayushi Sharma Jyotsna Pathak Muskan Prakash

B.Tech CSE(AI&ML) B.Tech CSE B.Tech CSE
Galgotias University Galgotias University Galgotias University
Greater Noida, India Greater Noida, India Greater Noida, India
[email protected] [email protected] [email protected]

J N Singh
Galgotias University
Greater Noida, India
[email protected]

Abstract- The paper has covered topics ranging from how

artificial intelligence and machine learning algorithms help in
object detection to how OpenCV is such a useful tool for How Video Object Detection works:
beginners who wish to learn how real time object identification So, how do we detect a moving object?
and tracking can be done. It also shows the flexibility of a The purpose is to record the coordinates of the entity in
tracking system to a moving camera, ideal for automotive safety motion and then encircling the object in the video.
applications. Image identification makes use of techniques like
detection of an object, its recognition, and segmentation. The use
of artificial intelligence and machine learning enhances the rate
of processing the data and maintaining the standard of the
outcome. Example, by using artificial intelligence, we can very
easily complete difficult tasks.

Keywords—Pedestrian tracking, Moving vehicle, Intelligent

vehicles, Unattended driving systems, Intelligent driving decisions

I. INTRODUCTION

OpenCV is a library catering to thousands of ml algorithms

and hundreds of functions that support these algorithms.
OpenCV approach includes Python, C++ and java and runs
on all desktop and mobile systems. It stands for Open-Source
Computer Vision Library. OpenCV contains many modules
including module for image processing, for identification of
object, and ml. By the use of it, we achieve, constrict, build
up, replace, retrieve information. The new algorithms, that
represents the software is categorized into groups, that Figure 1: Object Recognition
include features, learning, and mixture of both known as
hybrid. Yet, this tracking has many uses in different areas, Then we want our model to detect the operating entity as
some of which are monitoring traffic flows, speeds on shown in Figure1. The unit in motion (i.e., car) will be
different lanes, latest innovations in robotics, supervision, recognized and then encircled as a rectangle.
security and video connection. Therefore, this application
requires optimal usage of computation, basic transmission, Either, a model can be trained for object identification or
and precision over the network. Returns related to trained model could be used which would identify it the data.
computation and transmission depend upon quantity as well Although, both are techniques used in learning. They need
as variety of coordination between cameras used to collect labelled information to teach the object identification model.
data, distribution and checking to verify the verdict so as to Or we could also use frame differencing approach. Whenever
make the estimation errors minimum. an object is seen in motion, it means that the object is at a
distinct position at each consecutive frame.
II. WORKING EXAMPLES OF OBJECT IDENTIFICATION:
Nowadays, video identification is established across many We presume that other than the object nothing moves in a pair
domains of corporations. Its usage ranges from video of successive frames. This is how frame differencing works.
surveillance, sports broadcasting, electrical cars, robot
navigation and many more.
Cascading:
TrainCascadeObjectDetector is a classifier function.
CascadeObject Detectors system object is in MATLAB’s
ISBN: 978-1-6654-3811-7/21/$31.00 ©2021 IEEE
501

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on December 29,2024 at 13:28:33 UTC from IEEE Xplore. Restrictions apply.
 2021 3rd International Conference on Advances in Computing, Communication Control and Networking (ICACCCN)
computer vision toolbox One of the uses is to build a custom IV. COMPARATIVE STUDY
classifier to detect categories of objects which are fixed, for
A. INTRODUCTION:
instance face, human body, cars, etc. A group of classifiers
arranged in stages is known as Cascade object detector. For To acknowledge individuals in a real video surveillance
instance, classification output from every expert E n for input system, we spot humans automatically. Pedestrian detection
xi is given as En(xi). To distinguish amongst training, En(xi) detects areas where a human is present.
only two results expected that are written as +1 or −1, i.e., Person detection is the process of predicting and localizing
En(xi) ∈ [−1, +1]. The joint opinion of experts is written as each person in the image, represented in a rectangular box
L(xi). It shows the combination as below: that is made around the image.
B. METHOD OF PEDESTRIAN TRACING:
L(xi) = w1E1(xi) + w2E2(xi) + · · · + wnEn(xi)
It mainly has 2 models:
where E1(xi), E2(xi), . . ., En(xi) represents the decisions from 1. Hand-crafted model which is based on hand-crafted
n experts and w1, w2, . . ., wn are weights specified. The steps features and is widely used for object detection.
of for classifier training are: 2. Deep learning model uses convolutional neural
networks to focus on improved performance.
1. Provided a set to train data with a pair of images (x i, yi),
where xi is a true /false picture, and yi is the tag assigned to C. PROBLEM FORMULATION:
each image. The value for true pictures is 1, and that of false
pictures is 0. Safety standards and accident prevention system in cars have
gained notable development. Advanced driver assistance
systems are commercially available now and mostly found
2. Setting up weights w1, i= 1/2p, 1/2n for true and false
pre-installed by car manufactures. The aim of these systems
pictures, where p and n are the count of true and false
is to provide automobiles with sensors capable of detecting
pictures, correspondingly.
and acting when threatening situations are faced, so that the
driver would be able to avoid a collision. Detecting
3. For f = 1, . . ., F, where F is the number of levels of training pedestrians, launches an alert and then acts upon their
and n sets of images. unpredictable behavior.

4. Upgrade weight wf+1, i = wf, iβ 1−fi y, when ei = 0 then xi is Thus, the full potential of ADAS systems can be achieved by
sorted in a correct manner, ei = 1 otherwise and βy = ef/ (1 – including AI car and pedestrian tracing. To utilize such
ef). strong ML with computer vision methods, we employ
a cascade of classifiers. Each classifier is trained to model a
III. IMAGE RECOGNITION pedestrian. The targets are detected Haar features. This
process is done to be able to operate in real-time and will help
Image identification makes use of techniques like detection in future, especially in self-driving cars.
of an object, its recognition, and segmentation.
The use of artificial intelligence and machine learning
MERITS:
enhances the rate of processing the data and maintaining the
standard of the outcome. Example, by using artificial While driving if you take your eyes off of the road to reach
intelligence, we can very easily complete difficult tasks. for your coffee turn to tell your kids to quiet down, and then
suddenly you see a pedestrian is crossing the road right in
The more conventional computer vision and ML approach as front of you. You push the brakes—but it might be late.
opposed to deep learning, the following have been useful:
• determining the most appropriate features (HAAR It is a very scary yet usual scenario. One out of three crashes
features, image color histogram, etc.) involve a vehicle going straight as a pedestrian crosses the
• inspecting and learning about various techniques road. To prevent these mis-happenings, automakers now
provided by OpenCV provide a “pedestrian detection” system in some of their
• applying grid search to search the suitable classifier models. The ability to reliably detect vehicles provides huge
advantage to everyone.
Also, the following problems could arise:
• To decide correct position of our sliding windows Haar Cascade Classifiers:
• To make sure that they don’t overlap
• Recognizing the threshold for overlapping detection Object Detection using Haar cascade classifiers is a useful
• Using appropriate frame sampling rate detection method. It is a ML method in which a cascade
• Keeping min detection count over multiple frames function is taught by providing a lot of true and false images.
• Collecting combined window dimensions for avoiding
overlapping detections For example, when we discuss about face identification.
Initially, the algorithm needs lots of true (pictures of

502

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on December 29,2024 at 13:28:33 UTC from IEEE Xplore. Restrictions apply.
 2021 3rd International Conference on Advances in Computing, Communication Control and Networking (ICACCCN)
appearances) and false (pictures without faces) to train the go. It also finds objects that it might have been lost in the
classifier. middle of the frames

The tracking algorithm depends upon Euclidean distance that

exists in the middle of objects which the tracker has been
trained on before and new objects between consecutive
frames in a video.

The tracking algorithm presumes that when a rectangular

outline is made along the entity, it recognises it as an object.
These rectangular boxes are built using detectors for instance,
solid state drives, haar cascade classifier, etc, using which
they identify and run each frame in a video.
VI. IMPLEMENTATION AND DESCRIPTION OF PROJECT
MODULES:
We start by giving an input image and video, which then gets
converted into black and white, so that processing can be
done faster since there are many screens within fraction of
seconds. After this, the black and white image is converted to
colored when identified as the object and a rectangular box is
Figure 2: HAAR-CASCADE CLASSIFIER created around the object.
The same process is done over and over until the last frame
Presently, all potential dimensions and areas of each kernel is of the video or photo is achived.
used to determine the features.
Module 1(Figure 3,4):
For this, we apply every single component on all the training
pictures. For each feature, it finds the best limit which will
characterize the appearances to positive and negative. There
will be mistakes yet we select the highlights with least
blunder rate, which implies they are the highlights that most
precisely characterize the face and non-face pictures.

V. CAR AND PEDESTRIAN TRACKER:

Intelligent vehicles have the ability to recognize cars and
pedestrians and thus, detect dangerous situations ahead of
time. We have shown how to use ai and ml algorithms and
make a project out of it for beginners using OpenCV and
python. An illustration for beginners.
BASIC TOOLS REQUIRED:
Figure 3: Colored image processing
▪ Visual Studio Code
▪ Open CV
▪ Python
▪ Haar Cascade classifier
▪ XML

OBJECT IDENTIFICATION:
It is defined as procedure:
- Using the primary group of object detection
- Creating a distinctive identity for every primary detection
- Subsequently, tracing all objects when they shift throughout
the frames in clip, while preserving the distinctive identities.

Moreover, tracking permits to use distinctive identity all

tracked object, hence, attainable since the unique entities in a
clip can be accounted for. A good algorithm is fast, has the Figure 4: Colored image output
capacity to manage the object when being tracked and creates
a rectangular outline. It also, detects the primary object in one

503

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on December 29,2024 at 13:28:33 UTC from IEEE Xplore. Restrictions apply.
 2021 3rd International Conference on Advances in Computing, Communication Control and Networking (ICACCCN)
Module 2(Figure 5,6):

Figure 8: Output of outline around the object

Figure 5: Black & White image processing For Example (Code with implementation):

Figure 9: Final Code Snippet

Figure 6: Black & White image output

Module 3(Figure 7,8):

Figure 10(Black & white o/p)

Figure 7: Making rectangles(outline)

504

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on December 29,2024 at 13:28:33 UTC from IEEE Xplore. Restrictions apply.
 2021 3rd International Conference on Advances in Computing, Communication Control and Networking (ICACCCN)
and IoT. Using machine learning and computer vision for
detection and classification of different activities is very
important. Like observing which direction, the driver is
looking when he operates the vehicle, how fast he’s driving,
which direction he’s driving, locating the people that
surround him, etc. Therefore, main goal is to prevent
accidents by increasing efficiency.

REFERENCES:
[1]. [1] Viola, P & Jones, M. (2001). Rapid object detection using a
boosted cascade of simple features. Proceedings of the 2001 IEEE
Computer Society Conference on Computer Vision and Pattern
Recognition (CVPR, 2001), December 8-14, 2001, Kauai, HI, USA.
[2]. [2] Liao, S., Jain, A.K., Li, S. Z. (2016). A fast and accurate
unconstrained face detector. IEEE Transaction of Pattern Analysis
Figure 11: Code Snippet and Machine Intelligence, Vol 38, No 2.
[3]. [3] Luo, D., Wen, G., Li, D., Hu, Y., and Huna, E. (2018). Deep
learning-based face detection using iterative bounding-box
regression. Multimedia Tools Applications.
[4]. [4] Zhang, Y., Wang, X., and Qu, B. (2012). Three-frame difference
algorithm research based on mathematical morphology. Proceedings
of 2012 International Workshop on Information and Electronics
Engineering (IWIEE), pp. 2705 – 2709.
[5]. [5] Canny, J. (1986). A computational approach to edge detection.
IEEE Transactions on Pattern Analysis and Machine Intelligence,
Volume: PAMI-8, No: 6, pp. 679-698, November 1986.
[6]. [6] Li, J. and Ding, S. (2011). A research on improved Canny edge
detection algorithm. Proceedings of the International Conference on
Applied Informatics and Communication, pp. 102 – 108,
Communications in Computer and Information Science (CCIS), Vol
228, Springer-Verlag.
[7]. [7] Lucas, B. D. & Kanade, T. (1981). An iterative image registration
technique with an application to stereo vision.
[8]. [8] Ren, Z., Yang, S., Zou, F., Yang, F., Luan, C., and Li, K. (2017).
Figure 12: Running output A face tracking framework based on convolutional neural networks
and Kalman filter. Proceedings of the 8th IEEE International
VII. ARCHITECTURE DIAGRAM FOR PROPOSED METHOD: Conference on Software Engineering and Services Science, pp. 410-
413.
[9]. [9] Mingxing, J., Junqiang, D., Tao, C., Ning, Y., Yi, J., and Zhen, Z.
(2013). An improved detection algorithm of face with combining
AdaBoost and SVM. Proceedings of the 25th Chinese Control and
Decision Conference, pp. 2459-2463.
[10]. [10] Altun, H., Sinekli, R., Tekbas, U., Karakaya, F. and Peker, M.
(2011). An efficient color detection in RGB space using hierarchical
neural network structure. Proceedings of 2011 International
Symposium on Innovations in Intelligent Systems and Applications,
pp. 154-158, Istanbul, Turkey.

Figure 13. Flow Chart

VIII. CONCLUSION:
SO, artificial intelligence effects the future of each and every
single industry and humans as a whole. It also acts as the main
operator of emerging technologies such as big data, robotics

505

Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY DELHI. Downloaded on December 29,2024 at 13:28:33 UTC from IEEE Xplore. Restrictions apply.