CONTENT ONLY PPT

Part division
Shathyahan - Intro, aim, objectives, weapon det, access det, data processing tech,
conclusion, further studies, References

Pasindu - obj det tech, fire and smoke det, aband obj det, methodology,
management
Table of Contents
Background and Motivation
● Digital surveillance systems are gaining popularity due to their diverse applications, but manual
intervention is still needed for many tasks.
● Automation of surveillance using computer vision is an active area of research and in high demand.
● However, most existing methods only focus on one application, leaving other tasks still reliant on
human observers.
● Additionally, emergency management is still entirely operated manually.
● The goal of this project is to develop an automated surveillance system for galleria environments.
● The system will provide hazard alerts and navigational guidance assistance during emergencies.
● The six main areas of focus for the system are weapon detection, fire and smoke detection,
abandoned object detection, person detection, identification, and authorization, alerting, and
guidance.
● The project aims to create a converged system that integrates all six areas of focus.
Aim and Objectives
● The project aims to develop an automated surveillance system for galleria environments to
improve security and safety.
● The system will be able to detect weapons, fire and smoke, abandoned objects, and
unauthorized access.
● The system will provide hazard alerts and notify relevant authorities during emergencies.
● The system will also provide navigational guidance assistance to people during hazardous
situations.
● The overall aim will be achieved through the following objectives;
○ Implementing a CCTV surveillance camera system and getting video feeds into the edge computing system
○ Creating a dataset to train the detection model
○ Developing a multi-class object detection model using YOLO algorithm and CNNs
○ Developing required UIs for the system
○ Developing an alert system and visualization system.
○ Testing and improving the accuracy of the system
Literature Review
Literature Review - Object detection techniques
Object detection techniques

Supervised image classification Unsupervised image classification

Traditional methods Deep learning methods

CNN DNN RNN

RCNN FCNN Yolo

Literature Review - Fire and smoke detection techniques
Papers Used:
1. Real time video fire / smoke detection based on CNN in antilife surveillance system by Sergio Saponara1,
Abdussalam Elhanashi1, Alessio Gagliardi1
2. Real time fire and smoke detection by Komal Mahesh Dhotre, Vaishnavi Ganesh Choudhari, Neelam Kalidas
Gali,

Detailed overview:
Paper No. Methodology Limitations

1 ● YOLOv2 ● Had to used GPU

● Used Deep neural network designer tool to build network layers based embedded
● Used a data set with 400 images, label with Ground Truth unit since YOLOv2
Application and trained on MatLab. is based on GPU
● Implemented on NVIDIA Jetson Nano. calculation.
● Validated the accuracy using 200 images and accuracy was ● Could handle only
93%. 1 camera input.

2 ● Yolov3 ● Lack of control

● Used Tenserflow framework with Keras libraries. due to usage of
frameworks.
Methodology
Methodology - workflow
 Methodology - Data processing architecture Cloud layer

Each layer will have a

assigned color to its
Edge layer devices and parts
(Not device parts like
a) will be faded into
background
Device layer

a
Methodology - YOLO algorithm

Here CNN model and

“a” component will be
emphasize and
others will be faded to
background.

a
Management
Project Budget

Item Description Quantity Cost Total

Cloud Google Cloud SQL for 6 months and 1 15,500 15,500

storage 100GB per month (0.08USD * 100 *
6 = 48 USD => 15500 LKR)

Web cam 2 Webcams to test parallel processing 2 5,000 10,000

capabilities.

Total 25,500
 Color 01 - assigned to shathya

Workload division color 02 - assigned to pasindu

Color 03 - assigned to both

Further Business
GUIs
Databases CNN analysis logic
algos

Fire, smoke Weapon,

,aban objs authority
Training Developing related relted

Not the final design

Project Timeline

Design will be changed

later to be matched with
the theme.
Conclusion
● The project focuses on providing a complete surveillance and navigational
guidance system for Galleria environments.
● The system will be able to detect people, suspicious objects, fire outbreaks,
and unauthorized access.
● The system will provide emergency management to inform the appropriate
authorities and guide the public.
● The accuracy for detecting the above-mentioned criteria is more than 80%
based on past research work using identified techniques.
● The proposed system will be more efficient and effective than existing
systems by supporting full automation in both surveillance and emergency
management operations.
Further Studies
● The study focuses on several areas such as weapon detection, fire and
smoke detection, abandoned object detection, person identification for
authorization, alerting, and guidance.
● Further studies will be conducted to work on human behavior analysis to
detect full face cover, theft, and pickpocketing.
● The ultimate goal is to enhance security measures through the development
of advanced detection and identification technologies.
References
1. Domadia, S. G., & Zaveri, T. (2011, May). “Comparative analysis of unsupervised and supervised image classification techniques”. 2011 National
Conference on Recent Trends in Engineering & Technology pp. 1-5, 2011
2. “Decoding the dichotomy: Image processing vs deep learning,” HCLTech. [Online]. Available:
https://www.hcltech.com/white-papers/engineering/decoding-dichotomy-image-processing-vsdeep-learning. [Accessed: 22-Mar-2023].
3. J. Ahmad, H. Farman, and Z. Jan, “Deep Learning Methods and Applications,” Deep Learning: Convergence to Big Data Analytics, SpringerBriefs in
Computer Science, 2019.
4. J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once: Unified, real-time object detection,” 2016 IEEE Conference on Computer Vision
and Pattern Recognition (CVPR), 2016.
5. V. Magare, A. V. Raut, G. B. Bandgar , V. D. Ghule, and vrushali kondhalkar, “FIRE & SMOKE DETECTION USING YOLO,” Journal of Analysis and
Computation (JAC), vol. 15, no. 12,Dec. 2021.
6. K. Dhotre, V. Choudhari, N. Gali, P. Sandip, Student, and S. Chavan, ‘REAL-TIME FIRE AND SMOKE DETECTION’, 04 2022.
7. M. E. Atik, Z. Duran, and R. Özgünlük, ‘Comparison of YOLO versions for object detection from aerial images’, International Journal of Environment and
Geoinformatics, vol. 9, no. 2, pp. 87–93, 2022.
8. H. Ashraf et al., ‘Weapons Detection for Security and Video Surveillance Using CNN and YOLO-V5s’, Computers, Materials and Continua, vol. 70, pp.
2761–2775, 01 2022.
9. [9] Su Pyae Lwin, Myat Thida Tun, “Deep convolutional neural network for abandoned object detection,” International Research Journal of Modernization in
Engineering Technology and Science, 2022.vol.:04 /Issue:01/January-2022
10. T. Santad, P. Silapasupphakornwong, W. Choensawat, and K. Sookhanaphibarn, “Application of yolo deep learning model for real time abandoned baggage
detection,” 2018 IEEE 7th Global Conference on Consumer Electronics (GCCE), 2018.
11. [11] V. Jain, Y. Jain, H. Dhingra, D. Saini, M.C. Taplamacioglu, M. Saka , “A systematic literature review on qr code detection and pre-processing,” Medium,
March-2021. In International Journal on “Technical and Physical Problems of Engineering (pp. 111-119)
12. D. K. Hansen, K. Nasrollahi, C. B. Rasmusen, and T. B. Moeslund, “Real-time barcode detection and classification using Deep Learning,” Proceedings of
the 9th International Joint Conference on Computational Intelligence, 2017.
13. G. S. Sriram, ‘Cloud computing: an overview of big data challenges and opportunities for large enterprises’, International Research Journal of Modernization
in Engineering Technology and Science, vol. 4, no. 1, pp. 1331–1337, 2022.