Suspicious Activity Detection and Alert System Using Raspberry Pi
Suspicious Activity Detection and Alert System Using Raspberry Pi
Volume 8, Issue 4, April 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Suspicious Activity Detection and Alert
System using Raspberry Pi
K. Lalitha [1], K.P. Dhivyesh Anand[2], S. Mukundhan[3]
Asst Professor1, Student2,3
Dept. of Information Technology,
Panimalar Engineering College, Ponnamalle, Chennai, Tamil Nadu, India
Abstract:- An important branch of computer science II. PROBLEM STATEMENT
called computer vision processes digital photos and
movies to increase the intelligence of machines. Activity The present CCTV system only helps to identify criminals
recognition, which automatically classifies an agent's after the fact, making it ineffective at spotting suspicious
behavior, is a significant use of computer vision. activity before a crime takes place.
Uncovering a person's actions through a series of The current system depends on recording CCTV data,
observations while considering many difficult contextual making it labour- and time-intensive to continuously
conditions is the goal of the key application of suspicious monitor the data manually.
activity recognition. The internet tells the user or
commando when the surveillance system's motion III. OBJECTIVE
detection and tracking are activated.
To convert video to image.
I. INTRODUCTION To frame segmentation using K means clustering.
To extract frame by background subtraction and frame
Due to recent rapid technology breakthroughs, sequence.
automated human activity identification has gained To detection the object.
popularity as a research area. Applications for video To action recognition using Deep Belief Network (DBN).
surveillance include spotting regular and unusual activity, To classify normal activity or suspicious activity with
such as gaming, human-computer interaction, exam trained dataset.
monitoring, chaos detection, sports analysis, crowd To alert a security.
behaviour forecasting, and more. It is important for
guaranteeing safety both inside and outside of the border. IV. EXISTING SYSTEM
The volume of video data that needs to be analysed The term "object detection" describes a technique for
and the rapid pace of invention make manual intervention spotting real-world things like cars, bikes, TVs, flowers, and
not only impossible but also error-prone. Also, maintaining people in pictures and videos. This method facilitates the
regular surveillance of critical areas is a very difficult identification, localization, and detection of numerous
endeavour. Deploying intelligent video surveillance systems objects existing in an image, allowing for a thorough
that can track people's movements in real-time, classify comprehension of an image or video. In essence, it enables
them as regular or odd behaviours, and produce alerts the identification of distinctive things in a visual medium.
accordingly is therefore essential.
It is usually utilized in applications like image
Radar, cameras, and cell phones are just a few of the retrieval, security, surveillance, and advanced driver
sensors used in a variety of applications, including security, assistance systems (ADAS). Object Detection is done
surveillance, human-computer interface, and monitoring through many ways:
suspicious activity. Nowadays, the majority of systems use Feature Based Object Detection
CCTV cameras to record video material that can be used in Viola Jones Object Detection
criminal or violent act investigations. It would be preferable SVM Classifications with HOG Features
to have a system that can proactively recognise anomalous Deep Learning Object Detection
or unexpected occurrences and warn the authorities as soon
as they occur.
IJISRT23APR070 www.ijisrt.com 28
Volume 8, Issue 4, April 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Fig. 1: Advanced driver assistance systems (ADAS)
V. PROPOSED SYSTEM Additionally, the model is frequently trained using methods
from machine learning, deep learning, and data analytics.
Depending on the system components used, activity
recognition can be divided into two categories: sensor-based In contrast, vision-based activity recognition uses a
and vision-based activity recognition. Using a network of camera-based system to examine and process video data in
sensors, sensor-based activity recognition keeps track of order to find and classify behaviours within a specific
both the actions of a person and outside factors. Important environment. The video, which is basically a collection of
information can be extracted from the data collected by images, is often processed using digital image processing
these devices after it has been processed and aggregated. techniques to extract pertinent information.For this
particular project, a vision-based system was utilized.
VI. BLOCK DIAGRAM
SOFTWARE SIDE:
HARDWARE SIDE:
Fig. 2: Block Diagram Explation
IJISRT23APR070 www.ijisrt.com 29
Volume 8, Issue 4, April 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
VII. BLOCK DIAGRAM EXPLANATION VIII. MODULE DESCRIPTION
Using Python and the OpenCV library, a Raspberry Pi The system's software converts the videos into frame
4 and Raspberry Pi camera have been used to create a house pictures using data from the camera. Following that, these
security surveillance system. The system was tried by using pictures are preprocessed using image processing methods.
the "raspistill -o image.jpg" command to take a picture after The data is stored in memory and the software has been
configuring the Raspberry Pi 4 with the Raspbian OS and trained using images. The system sends an alert message if it
connecting the camera.In order to increase functionality, discovers any suspicious behaviour.
extra libraries and dependencies were installed after the
coding was finished in the terminal. In order to allow email The Raspberry Pi receives information from the
notifications with an attached image when motion is camera via the USB port in the hardware component. The
detected, the mail.py file was configured with the user's and user receives a text message via email if the system notices
system's email addresses. The system constantly watches for any suspicious behaviour.
movement and emails the user's email address an alert with
the image it has just taken. This project was successful in A. HARDWARE REQUIREMENTS
giving users an easy method to spot activity and take the Raspberry pi
necessary action. Camera
B. SOFTWARE REQUIREMENTS
Raspberry pi OS
Python IDE
OpenCV library
IX. OUTPUT RESULTS
The confidence curve and other result images are mentioned below.
Fig. 3: Confusion Matrix
IJISRT23APR070 www.ijisrt.com 30
Volume 8, Issue 4, April 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Fig. 4: F1 Confidence curve
Fig. 5: Loss Graph
X. CONCLUSION
This model has produced results that are correct. Anorange frame is created to emphasize the area of motion when the
camera detects motion. Quickly after being taken, a picture is sent to the user via Outlook. The USB Raspberry Pi camera used for
this project provides clear images that make it simple to spot any suspicious activity. A sample output is shown in the below fig4.
Fig. 6: The expected classification of output
IJISRT23APR070 www.ijisrt.com 31
Volume 8, Issue 4, April 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
REFERENCES [6.] T. Huynh-The, B. V. Le, S. Lee, and Y. Yoon,
“Interactive activity recognition using pose-based
[1.] J. K. Aggarwal and M. S. Ryoo, “Human activity spatio–temporal relation features and four-level
analysis: A review,” ACM Comput. Surv., vol. 43, Pachinko Allocation Model,” Inf. Sci. (Ny)., vol.
no. 3, 2011, doi: 10.1145/1922649.1922653. 369, pp. 317–333, 2016, doi:
[2.] A. G. D’Sa and B. G. Prasad, “A survey on vision 10.1016/j.ins.2016.06.016.
based activity recognition, its applications and [7.] S. Abdelhedi, A. Wali, and A. M. Alimi, “Fuzzy
challenges,” 2019 2nd Int. Conf. Adv. Comput. logic based human activity recognition in video
Commun. Paradig. ICACCP 2019, pp. 1–8, 2019, surveillance applications,” Adv. Intell. Syst.
doi: 10.1109/ICACCP.2019.8882896. Comput., vol. 427, pp. 227–235, 2016, doi:
[3.] G. Cheng, Y. Wan, A. N. Saudagar, K. Namuduri, 10.1007/978-3-319-29504-6_23.
and B. P. Buckles, “Advances in Human Action [8.] P. Guo, Z. Miao, Y. Shen, W. Xu, and D. Zhang,
Recognition: A Survey,” no. February, 2015, “Continuous human action recognition in real
[Online]. Available: time,” Multimed. Tools Appl., vol. 68, no. 3, pp.
http://arxiv.org/abs/1501.05964. 827–844, 2014, doi: 10.1007/s11042-012-1084-2.
[4.] C. Dhiman and D. K. Vishwakarma, “A review of [9.] A. Jalal, M. Uddin, and T. S. Kim, “Depth video-
state-ofthe-art techniques for abnormal human based human activity recognition system using
activity recognition,” Eng. Appl. Artif. Intell., vol. translation and scaling invariant features for life
77, no. August 2018, pp. 21–45, 2019, doi: logging at smart home,” IEEE Trans. Consum.
10.1016/j.engappai.2018.08.014. Electron., vol. 58, no. 3, pp. 863– 871, 2012, doi:
[5.] S. A. R. Abu-Bakar, “Advances in human action 10.1109/TCE.2012.6311329.
recognition: An updated survey,” IET Image [10.] J. Hu and N. V. Boulgouris, “Fast human activity
Process., vol. 13, no. 13, pp. 2381–2394, 2019, recognition based on structure and motion,” Pattern
doi: 10.1049/ietipr.2019.0350. Recognit. Lett., vol. 32, no. 14, pp. 1814–1821,
2011, doi: 10.1016/j.patrec.2011.07.013.
IJISRT23APR070 www.ijisrt.com 32
