Development of an anomaly recognition system for video surveillance systems inside an educational institution using deep learning algorithm / Christian Paul R. Altar, Christian Pol C. Macuja, Adriane Vergel C. Riola, Lorenzo John D. Semilla.

By:

Altar, Christian Paul R

Contributor(s):

Material type: Text

TextManila : Technological University of the Philippines, 2022Description: xi, 94 pages : illustrations ; 28 cm. + 1 CD-ROM (4 3/4 in.)Subject(s):

LOC classification:

BTH QA 76 A48 2022

Dissertation note: College of Science -- Bachelor of Science in Computer Science, Technological University of the Philippines, 2022. Abstract: Surveillance cameras are being set up in establishments, one is in educational institutions, to monitor areas from any abnormal events that might happen at the said location. However, these abnormal events need to be reviewed manually by personnel to determine if an abnormal event has happened, thus spending more time in viewing a lengthy video. An Anomaly Detection System, that will be developed through Python and with Deep Learning algorithm, aids in detecting abnormal events in videos that the surveillance system has recorded. Python is used in making the system for usable modules in video and image manipulation like OpenCV, Keras, and TensorFlow. Python also provides modules for data treatment like NumPy and Scikit. With these, training the system can be made more accurate with higher computer specifications, specifically in CPU, Graphics Processing Unit, and Random Access Memory usage. This system enhances security measures in a way that this will make anomaly detection faster for videos recorded by surveillance systems. A spatiotemporal model consisting of convolutional layers and convolutional LSTM layers can showcase a considerably good score in correctly predicting normal events from anomalous events. Alongside this, the system yielded an accuracy of at most 75% in correctly predicting abnormal events and normal events from test dataset. Furthermore, the results revealed that the system is highly rated in usability as the mean rating is 3.7 which is considered highly acceptable, and the lowest rate the system obtained is 3.52. A system deemed as highly acceptable in usability can achieve specified goals with efficiency in a designated utilization. --Author's Abstract

Tags from this library: No tags from this library for this title. Log in to add tags.

Average rating: 0.0 (0 votes)

Holdings
Item type	Current library	Shelving location	Call number	Copy number	Status	Date due	Barcode
Bachelor's Thesis COS	TUP Manila Library	Thesis Section-2nd floor	BTH QA 76 A48 2022 (Browse shelf(Opens below))	c.1	Not for loan		BTH0003448

Thesis (Undergraduate)

College of Science -- Bachelor of Science in Computer Science, Technological University of the Philippines, 2022.

Includes bibliographical references.

Surveillance cameras are being set up in establishments, one is in educational institutions, to monitor areas from any abnormal events that might happen at the said location. However, these abnormal events need to be reviewed manually by personnel to determine if an abnormal event has happened, thus spending more time in viewing a lengthy video. An Anomaly Detection System, that will be developed through Python and with Deep Learning algorithm, aids in detecting abnormal events in videos that the surveillance system has recorded. Python is used in making the system for usable modules in video and image manipulation like OpenCV, Keras, and TensorFlow. Python also provides modules for data treatment like NumPy and Scikit. With these, training the system can be made more accurate with higher computer specifications, specifically in CPU, Graphics Processing Unit, and Random Access Memory usage. This system enhances security measures in a way that this will make anomaly detection faster for videos recorded by surveillance systems. A spatiotemporal model consisting of convolutional layers and convolutional LSTM layers can showcase a considerably good score in correctly predicting normal events from anomalous events. Alongside this, the system yielded an accuracy of at most 75% in correctly predicting abnormal events and normal events from test dataset. Furthermore, the results revealed that the system is highly rated in usability as the mean rating is 3.7 which is considered highly acceptable, and the lowest rate the system obtained is 3.52. A system deemed as highly acceptable in usability can achieve specified goals with efficiency in a designated utilization. --Author's Abstract

There are no comments on this title.

to post a comment.