Please use this identifier to cite or link to this item:
http://hdl.handle.net/10603/546275
Title: | Anomaly Detection in Surveillance Video Using Deep Learning |
Researcher: | PATEL, PRIYANKA |
Guide(s): | NAYAK, AMIT |
Keywords: | Computer Science Computer Science Artificial Intelligence Engineering and Technology |
University: | Charotar University of Science and Technology |
Completed Date: | 2023 |
Abstract: | Anomaly detection in pedestrian pathways is a critical research area that aims to newlineimprove pedestrian safety. However, the traditional manual assessment of newlineidentifying abnormal occurrences in video surveillance systems is becoming newlineincreasingly time-consuming due to the growing volume of captured videos. To newlineovercome this challenge, this research proposes an automated deep learningbased approach using the Time Distributed Spatial Temporal Convolutional Long newlineShort-Term Memory (TDSTConvLSTM) model to recognize and classify different newlineabnormalities in pedestrian walkways. newlineThe TDSTConvLSTM is a deep learning-based approach that has shown newlineexcellent performance in computer vision processes, such as object classification newlineand object detection. In this study, the model is used to detect anomalies in newlinepedestrian pathways, including autos, skates, jeeps, cars, cycles, tempos, and newlineother vehicles. The TDSTConvLSTM model includes preprocessing as the first newlinestep, which is used to remove noise and resize frames to improve picture quality. newlineThe regression model is then used for the detection of errors and cost newlinereconstruction, enabling the detection of anomalies with greater accuracy and newlineefficiency. The classification model is used for the detection of the type of newlineanomaly. newlineComprehensive simulations are conducted under experimental settings to newlineevaluate the performance of the TDSTConvLSTM model to detect anomalies in newlinepedestrian walkways for campus datasets like UCSD PED-1, UCSD Ped-2, newlineShanghaiTech Campus, and CHARUSAT Gate-1, and CHARUSAT Gate-2. The newlinevi newlineresults of the simulations demonstrate exceptional performance and maximum newlinedetection accuracy, indicating that the proposed TDSTConvLSTM approach is a newlinepromising solution for improving pedestrian safety in various scenarios. newlineThis study proposes a deep learning-based approach using the TDSTConvLSTM newlinemodel for anomaly detection in pedestrian walkways, which can improve newlinepedestrian safety by detecting abnormal occurrences in real time. newlineThe proposed TDSTConvLSTM approach has several advantages, i |
Pagination: | |
URI: | http://hdl.handle.net/10603/546275 |
Appears in Departments: | Faculty of Technology and Engineering |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
01_title.pdf | Attached File | 29.75 kB | Adobe PDF | View/Open |
02_prelim pages.pdf | 215.53 kB | Adobe PDF | View/Open | |
03_content.pdf | 193.28 kB | Adobe PDF | View/Open | |
04_abstract.pdf | 8.24 kB | Adobe PDF | View/Open | |
05_chapter 1.pdf | 63.39 kB | Adobe PDF | View/Open | |
06_chapter 2.pdf | 460.93 kB | Adobe PDF | View/Open | |
07_chapter 3.pdf | 1.24 MB | Adobe PDF | View/Open | |
08_chapter 4.pdf | 1.26 MB | Adobe PDF | View/Open | |
09_chapter 5.pdf | 11.68 kB | Adobe PDF | View/Open | |
10_annexures.pdf | 96.38 kB | Adobe PDF | View/Open | |
11_chapter 6.pdf | 10.06 kB | Adobe PDF | View/Open | |
80_recommendation.pdf | 48.76 kB | Adobe PDF | View/Open |
Items in Shodhganga are licensed under Creative Commons Licence Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0).
Altmetric Badge: