Wireless Violence Detection

Student Classification

Jordan Holeman, Senior, Computer Science

Faculty Mentor

Xiaohong Yuan, Computer Science

Department

Computer Science

Document Type

Poster

Publication Date

Spring 2023

Abstract

Computer vision continues to improve its ability to detect human activity. However, detecting violent scenes intelligently in video surveillance has been a challenge to implement effectively. AI-Assisted Edge Vision has been used to detect violent videos in IoT-Based Industrial Surveillance Networks. Convolution long short-term memory (ConcLSTM) model is used in this detection system. Video frames with objects such as weapons are forwarded to cloud for a detailed investigation by extracting specific features using ConcLSTM. Literature shows that there are many techniques that have been used for computer vision. For example, Lightweight Deep Learning Based Intelligent Edge Surveillance, Multi-manifold Positive and Unlabeled Learning, Object Detection Binary Classifiers, Detecting Anomalies in Individual Trajectories, Attention-based Encoder-Decoder Networks, Low Level features, MoBSIFT and Movement Filtering Algorithm, and Optical Flow Classification. These methods are related to the implementation of our system. For instance, the Lightweight Deep Learning Based Intelligent Edge Surveillance method gave us useful information on a step to screen the frames in both indoor and outdoor scenes. This poster introduces these approaches that were investigated for computer vision.

This document is currently not available here.

Share

COinS