DETECTION OF VIOLENCE IN FOOTBALL STADIUM THROUGH BIG DATA FRAMEWORK AND DEEP LEARNING APPROACH
Journal Title: International Journal of Data Science and Artificial Intelligence - Year 2023, Vol 1, Issue 02
Abstract
Football is the most famous game in the world, with over 4 billion supporters worldwide. Football hooliganism refers to the aggressive or destructive actions of a supporter or player in a stadium while watching or participating in a game. To avoid violence, a real-time violence detection system is required to observe the audience and players behaviour in order to take appropriate action before violence occurs. The input of the system is a massive volume of real-time video feeds from various sources, that are processed using the Flink structure. Using the Histogram of Oriented Gradients (HOG) function in the Flink framework, pictures are partitioned into frames and their characteristics are retrieved. The frames are then labelled on the basis of attributes including Groundside-violence model, Crowdside-violence model, human part model, and Non-violence model, are utilised to train the multihead attention based Bidirectional Long Short-Term Memory network for violent scene detection. The RWF-2000 dataset, which contains the training set (80%) and the test set (20%) was used to train the network and also a dataset comprising 410 video footages with non-violence scenes and 409 video footages with violent situations is created by the videos obtained from a football stadium, to make the algorithm more strong to violence detection. Other existing approaches are used to validate the model's performance. When compared with existing systems, the proposed violence detection methodology significantly increases accuracy upto 1.6453%, precision upto 0.646%, recall upto1.959%, and reduces execution time upto 60% than other existing methods.
Authors and Affiliations
M. Dhipa and D. Anitha
Keywords
Related Articles
- EP ID EP742456
- DOI -
- Views 18
- Downloads 0
SELECTIVE FORWARDING ATTACKS DETECTION IN WIRELESS SENSOR NETWORKS USING BLUE MONKEY OPTIMIZED GHOST NETWORK
Wireless Sensor Networks (WSNs) are increasingly the technology of choice due to their wide applicability in both military and civilian domains. The selective forwarding attack, one of the main attacks in WSNs, is the ha...
EFFICIENT DATA SEARCH AND RETRIEVAL IN CLOUD ASSISTED IOT ENVIRONMENT
Internet of Things (IoT) is expanding across a number of industries, including the medical field. Such a scenario might easily reveal sensitive information, such as private digital medical records, presenting potential s...
IOT-ENABLED PROTEIN STRUCTURE CLASSIFICATION VIA CSA-PSO BASED CD4.5 CLASSIFIER
Data mining is a technique for obtaining useful information from vast amounts of information. Big data refers to large amounts of complicated information that is processed, particularly in relation to biological processe...
DYNAMIC LOAD BALANCING IN CLOUD COMPUTING USING HYBRID KOOKABURRA-PELICAN OPTIMIZATION ALGORITHMS
Cloud Computing (CC) technology facilitates virtualized computer resources to users via service providers. Load balancing assumes a critical role in distributing dynamic workloads across cloud systems, ensuring equitable...
HYBRID OPTIMIZATION INTEGRATED INTRUSION DETECTION SYSTEM IN WSN USING ELMAN NETWORK
Wireless Sensor Networks (WSNs) increases the usage of integrated systems and areas which attracts the attention of attackers. However, WSNs are vulnerable to different kinds of security threats and attacks. To ensure th...