International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 08 | Aug 2024
p-ISSN: 2395-0072
www.irjet.net
Human Suspicious Activity Detection using Machine Learning Joshitha Lakshminarayana1, Bhoomika K2, Keerthana M3, Dr.Vrunda Kusanur4 1Dept. of ECE BNM Institute of Technology Bengaluru, India, 2Dept. of ECE BNM Institute of Technology Bengaluru, India, 3Dept. of ECE BNM Institute of Technology Bengaluru, India,
4Associate Professor, Dept. of Electronics and Communication Engineering, BNM Institute of Technology,
Karnataka India ---------------------------------------------------------------------***--------------------------------------------------------------------Low level features are extracted from the first layer and Abstract--Current urban environments require advanced the following layers will provide these features. Creates a complete feature representation. It solves many problems that arise in traditional analysis. Provides better performance by deep neural networks.
surveillance systems that can identify unusual activity in densely populated regions in order to protect public safety and security. This project uses an advanced multi-model approach to detect anomalous behavior in pedestrians, in order to meet this need. The combination of ResNet50, Inception V3, and the Slow Fast model represents an allencompassing approach that utilizes deep learning to identify minor variations achieving an accuracy of 94%. By improving the effectiveness of surveillance systems in public areas, this project hopes to contribute to the general safety and security by offering an effective way of identifying and handling anomalous behavior.
Research literature has gathered the following point’s supports the use of deep learning in analysis Analyzing human experience activity by observing video is a frequently conducted research in image processing and computer vision operates in its field. Bus stations, train stations, airports, banks, shops, homes, schools and colleges, car parks, roads, etc. through visual monitoring. Human activities in public areas such as can be monitored. Prevent violence, theft, accidents and illegal parking, torture, fighting, theft, crime and other activities no. It is very difficult to constantly monitor public places, so there is a need for intelligent video surveillance that can monitor human activities and distinguish them into normal and abnormal; and can generate an alarm.
Key words—Anomalous behavior, Pedestrians, ResNet50, Inception V3, Slow Fast model, Deep learning, Advanced multimodel approach, Accuracy (94%), Public safety.
1.INTRODUCTION In today’s world, abnormal activity indicates threats and risks to others. An anomaly can be defined as something that deviates from what is expected, common, or normal. Because it is difficult to continuously monitor public spaces and classify them, intelligent video surveillance is necessary. So choosing an appropriate framework for identifying suspicious activity plays a very significant role.
2.SYSTEM ARCHITECTURE 2.1 Inception V3: Introduction: Inception V3 stands as a pinnacle achievement in the realm of convolutional neural networks (CNNs), a product of continuous refinement within the Inception series pioneered by Google. This state-of-the-art architecture is specifically crafted for image classification and object recognition, embodying a synthesis of innovative design principles that elevate its performance in comparison to its predecessors.
Abnormal activity detection, monitoring and identifying abnormal conditions is a very difficult task. In today’s life, the use of security cameras is increasing due to serious crimes. For some the performance is incredible, for others it is not. Therefore, choosing an appropriate framework to identify suspicious activities plays an important role . New monitoring systems use deep learning models to identify malicious programs. Deep learning is very popular and is a type of machine learning.
Architecture Overview: The evolution of Inception V3 is marked by a departure from conventional approaches. The architecture comprises a meticulously designed ensemble of convolutional modules, each tailored for a specific purpose. These modules, including 1x1, 3x3, and 5x5 convolutions, collectively enable the network to capture features at varying scales. Furthermore, the introduction of parallel pathways, often referred to as Inception modules, allows simultaneous feature extraction at different receptive field sizes. This parallelism empowers the model to discern fine details alongside broader contextual information, contributing to its robust recognition capabilities.
It is popular because it can process unnecessary information. Deep learning utilizes and utilizes deep learning neural networks. He has good calculations. It also provides great flexibility when it needs to handle many features. These features are taken from data and do not cause any problems. The deep learning model has many layers through which it passes data. Each layer comes with features derived from non-standard data. This feature is transferred to the next layer of the network.
© 2024, IRJET
|
Impact Factor value: 8.226
|
ISO 9001:2008 Certified Journal
|
Page 524