ANN-Based Scalable Video Encoding Method for Crime Surveillance-Intelligence of Things Applications

The proliferation of high-quality, networked CCTV cameras in public areas has raised the volume of constantly generated video data significantly in recent years. The need for scalable, distributed video processing infrastructure is growing as a result. In order to make unstructured videos supplied to the platform searchable and comprehensible by both humans and machines, an intelligent video analytics platform runs them through a series of trans-disciplinary algorithms. The use of video analytics extends well beyond just surveillance to include things like video asset management. Many different commercial and academic approaches exist here. Most current systems for face and object recognition, however, have a time-honored client-server architecture that leaves out support for increasingly intricate use cases. Furthermore, distributed computing is seldom used to manage such frameworks at a scale. Furthermore, no current works provide any kind of assistance with low-level networked video processing APIs. Not only did they ignore the expanding needs of customers, academics, and developers, but they also neglected to address a comprehensive service-oriented environment. In this work, we offer an Artificial Neural Network based Scalable Video Coding (ANN-SVC) framework for intelligent video surveillance as a means of resolving these problems. The suggested system can handle both live video feeds and video analytics in batches. Similarly, the batch processing data for each realtime stream is available. This work therefore has a connection to the idea of symmetry. We also provide Spark's first distributed video processing library. To guarantee scalability, efficacy, and fault-tolerance, SIAT takes use of cutting-edge distributed computing technology. Finally, in an effort to back up our assertions, we implement and test our suggested framework. The simulation shows better results in throughput, gooodput and PSNR.