Live Video Stream Analysis in Real-Time Using Edge Enhanced Clouds

With the rise in streaming data production thanks to the increasing use of Internet of Things (IOT) tools like sensors and video cameras, certain applications such as video surveillance, augmented reality, and autonomous vehicles need quick responses and speedy analysis of data. Right now, they rely on cloud-based setups to store and run machine learning analytics on the information they have. But, this centralized method doesn't fully support analyzing real-time, large-scale streaming data due to issues like network bandwidth and delays between the data source and cloud. In proposed system we have two phases such as filtration and identification. In filtration, all the low value stream objects are filtered using configuration rule which reduces the amount of data. In identification, analytics are performed in the stream of interest using deep learning inference. An algorithm is utilized to group both in-transit and cloud resources strategically in order to meet the quality of service (QOS) constraints set by the user. The study reveals that when dealing with 15k data stream elements at a frame rate ranging from 10 to 100 per second, the time required completing a task in the proposed system is at least 45% quicker and conserves more than 95% of bandwidth compared to a cloud-based approach.