QoS in multimedia application for IoT devices through edge intelligence

Abstract

Abnormal images predicted from video captured in surveillance system is considered as a primary concern in multimedia based IoT (Internet of Things).Conventional techniques comprises of poor quality of service (QoS) while making prediction of images from the videos, in order to overcome these issues proposed method incorporated greedy based genetic algorithm (GGA) for feature selection as traditional methods and Deep CNN for classification. The process involved in building an effective proposed model includes pre-processing, which helps in performing dimensionality reduction in order to minimize the over time of computation. Then the processed images are passed on to the perform object detection with v5 algorithm, which converts videos into images segments and array of images are stored in cloud, after that feature selection proceeded, in which feature selection method with GGA is employed for feature selection in order to select the best features. The key frames which are selected from the image segments of videos assist in the detection of the significant region. Finally, classification process is performed with deep CNN for classification of normal images and abnormal images which are captured from the video. Computational analysis such as time complexity are identified for the proposed framework in addition to it, stability analysis using k-fold cross validation and convergence analysis are detected for the proposed method. Finally performance metrics such as accuracy, precision, F1 score, and recall are implemented in order to evaluate the efficiency of the proposed framework. And various existing algorithms are compared with proposed method as well.