Detection and Classification of Objects in Video Content Analysis Using Ensemble Convolutional Neural Network Model

Abstract

Video content analysis (VCA) is the process of analyzing the contents in the video for various applications. Video classification and content analysis are two of the most difficult challenges that computer vision researchers must solve. Object detection plays an important role in the VCA and is used for identification, detection and classification of objects in the images. The Chaser Prairie Wolf optimization-based deep Convolutional Neural Network classifier (CPW opt-deep CNN classifier) is used in this research to identify and classify the objects in the videos. The deep CNN classifier correctly detected the objects in the video, and the CPW optimization boosted the deep CNN classifier’s performance, where the decision-making behavior of the chasers is enhanced by the sharing nature of the prairie wolves. The classifier’s parameters were successfully tuned by the enabled optimization, which also aids in producing better results. The Ensemble model developed for the object detection adds value to the research and is initiated by the standard hybridization of the YOLOv4 and Resnet 101 model, which evaluated the research’s accuracy, sensitivity, and specificity, improving its efficacy. The proposed CPW opt-deep CNN classifier attained the values of 89.74%, 89.50%, and 89.19% while classifying objects in dataset 1, 91.66%, 86.01%, and 91.52% while classifying objects in dataset 2, compared to the preceding method that is efficient.