Novel Framework for the Improvement of Object Detection Accuracy of Smart Surveillance Camera Visuals using Modified Convolutional Neural Network Technique compared with Support Vector Machine

Abstract

The aim of this research work is to appraise the accuracy of Support Vector Machine (SVM) and Modified Convolutional Neural Network Technique (MCNNT) by replacing the hierarchical data processing for Smart Surveillance System. Materials and Methods: With MCNNT our novel object detection framework utilizes hierarchical data models of data processing, it is made up of layers that are completely interconnected to each node to control the complexities of object detection and using model image dataset calculated the data with sample size of 20 per group using p-value as 0.05. Result: The acquired mean accuracy of MCNNT (96.16%) obtained greater than SVM (94.40%). There is statistically significant deviation between obtained accuracies of two algorithms and for confidence interval (CI) 95% independent sample test was performed. Conclusion: Based on obtained results MCNNT acquired better accuracy than SVM of object detection.