Expectation Maximization Method for Effective Detection and Tracking of Object Using Machine Learning Technique for Secure Wireless Communication

Abstract

Object detection is one of the most important computer vision tasks that is used synonymous to object recognition which comprises the mission of identifying the class of the object as well as locating its presence. The main difficulty in the recognition is that there is no single integrated system for detecting the shape or size of objects. Again, the difficulty still persists for the moving object detection due to the varying frames size and speed. Extracting the frames is a more trivial procedure. When frames are extracted the information about the true size, position, orientation of the moving object may be altered. Hence it necessitates to develop an automated object orientation, location identification algorithms to solve the above said problem. To avoid the above circumstances, the proposed work uses the machine learning technique with latest procedures containing different training parameters. This work concentrates on the procedures for the identification of an exact machine learning technique to train the features, which were extracted from the Expectation Maximization (EM) based segmentation process. Before segmentation intensity estimation and clustering procedures were implemented. The image obtained after preprocessing stages is segmented for generating the exemplars and given to a Back Propagation Neural network. Based on the proposed Intensity based clustering and EM based segmentation and feature extraction procedures, the moving object is detected. Hue, Saturation, Gradient Intensity values and GLCM features are extracted and Precision and Recall were used to evaluate the proposed algorithm. Precision was found to be 0.982, Recall is 0.876. accuracy of Back propagation was found to be 96.28%.The experimental results obtained in this proposed work proves the novelty in terms of the computational performance and detection performance.