RETRACTED: Design and Development of Person Re-Identification system based spatial features

Abstract

Primary intention of this research is to design and develop an effectual approach for re-identifying the persons. The proposed approach involves four stages named human identification, object tracking, feature extraction, and matching. The four stages will be implemented by matching the probe videos and gallery videos. Initially, the videos will be subjected to the human identification phase. First stage will be the identification of human depending on the spatial features. It is important to consider that person re-identification will be carried out using the video surveillance data which can be probe data or gallery data. For human identification, the spatial features will be extracted and subjected to gait based Bayesian network for identifying the humans. Then, the object tracking phase will be carried out for tracking the objects from the videos. Here, the Neighborhood Search Algorithm will be used for tracking the objects. Then, the tracked objects will be given to the feature extraction process in which the features like mean, variance, speed, and motion deviations will be extracted. Finally, in matching phase, the matching will be performed between the probe videos and gallery videos and the person re-identification will be done using the extracted features based on the Bayesian probability-based matching. Finally, the matched outputs will be utilized for re-identifying the persons. The implementation of the proposed approach will be done using the Grid and Caviar Data sets. The performance of the proposed technique will be evaluated using three metrics, such as accuracy, sensitivity and specificity and will be compared with that of existing methods.