Efficient Deeplearning Technique to person reidendification in content based video system

• Published in: 2020 International Conference on Computational Intelligence for Smart Power System and Sustainable Energy (CISPSSE) pp. 1 - 6
• Main Authors: Kumar, Kolluru Pavan, Vijaya, K Sri, Jukuntla, Amar
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2020
• Subjects: Cameras; CNN; content-based; Feature extraction; Image retrieval; Lighting; local features; Person reidentification; ResNet50; Safety; Siamese Network; Video surveillance
• DOI: 10.1109/CISPSSE49931.2020.9212271

Person re-identification is very interesting and major difficult in Video surveillance systems which have great impact on safety of public. In spite of broad research endeavors' for quite a long time, it stays one of the most testing open issues that extensively impedes the triumphs of genuine Content Based Image Retrieval frameworks Person re-ID is only issue of distinguishing individuals crosswise over pictures that have been caught by various observation cameras without covering fields of view. With the expanding requirement for computerized video examination, this errand is getting expanding consideration. What's more, it has numerous basic applications, for example, cross camera following, multi-camera conduct investigation and measurable pursuit. In any case, this issue is trying because of the huge varieties of lighting, posture, perspective and foundation. To address these various troubles, in this paper, we propose a few profound learning-based ways to deal with acquire a superior individual re-recognizable proof execution in various manners. We applied ResNet50 and Siamese Network. We use local feature methods to extract the required image from database. We observed the ResNet50 producing great accuracy compare to remaining methods.