A Novel Video Analytics Framework for Microscopic Tracking of Microbes

Abstract

Micro-organisms or microbes are single- or multi-cellular living organisms viewed under a microscope because they are too tiny to be seen with naked eyes. Tracking them is important as they play a vital role in our lives in terms of breaking down substances, production of medicines, etc., as well as causing several diseases like malaria, tuberculosis, etc., which need to be taken care of. For a pathological study, the images of these microbes are captured from the microscope and image processing is done for further analysis. These operations involved for the analysis requires skilled technicians for error-free results. When the number of images increases, it becomes cumbersome for those technicians as there is a chance of ambiguity in results, which hampers the sensitivity of the study. Further, image processing is a bit challenging and time-consuming as a single image provides only a snapshot of the scene. In this situation, video has come into the picture which works on different frames taken over time making it possible to capture motion in the images keeping track of the changes temporally. Video combines a sequence of images, and the capability of automatically analyzing video to determine temporal events is known as video analytics. The aim of this paper is to develop a new computing paradigm for video analytics which will be helpful for the comprehensive understanding of the microbial data context in the form of video files along with effective management of that data with less human intervention. Since video processing requires more processing speed, a scalable cluster computing framework is also set up to improve the sensitivity and scalability for detecting microbes in a video. The HDP, an open source data processing platform for scalable data management, is used to set up the cluster by combining a group of computers or nodes. Apache Spark, a powerful and fast data processing tool is used for the analysis of these video files along with OpenCV libraries in an efficient manner which is monitored with a Web UI known as Apache Ambari for keeping in track all the nodes in the cluster.