An improved crowd detection system using text and image filtration methods for precautions from COVID-19

Abstract

The text feature is an important descriptive feature of many image analysis applications. Objectives of this study aim to determine the different texture characteristics of them estimation and calculation of population density. In this paper, we have various reviews of the texture and their own have been extensively reviewed, a different combination that is possible to test their effectiveness in crowds of pedestrians. Divide into two categories and retreat. A framework has been proposed to evaluate performance of all aspects of the density coefficient of human density as well to count. According to the framework, the input images are categorized into blocks and blocks into cells of different sizes, having various levels of overcrowding. Because of a distorted view of people, people's visibility near the camera contributes greatly to this feature vector than distant people. Therefore, the features released are usually using a standard visual map of the scene. In the first stage, picture blocks are classified using multiple classes SVM has been at a different level of congestion. In the second phase Gaussian Process Deactivation is used to restore undoing low-level features for calculation. Various texture features and their possible combinations checked on publicly available dataset.