Detection of water-risk zones based on color and texture in home video surveillance

Abstract

Automatic detection of risk areas is a requirement of home and outdoor video surveillance, it is used as a tool for prevention and warning in case of imminent danger. Swimming pools, for example, are a real danger for children. Accidental drownings are the second leading cause of death among young people under the age of 15 just after traffic accidents.The purpose of this paper is to automate the detection of the aquatic risk zone by segmenting it with precision. An approach based on the gathering of two important indices that characterize water surfaces, namely color and texture, is proposed. Chromatic detection is based on experimentally defined color and brightness values in the HSV color space, while textural detection is based on the local properties of the shape of the water surface by calculating the variance and applying the local binary pattern method.In order to validate our approach, the experiments were applied on a proprietary database that contains a variety of images containing swimming pools in daytime.