Stabilization of environmental conditions to improve the performance of a mobile application for the state of the driver monitoring

Abstract

The use of modern digital technologies in the field of security enhancement is a promising area of research. The results of a study in the field of increasing the level of human security in the case of monotonous, routine work using neural networks were reported at the AIS2021 conference. It has been developed a model and an algorithm for detecting a dangerous state of the driver (wakeful or sleeping). This work has been continued. The conditions of the environment in which the developed software is intended to be used depend on the smartphone used and how the smartphone is mounted in the vehicle. It should be noted that in a real environment, smartphones work with not always complete and accurate data. These factors directly affect the performance of the mobile app by changing the speed, accuracy, and functionality of the mobile app. This paper presents algorithms for manual and automatic calibration of the system based on data from the camera, sensors and settings of the driver's smartphone. It is proposed to use a mandatory calibration method that adapts to the current context of the driver and vehicle in order to reduce errors in various calculations of physical quantities numerical values and ensure the uniformity of measurements during the operation of the system. The calibration takes into account the input information about the driver, vehicle and smartphone. The processing of the critical area and the state of the driver in the image were implemented in the C ++ programming language using the computer vision software libraries OpenCV and Dlib, and the JNI interfaces were realized in the Java language.