Image and noise reduction for assessing driver incompetence in cases of sudden unintended acceleration

Abstract

This paper explores using cameras aimed at the accelerator and brake pedals during sudden unintended acceleration in cars, removing noise from captured images to determine driver incompetence. A car model was constructed using Raspberry Pi to simulate brake malfunction using random functions, increasing the revolutions per minute (RPM) to simulate sudden acceleration. By employing a DC encoder motor to measure the motor's rotational speed through waveform counts, the RPM was calculated. The study recognized sudden acceleration when the brake malfunctioned through the DC encoder motor, causing an abnormal RPM increase, allowing camera capture toward the accelerator and brake during sudden acceleration events. Precautions were taken for problems arising from noise in captured images. The Unix operating system was utilized to apply Gaussian filter image processing techniques for noise removal. While using an average value filter led to abrupt changes by replacing with the average of surrounding signals, resulting in an unsmooth image, a Gaussian filter was used in this study to decrease weights as distance from the center increased, mitigating these issues.