자율주행 카메라 인지성능에 작은 오염이 미치는 영향 연구

Abstract

The environmental awareness sensors of self-driving cars include cameras, radars, and lidar. While various cognitive sensors are employed in combinations as per self-driving manufacturers' unique development strategies, the camera is a consistently utilized sensor. Camera sensors are the only ones capable of capturing texture, color, and contrast information, as well as recognizing objects such as road lanes, signals, signs, pedestrians, bicycles, and surrounding vehicles. Due to the ever-increasing pixel resolution and relatively low prices, camera sensors are gaining importance in autonomous vehicles. However, they are susceptible to environmental changes like dust, sun, rain, snow, or darkness. Furthermore, due to their relatively small, lens-like shape in comparison to radar and lidar sensors, camera sensor performance can be compromised by visual obstructions such as small dust particles (referred to as ‘blockage’ hereafter), which can significantly impact the safety of autonomous driving. In this study, a camera simulator was employed to simulate a virtual accident scenario based on an actual accident, projecting the virtual scenario screen directly through a camera. Object recognition delay time was assessed based on the density and color of the blockage. This study demonstrates that cognitive delays caused by blockage can lead to major accidents and draws a parallel with the idea that a small speck of dust in one's faith can cause significant trials. In addition to emphasizing the importance of cleaning the camera lens to prevent blockage, I would like to suggest the periodic purification of one's faith from external temptations.