The Application of Computer Vision in Responding to the Emergencies of Autonomous Driving

摘要

Nowadays, artificial intelligence is developing rapidly, and greatly promotes the appearance of relevant industries. Automated driving is one of those hot industries. With the development of autonomous vehicles, higher requirements such as intervention when an accident or system failure occurs mean more problems. Many shortcomings arose during emergencies. For example, the system cannot make a correct judgment in a short time and calculate the most suitable solution. This paper analyzes several algorithms for emergency situations at the current stage of autonomous vehicles from the perspective of computer vision. In addition, this paper finds suitable algorithms by collecting, collating and researching literature. Although these algorithms can solve the problems in most cases, there are still defects. For example, there are problems such as incomplete consideration and insufficient optimization. This paper also proposes possible solutions to the defects of each algorithm: The algorithm of edge recognition algorithm under extreme weather is added. Besides, the HMM model is used to optimize the pedestrian motion recognition algorithm. In addition, the standard of driver behavior recognition is also optimized. It is hoped that autonomous vehicles will have a more comprehensive emergency handling system in the future such as faster emergency response and more accurate decision.