Single Camera Object Detection for Self-Driving Vehicle: A Review

Abstract

The development of technologies for autonomous vehicle (AV) have seen rapid achievement in the recent years. Commercial carmakers are actively embedding this system in their production and are undergoing tremendous testing in the real world traffic environment. It is one of today’s most challenging topics in the intelligent transportation system (ITS) field in term of reliability as well as accelerating the world’s transition to a sustainable future. The utilization of current sensor technology however indicates some drawbacks where the complexity is high and the cost is extremely huge. This paper reviews the recent sensor technologies and their contributions in becoming part of the autonomous self-driving vehicle system. The ultimate focus is toward reducing the sensor count to just a single camera based on the single modality model. The capability of the sensor to detect and recognize on-the-road obstacles such as overtaking vehicle, pedestrians, signboards, bicycle, road lane marker and road curvature will be discussed. Different feature extraction approach will be reviewed further with the selection of the recent Artificial Intelligent (AI) methods that are being implemented. At the end of this review, the optimal techniques of processing information from single camera system will be discussed and summarized.