Abstract:The past decade has witnessed the rapid development of autonomous driving systems. However, it remains a daunting task to achieve full autonomy, especially when it comes to understanding the ever-changing, complex driving scenes. To alleviate the difficulty of perception, self-driving vehicles are usually equipped with a suite of sensors (e.g., cameras, LiDARs), hoping to capture the scenes with overlapping perspectives to minimize blind spots. Fusing these data streams and exploiting their complementary properties is thus rapidly becoming the current trend. Nonetheless, combining data that are captured by different sensors with drastically different ranging/ima-ging mechanisms is not a trivial task; instead, many factors need to be considered and optimized. If not careful, data from one sensor may act as noises to data from another sensor, with even poorer results by fusing them. Thus far, there has been no in-depth guidelines to designing the multi-modal fusion based 3D perception algorithms. To fill in the void and motivate further investigation, this survey conducts a thorough study of tens of recent deep learning based multi-modal 3D detection networks (with a special emphasis on LiDAR-camera fusion), focusing on their fusion stage (i.e., when to fuse), fusion inputs (i.e., what to fuse), and fusion granularity (i.e., how to fuse). These important design choices play a critical role in determining the performance of the fusion algorithm. In this survey, we first introduce the background of popular sensors used for self-driving, their data properties, and the corresponding object detection algorithms. Next, we discuss existing datasets that can be used for evaluating multi-modal 3D object detection algorithms. Then we present a review of multi-modal fusion based 3D detection networks, taking a close look at their fusion stage, fusion input and fusion granularity, and how these design choices evolve with time and technology. After the review, we discuss open challenges as well as possible solutions. We hope that this survey can help researchers to get familiar with the field and embark on investigations in the area of multi-modal 3D object detection.

Vision-Based Traffic Accident Detection and Anticipation: A Survey

&Lt;title>automatic Traffic Real-Time Analysis System Based on Video</title>

TAD: A Large-Scale Benchmark for Traffic Accidents Detection from Video Surveillance

Vision Technologies with Applications in Traffic Surveillance Systems: A Holistic Survey

3D Object Detection for Autonomous Driving: A Survey

Vision-Based Accident Anticipation and Detection Using Deep Learning

Unsupervised Traffic Accident Detection in First-Person Videos

Vision-Based Traffic Sign Detection and Analysis for Intelligent Driver Assistance Systems: Perspectives and Survey

DoTA: Unsupervised Detection of Traffic Anomaly in Driving Videos

3D Object Detection for Autonomous Driving: A Comprehensive Survey

A Memory-Augmented Multi-Task Collaborative Framework for Unsupervised Traffic Accident Detection in Driving Videos

CRASH: Crash Recognition and Anticipation System Harnessing with Context-Aware and Temporal Focus Attentions

Automated Road Defect and Anomaly Detection for Traffic Safety: A Systematic Review

A Vision-Based Driver Assistance System with Forward Collision and Overtaking Detection

Text-Driven Traffic Anomaly Detection with Temporal High-Frequency Modeling in Driving Videos

3D Object Detection from Images for Autonomous Driving: A Survey

Surround-View Vision-based 3D Detection for Autonomous Driving: A Survey

Multi-Modal 3D Object Detection in Autonomous Driving: A Survey

Deep Learning Serves Traffic Safety Analysis: A Forward-looking Review

Deep learning-based perception systems for autonomous driving: A comprehensive survey