دانلود کتاب فیوژن عمیق دوربین و لیدار

This dissertation addresses open challenges in the field of autonomous driving by fusing LIDAR and camera data leveraging deep neural networks. This chapter will first introduce the goals and motivation of this research. To that end, we start with the history and current state of autonomous driving in Section 1.1. We continue with an overview of sensors in autonomous vehicles (Section 1.2) and discuss how to fuse sensor data in Section 1.3. More recently, multimodal data fusion methods leveraging deep neural networks have been proposed. An overview and discussion on them is presented in Section 1.4. Finally, the goals of this dissertation and its outline are provided in Section 1.5 and Section 1.6, respectively. Mobility has always been an integral part of our society. Especially the invention of the automobile in the 19th century has had a drastic impact on the everyday life of humans. Since then cars and other vehicles have shaped the way of transportation. Today, a significant amount of time is spent in cars [57], making them an important space of living. However, the increase in vehicle usage is constantly putting a burden on our traffic infrastructure, resulting in significant road congestion and full parking lots. Every day, road traffic accidents cause thousands of fatalities [82] and even more injuries. And with more vehicles on the road, the impact on the environment has increased as well. Large amounts of energy and resources for infrastructure, production, and fueling of vehicles are required. Motivated by this development, car manufacturers, big technological companies and many startups compete in bringing a groundbreaking solution to the market—autonomous driving. Autonomous driving is expected to improve the quality and availability of mobility in many aspects. It will allow to optimize the time spent in vehicles by enabling the passenger to work, leisure or sleep. It will enable elderly or impaired people to be picked up directly from home and be dropped at their target. Furthermore, traffic accidents could be drastically reduced, resulting in fewer fatalities per year. However, autonomous vehicles are not only a technological improvement but can also revolutionize mobility itself. They enable complex shared mobility services, reducing the need of privately owned vehicles. Shared Mobility would increase the average number of passengers per car. The accompanying reduction of the total number of vehicles on the roads could significantly relieve both the traffic infrastructure and the environment. In recent years, autonomous driving has been moving from the realm of science-fiction to reality, especially due to increased processing power and the advances in machine learning. Yet, many problems are still unsolved and autonomous driving remains one of the most challenging technological problems of the 21st century.