3. LiDAR perception system

Perception is essential to any autonomous machine application involving sensory data and artificial intelligence techniques. The ultimate goal of a perception system is to extract spatial and semantic information from the raw sensor data, enabling the machine to construct a complete representation of its immediate environment. This requires descriptions of all perceived objects, including the types, positions, headings, speeds and dimensions of all objects in the environment. Perception processes applied to intelligent vehicles can be divided into two types: the model- or geometry-based approach, and the data- or machine-learning-based approach. The former is used to extract positional and dimensional information from observed objects. Structured 3D information is also used to infer semantic data by matching different a priori models. Until the emergence of deep learning, machine perception...