Point clouds serve not only as a type of spatiotemporal data but also as a 3D representation model, providing a fundamental method for 3D digitization and semantic expression. With the advancement of 3D equipment, such as LiDAR, the volume of point cloud data has been rapidly increasing, necessitating the use of deep-learning-based analytics to manage these data effectively. Consequently, point cloud machine vision analysis has garnered significant attention in the field of computer vision and various applications, including smart cities, digital preservation of cultural heritage, autonomous driving, film and television entertainment, and infrastructure security monitoring. In this chapter, we present a comprehensive overview of foundational methods for deep-learning-based point cloud analysis. We commence with an examination of traditional techniques for point cloud classification and semantic segmentation. This is followed by an exploration of methodologies for point cloud object detection and tracking. Each method is detailed starting with a problem statement, followed by an exposition of the general solution processes, representative works, and prevailing trends. Collectively, this chapter aims to elucidate the core methods underpinning deep-learning-based point cloud analysis.