Multi-Level Bayesian Models for Environment Perception

This book provides a comprehensive overview of existing techniques for automated multi-level interpretation of the observed static or dynamic environment,with a focus on machine perception. It proposes new original methods and utilizes well-established Bayesian frameworks to ensure a sound theoretical basis. Case studies are provided on passive radar and Lidar-based Bayesian environment perception tasks, demonstrating the effectiveness of the proposed contributions in real-world test images and videos.

Format: Paperback / softback
Length: 202 pages
Publication date: 20 April 2023
Publisher: Springer Nature Switzerland AG

This comprehensive book delves into the intricate realm of machine perception, employing a diverse array of 2D and 3D imaging sensors to address a selection of challenging problems. It presents a multitude of novel and innovative methods, while also offering a thorough and up-to-date survey of existing techniques for automated, multi-level interpretation of the observed static or dynamic environment. To provide a robust theoretical foundation for these novel models, extensive surveys and algorithmic developments are conducted within well-established Bayesian frameworks.

Low-level scene understanding is formulated as a series of image segmentation challenges, where the advantages of probabilistic inference techniques such as Markov Random Fields (MRF) or Mixed Markov Models are explored. For the object-level scene analysis, the book predominantly relies on the literature of Marked Point Process (MPP) approaches, which emphasize strong geometric and prior interaction constraints in object population modeling. Key advancements are introduced in the spatial hierarchical decomposition of observed scenarios and the temporal extension of complex MRF and MPP models.

In addition to utilizing conventional optical sensors, case studies are presented on passive radar (ISAR) and Lidar-based Bayesian environment perception tasks. Through rigorous experimentation, it is demonstrated that the proposed contributions embedded within a rigorous mathematical toolkit can yield significant improvements in the results obtained from real-world 2D/3D test images and videos, with applications spanning video surveillance, smart city monitoring, autonomous driving, remote sensing, and optical industrial inspection.

Weight: 338g
Dimension: 235 x 155 (mm)
ISBN-13: 9783030836566
Edition number: 1st ed. 2022