BLAST: Bayesian online change-point detection with structured image data

2025-04-14

AI Summary

This paper addresses three key challenges in real-time change-point detection for high-dimensional image data: low computational efficiency, insufficient modeling of structural features, and difficulty in quantifying uncertainty. To this end, we propose the first Bayesian online change-point detection framework tailored for structured images. Our approach introduces a novel deep Gaussian Markov random field (Deep GMRF) prior to explicitly capture spatial dependencies within images. We further design an online-updatable posterior run-length distribution inference mechanism, enabling scalable Bayesian inference with computational complexity of O(p²). Evaluated on street-scene surveillance and metal additive manufacturing process monitoring tasks, our method reduces detection delay by 37% and uncertainty calibration error by 52% compared to state-of-the-art approaches. To the best of our knowledge, this is the first framework achieving structure-aware, computationally efficient, and rigorously uncertainty-quantified online change-point detection for image sequences.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Computer Vision: Other Foundations of Computer Vision

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

The prompt online detection of abrupt changes in image data is essential for timely decision-making in broad applications, from video surveillance to manufacturing quality control. Existing methods, however, face three key challenges. First, the high-dimensional nature of image data introduces computational bottlenecks for efficient real-time monitoring. Second, changes often involve structural image features, e.g., edges, blurs and/or shapes, and ignoring such structure can lead to delayed change detection. Third, existing methods are largely non-Bayesian and thus do not provide a quantification of monitoring uncertainty for confident detection. We address this via a novel Bayesian onLine Structure-Aware change deTection (BLAST) method. BLAST first leverages a deep Gaussian Markov random field prior to elicit desirable image structure from offline reference data. With this prior elicited, BLAST employs a new Bayesian online change-point procedure for image monitoring via its so-called posterior run length distribution. This posterior run length distribution can be computed in an online fashion using $mathcal{O}(p^2)$ work at each time-step, where $p$ is the number of image pixels; this facilitates scalable Bayesian online monitoring of large images. We demonstrate the effectiveness of BLAST over existing methods in a suite of numerical experiments and in two applications, the first on street scene monitoring and the second on real-time process monitoring for metal additive manufacturing.

Problem

• Detects abrupt changes in high-dimensional image data efficiently
• Incorporates structural image features to avoid delayed detection
• Provides Bayesian uncertainty quantification for confident change detection

Innovation

• Bayesian online change-point detection for images
• Deep Gaussian Markov random field prior
• Scalable O(p^2) computation for large images