Building Development Monitoring in Multitemporal Remotely Sensed Image Pairs with Stochastic Birth-Death Dynamics

Area of Science:

• Remote Sensing
• Computer Vision
• Geospatial Analysis

Background:

• Accurate building extraction and change detection are crucial for monitoring urban development and land use.
• Existing methods struggle with data heterogeneity and computational complexity in large-scale remote sensing datasets.

Purpose of the Study:

• To develop a novel probabilistic method integrating building extraction and change detection for remote sensing image pairs.
• To address challenges of data heterogeneity and computational efficiency in analyzing diverse aerial and satellite imagery.

Main Methods:

• A global optimization process considering observed data, prior knowledge, and spatial interactions.
• Object-change modeling using Multitemporal Marked Point Processes for recognizing changed and unaltered buildings.
• A flexible hierarchical framework for creating building appearance models from elementary features.
• Quick Multiple Birth and Death optimization and a nonuniform stochastic object birth process for efficient change detection.

Main Results:

• Simultaneous exploitation of low-level change information and object-level building descriptions.
• Successful separation of changed and unaltered buildings despite data heterogeneity.
• Efficient and optimal change detection with guaranteed convergence and manageable computational complexity.

Conclusions:

• The proposed probabilistic method effectively integrates building extraction and change detection.
• The approach demonstrates robustness in handling heterogeneous remote sensing data.
• The methodology offers a computationally efficient solution for large-scale geospatial analysis.