Facade Footprint Biasing for Reflective Surface 3D Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing photogrammetric reconstruction methods struggle to accurately reconstruct non-Lambertian surfaces, such as reflective or featureless surfaces, leading to incomplete or distorted 3D reconstructions with holes and artifacts, particularly in large buildings and water bodies.
Innovation Solution
A method and system that applies a bias towards filled-space in the footprints of non-Lambertian surfaces using convolutional neural networks and total variation (TV)-L1 voxel reconstruction algorithms to generate accurate 3D reconstructions by modifying observation models based on range-image estimates and land-cover classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If photogrammetric reconstruction is performed on non-Lambertian surfaces (reflective, moving, or featureless), then 3D reconstruction can be generated, but triangulation fails and produces holes in the reconstruction
Solution Approach 1:
The system performs preliminary classification of surface types (Lambertian vs. non-Lambertian) before reconstruction. By identifying reflective, moving, or featureless surfaces in advance, the system can apply appropriate handling strategies such as excluding problematic regions or using alternative reconstruction methods for those specific areas, thereby preventing triangulation failures before they occur.
Solution Approach 2:
The patent applies different reconstruction strategies to different regions of the scene based on their optical properties. Rather than treating the entire scene uniformly, the system identifies specific non-Lambertian regions and applies localized corrections or alternative methods only to those areas, preserving reconstruction quality while maintaining overall completeness.
2Ease of manufacture
If standard photogrammetric methods are used on large buildings with reflective windows, then processing is straightforward, but reflections cause triangulation failure and large holes in the building sides
Solution Approach 1:
The system modifies reconstruction parameters and observation models based on detected surface properties. For reflective surfaces like building windows, the system adjusts parameters such as reflection handling thresholds, observation volume configurations, and biasing factors to account for the non-Lambertian behavior, thereby maintaining reconstruction completeness without significantly complicating the overall process.
3Measurement precision
If observation models are modified with bias towards filled-space, then reconstruction accuracy improves, but processing complexity increases
Solution Approach 1:
The system pre-computes and stores biasing factors and observation model adjustments during a preliminary processing stage. By preparing these modifications in advance based on scene analysis, the actual reconstruction process can apply them efficiently without requiring complex real-time calculations, thus improving accuracy while managing processing complexity.
Data Source
Figure 1A~1B
Figure 2A~2B
Figure 3A~3B
AI summary
The present disclosure relates to systems and methods for automatically applying a bias towards filled-space in footprints of features that may have non-Lambertian surfaces during photogrammetric reconstruction of images with the features. The systems and methods may generate an observation model for the feature based on range-image estimates and determine whether the feature is a building or water using a land-cover classification. The systems and methods may determine a footprint of the features and may generate a modified observation model with a bias towards filled-space within the footprint of the feature. The systems and methods may apply a voxel reconstruction algorithm to the modified observation model to generate a three-dimensional (3D) reconstruction of the feature.