Accurate 3D modelling for automated inspection: A stereo vision approach
A. Bhatti, and S. Nahavandi
Deakin University, Australia
Motivation and Objective
Dense 2D disparity
Accurate 3D depth modelling
Visual Inspection System
A stereo vision approach using Multiresolution analysis
Matching Algorithm
Coarsest Level Matching
Scale Normalization
Wavelet Transform Modulus
Coarsest Level Matching
Wavelet coefficients vectors and Similarity by correlation
Consistency Measure
Probabilistic Measure
Process Constraints
Rest of the process is constrained with precise disparity range [
d
min
d
max], uniqueness, continuity, etc.
Consistency Measure
Interpolation
Reference 3D Model
3D re-projection Equations
Calculated 3D Model
Accuracy of the Depth
Conclusion and Future Work
Multiresolution Algorithm: capable of producing 3D depth model to reasonably high accuracy
Accuracy of the algorithm relies on the consistency and pre-calculated information rather than extra hardware tools
Enhancement in depth accuracy by translating the error tolerance in terms of the relationship between depth error and calibration information
