Error Propagation and Statistical Validation of Computer Vision Software
Abstract
Computer vision software is complex, involving many tens of thousands of lines of code. Coding mistakes are not uncommon. When a vision algorithm is run on controlled data which meet all the algorithm assumptions, the results are often statistically predictable. This renders it possible to statistically validate the algorithm and its associated theoretical derivations. In this paper we review the general theory of some relevant kinds of statistical tests and then illustrate the experimental methodology of statistical algorithm validation to validate a program that estimates parameters of buildings in aerial photographs. This program estimates the 3D positions of building vertices based on input data obtained from multi-image photogrammetric resection calculations and 3D geometric information relating some of the points, lines and planes of the building to each other.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 01, 2001
- Accession Number
- ADA458742
Entities
People
- Robert M. Haralick
- Tapas Kanungo
- Xufei Liu
Organizations
- University of Maryland