Digital image recognition methods are now being utilized for steel bridge painting surface assessment. Through the use of digital image processing methods, the rust percentage on steel bridge surfaces can be objectively computed. The rust percentage may be crucial for the successful implementation of steel bridge painting warranty contracting since it can be used to decide whether painting contractors should repair painting defects at the end of the warranty period or at any regular inspection. To implement painting warranty contracting, however, appropriately designed recognition methods must be developed first. In this paper, several recognition methods are proposed and evaluated based on their rust recognition performance, and these methods fall into two categories. One uses artificial intelligence methods such as neural networks and fuzzy logic as the backbone for recognition, while the others apply statistical approaches to detect rust areas. Illumination is considered in some of the proposed methods to determine how it affects the recognition results. To test the performance of the proposed methods, a series of comparisons were made based on different environmental conditions.

No abstract