Study on image diagnosis of deterioration and cracks in exterior finish coating using AI.

(Part 2) Study on improving the accuracy of diagnostic imaging.

Abstract

In this study, we investigated the improvement of the judgment accuracy of the crack density image diagnosis system for exterior finish coats using AI developed last year. Although a general-purpose AI engine is used in this AI image diagnosis system, the content of the analysis has not been disclosed, and the number of training sessions and the optimal method are unknown, so the conditions necessary for improving the judgment accuracy are unknown.However, there are several materials and patterns (textures) in finishing materials, and at first glance, it may seem difficult to distinguish between patterns and cracks. Therefore, in order to confirm the influence of the type and texture of the finish material on the image diagnosis of AI, we newly learned the deterioration cracks of a thin waterproof exterior coating material E (ripple pattern), and examined whether the image diagnosis would be different from that of the deterioration cracks of a multi-layer coating material E (sprayed pattern convex treatment), which was learned previously. The results showed that AI may be able to diagnose cracks in finish coats regardless of the material or pattern.