PSI - Issue 84

Regina Finocchiaro et al. / Procedia Structural Integrity 84 (2026) 1264–1269

1268

Links elements that change together over time/under action

Misses trends if photos are not comparable across campaigns

Georeferenced baseline imagery; repeatable viewpoints; time-series documentation

Common Fate

6. Critical discussion: real advantages and real risks of the proposed method In Author opinion, the most tangible advantage of the proposed method is not a “philosophical” but an operational one: when minimal perceptual metadata are added to inspection data reports, the data-base archive becomes readable over time by everyone. In a data-base context, such as AINOP or similar BMS systems, this means to be able to distinguish more rigorously between real deterioration evolutions (new cracking, progression, coherent changes) and differences due to the observer’s opinion or experience, lighting conditions, camera viewpoint, or non-homogeneous grouping criteria. The direct consequence is a more reliable “history” of the asset and, therefore, a stronger basis for defects prioritization and assessment planning. A second benefit concerns the correct and coherent selection of further investigations: when interpretative ambiguity and observational completeness become explicit and traceable, the transition toward L3/L4 is no longer a leap driven only by impression or perceived urgency but becomes a defensible decision. In practice, this reduces both unnecessary test and survey escalation (investigations triggered out for excessive caution or weak interpretations) and phenomena underestimation (due to apparently “clean” phenomena observed only partially or under operational constraints): in this manner the uncertainty is no longer shaded within the inspection narrative and inspection report, that is mandatory according to Guidelines. Finally, the framework integrates naturally with technologies such as drones, ARS (Augmented Reality System), and Computer Vision and Image Understanding: once a shared reference exists regarding what should be documented and how signs are organized into phenomena, the gap between what the artificial system captures and what the human operator interprets is reduced. The result is a more coherent work in which images, annotations, and automated outputs can be compared without “creative translations”: this approach permits to improve consistency, transferability, and auditability of results. Background shaded risks should not be concealed: an overly rigid standardization may miss “non-textbook” phenomena; moreover, Prägnanz may push toward overly-tidy interpretations. Mitigation doesn’t mean adding bureaucracy, but introducing the approach gradually, focusing on a limited set of high-utility attributes and on training with real cases (including ambiguity), supported by inter-inspector audit procedures or peer-review standardized processes. 7. Conclusion The proposed approach can be embedded in training as a grammar of observation, namely as a set of minimal practices that make defects evaluation and reading more reproducible. On the public archives side, the addition of perceptual metadata improves global auditability and future management of time series. On the technology side, a shared perceptual reference facilitates the use of advanced tools (drone surveys, Augmented Reality System, computer vision) because it better defines the bridge infrastructure between “detected patterns” and “interpreted phenomena”. In this perspective, the update of the operational instructions aimed at reducing discretion is naturally consistent with the idea of making grouping criteria and observational uncertainty declarable and traceable. The Bridge Guidelines introduce a solid and scalable multi-level framework, but the system remains structurally dependent on the quality of visual inspection and on the ability to organize defects and deterioration signs into interpretative phenomena. Gestalt principles describe mechanisms already active in any survey; making those mechanisms explicit, in a lightweight and workflow-compatible form, can surely reduce both discretion and improve

data quality without altering the Code structure. The contribution of this paper appears twofold: (i)

a critical reading of where perception affects the procedure