Technical Innovation in Engineering Quality Control Based on Micro Visual Recognition

Abstract

Engineering quality control is shifting from sampling-based acceptance and manual review to online perception, process intervention, and evidence traceability. Taking detailed quality problems such as prefabricated-building joints, rebar tying, component seams, and apparent concrete defects as research objects, this paper proposes a technical innovation path for engineering quality control based on micro visual recognition. This path uses close-range miniature cameras, controllable fill lighting, edge-computing gateways, and lightweight deep-learning models as support, and builds a closed-loop system of acquisition, recognition, judgment, rectification, reinspection, and archiving. The study shows that micro visual recognition can not only improve the discovery rate of small-scale defects such as microcracks, misalignment, holes, and rebar-spacing deviations, but also connect visual evidence with procedures, components, responsible parties, and BIM models, thereby shortening the delayed exposure time of quality problems. Implementation strategies are proposed from four aspects: system architecture, algorithm optimization, data governance, and management mechanisms, providing reference for refined, digital, and proactive engineering quality control.