Abstract: To address the core pain points of traditional concrete cast-in-place pile core drilling inspection, such as low efficiency, subjective interpretation, and poor traceability, and to promote the digital transformation of pile foundation engineering quality control, this paper proposes an intelligent solution that deeply integrates computer vision, deep learning, and Internet of Things technologies. The system adopts a three-layer collaborative architecture of "cloud-edge-terminal". Core sample images are collected through mobile terminals. After image enhancement and geometric correction preprocessing, the improved Mask R-CNN algorithm is used to achieve precise instance segmentation of core samples and defects, and automatically extract key parameters such as core recovery rate and fracture width. Integrate multi-model collaborative decision-making with large language models to generate standardized professional descriptions, and build a full-process trusted traceability chain based on the pile position serial numbers. Research and application show that this technology has increased on-site cataloging efficiency by 47%, achieved a defect identification accuracy rate of 97%, and formed an unalterable digital quality file. This technical system has realized the objectivity, automation and credibility of core drilling inspection, providing a reliable technical path for the digital transformation of quality control in pile foundation engineering.