Research on monitoring construction quality of manually vibrated concrete based on machine vision(PDF)
《建筑科学与工程学报》[ISSN:1673-2049/CN:61-1442/TU]
- Issue:
- 2026年01期
- Page:
- 56-66
- Research Field:
- 智能检测与建造技术专栏
- Publishing date:
Info
- Title:
- Research on monitoring construction quality of manually vibrated concrete based on machine vision
- Author(s):
- WANG Shaojie; CHEN Lijuan; WU Qinglin; SHI Peixin; CHEN Lei
- 1. School of Rail Transportation, Soochow University, Suzhou 215000, Jiangsu, China; 2. Intelligent Urban Rail Engineering Research Center of Jiangsu Province, Suzhou 215000, Jiangsu, China
- Keywords:
- concrete; manual vibration quality; digital superintendent; machine vision; action recognition
- PACS:
- -
- DOI:
- -
- Abstract:
- Aiming at the problems that the quality judgment in the process of concrete manual vibration construction is highly dependent on the experience of operators, and the manual supervision has limitations and cannot monitor the risk in real time, an intelligent monitoring method of concrete manual vibration quality based on machine vision was proposed. First, a recognition model base on EfficientNet was developed to classify target concrete surface into “unqualified”, “medium”, and “qualified” categories with site images from live feed. Then, a STGCN++based pose recognition model was designed to dynamically identify and track 3 actions “vibrating”, “moving vibrator”, and “resting” by using skeleton key point sequences method. The outputs of the 2 models were incorporated for analysis and assessment on vibration time and surface state. A visual monitoring interface and workflow were presented for automatic simultaneous assessment and feedback of both product quality and workmanship quality. The results show that the global accuracy of the automatic identification method of concrete surface state in vibration is 99.89%, and the global accuracy of the automatic identification method of worker vibration behavior is 96.93%. The proposed method can effectively complement and even partially replace manual supervision, providing realtime detection and alerts for undervibration or missedvibration risks. The research has been applied in a metro station at Suzhou, and the results have good engineering application value.
Last Update: 2026-01-20