Rapid bearing flatness measurement of bridges with a developed quadcopter drone system and lightweight object detection model

Bearing flatness is a critical for the construction quality inspection of highway and railway bridges during construction. Conventional methods may result in high costs, operational safety risks and low measurement efficiency for high bridge piers in mountainous terrain. To address these issues, this paper developed an innovative quadcopter drone with an optical prism target for rapid and automated measurement of bearing flatness. The contribution of this study includes two aspects: (1) An optimized YOLO-MiniFaster model is developed for bearing padstones detection based on the RepNCSPGELAN module, the Slim-FPN feature fusion architecture, and SimAM attention mechanism. This enhanced architecture enables real-time and accurate detection of bridge bearings with higher computational efficiency. (2) A custom hardware prototype is developed, consisting of a quadcopter drone platform, an optical target module, a high-resolution image acquisition unit, and flight control system. The effectiveness of the developed drone system has been validated through field tests of a newly constructed railway bridge. Experimental results demonstrate that the developed model achieves a mean average precision (mAP) of 87.14% in detecting the bridge bearing padstones with a 14.3% reduction in model parameters compared to the original model. Furthermore, the quadcopter drone system enables a 45% reduction in on-site operational time compared to traditional manual methods. The developed system offers an alternative method that could improve the operational safety and measurement efficiency for bridge construction in hard-to-access locations.