High-precision prediction of micro-displacement of workpiece using I-ResNet50V2 and dual-laser composite imaging

Abstract Monocular laser vision systems and machine learning-based micro-displacement measurement methods face critical limitations in feature representation accuracy and computational efficiency. Existing approaches rely on manually designed image metrics or suffer from excessive model complexity, leading to compromised precision and low computational efficiency. To address these challenges, we propose a lightweight deep learning framework integrating an optimized ResNet50V2 architecture with dual-laser composite imaging. Key innovations include: A structurally enhanced ResNet50V2 network that reduces parameters through depthwise separable convolutions and multi-scale feature extraction, while improving displacement-sensitive feature learning; A dual-laser imaging system that enhances robustness through adaptive fusion of complementary laser patterns. Experimental results demonstrate a measurement error below 0.07 mm with merely 12.98 million parameters, outperforming conventional machine learning methods in both precision and computational efficiency. The proposed system provides a cost-effective solution for high-precision micro-displacement measurement in resource-constrained industrial applications.