An aero-engine lockwire kink density detection method based on target-driven 3D reconstruction

As a critical mechanical anti-loosening device in aero-engine, the kink density detection of lockwire is essential for ensuring reliable connections and operational safety. However, there are currently no automatic detection methods for lockwire kink density. The optical 3D reconstruction is a feasible solution. Nevertheless, existing 3D reconstruction methods are not target-specific and introduce substantial redundant data. In this paper, we propose an aero-engine lockwire kink density detection method based on target-driven 3D reconstruction. Firstly, the acquired binocular aero-engine images undergo lockwire segmentation, followed by local bright spot extraction to construct the lockwire feature curves. A soft shape context approach is then designed, integrating both global and local information for lockwire stereo matching. Subsequently, we build an energy function by introducing spatial distance constraints. Afterwards, a curve–constrained bidirectional belief propagation framework is proposed to minimize the energy function for accurate lockwire keypoints 3D reconstruction. Experimental results demonstrate that our proposed method achieves superior performance with an accuracy of 98.04%.