Kaleidoscopic imaging system for three-dimensional reconstruction of wear debris

Abstract Wear debris analysis is useful for monitoring the mechanical equipment condition and diagnosing faults. Traditional oil wear debris imaging methods usually require multiple cameras or mechanical tumbling to capture multi-view features, which brings complexity and potential sample interference. This study proposes a 3D reconstruction method for wear debris based on a single-camera Kaleidoscopic Imaging system. It aims to address the limitations of conventional multi-camera systems, such as high cost and complex calibration, as well as the drawbacks of mechanical rotation devices, including large footprint and low operational speed. The core innovation of this method lies in utilizing a precisely designed polyhedral mirror cavity to enable multiple reflections of light within a single exposure, thereby simultaneously capturing multiple virtual view of the same particle on a single camera. The proposed technique eliminates the need for mechanical moving parts or multi-camera synchronization units. It fundamentally avoiding failures caused by mechanical fatigue or synchronization errors, thereby achieving static and high-speed multi-view imaging. A long short-term memory 3D reconstruction neural network was employed to reconstruct wear debris from preprocessed images. Experimental results indicate that this method can efficiently reconstruct three-dimensional models of wear debris with diverse morphologies. The successful implementation of this approach provides a highly reliable and cost-effective solution for industrial oil monitoring.