Thread pitch detection method based on watershed algorithm and geometric clustering

The parameter detection of threaded connectors is a crucial aspect of mechanical assembly quality control. This paper proposes a novel pitch detection method based on DIGIT (Dynamic Image-based Gelatinous Interface for Tactile) tactile sensors, achieving high-precision measurement through two core innovations: an improved distance-constrained watershed algorithm and geometric clustering with parallel line constraints.The system first captures thread texture images by leveraging the imaging characteristics of the sensor's elastomeric deformation. After image preprocessing to enhance contrast and reduce noise, an improved watershed algorithm is employed to resolve texture adhesion issues. Finally, the pitch value is calculated through parallel line clustering under geometric constraints. Experimental results show that this method achieves detection accuracy of over 90% for M1.5, M2, M2.5, and M3 threads, with a single detection time of 0.18 seconds.This approach not only enriches the modal technology path in the field of thread detection but also, with its advantages of lightweight design and high real-time performance, provides a feasible industrial application example for multimodal perception technologies. It aligns with the current the trends in intelligent manufacturing, particularly the integration of multimodal fusion in robotics.