Research On Key Techniques Of External Thread Geometric Parameter Measurement Based On Machine Vision

Thread is a very common connection form in aerospace fields, and its geometric parameters have a significant impact on performance of aircraft and spacecraft. The domestic and foreign scholars did a lot of researches in external thread measuring methods, however, the efficiency and accuracy of existing detection techniques are still unable to meet measurement demand. The purpose of the paper is to design a non-contact, efficient, high-precision MJ screw thread(metric aerospace thread) measuring system. A new method for external threads measuring based on machine vision is proposed and many measurement algorithms and system errors are studied in the paper. The paper mainly focus on the following aspects:(1) A fast and accurate external thread measuring system based on machine vision is proposed in the paper. In order to avoid the lower accuracy problems of direct three-dimensional reconstruction of external screw, two mutually perpendicular cameras are used to acquire screw image and thread contour is used to extract threads parameters in the paper.(2) Thread image processing and parameter extraction algorithms are studied in the paper. The image gradient and sub-pixel accuracy contour is extracted by screw image preprocessing. Extraction algorithms of multiple parameters(include major diameter, pitch diameter, minor diameter, pitch, thread angle and radius of rounded crest) are designed and validated in experiment.(3) Main errors of the measurement system are studied in the paper. For the optical imaging part, the influence of depth of field and light diffraction are studied. For the image processing algorithms, the thread contour extraction errors are studied. For the mechanical system, effects of object distance error, image distance error, angle error of cameras and camera rotation angle error are studied. The result of error analysis shows that, even installed with non-precision machinery, the relative measurement error can be far less than maximum error allowed in the national screw standard GJB3.2A-2003(Suitable for aerospace thread), and the system can be used to measure the MJ external threads accurately.In the paper, a new system based on machine vision is proposed to meet the MJ screw measurement need in the aerospace field. The overall scheme, algorithms and error analysis discussed in the paper are useful for the studies of other screw thread measurement. The paper is worthy of reference for the development of non-contact, efficient and high precision thread measurement systems.