Research On Auto Focusing And Image Tracing And Mosaic Technology Of Part Geometric Dimension Measurement Based On Machine Vision

In the application of machine vision measurement,there are often problems such as inaccurate focusing and inadequate view field of camera.To solve the main problems,this paper studies the key technologies of measuring geometric dimensions of parts such as auto focusing and image tracing and mosaic with machine vision.Based on actual machine vision platform,this paper studies its overall scheme and work flow.Hardware composition and working principle of light source illumination unit,image acquisition unit,motion control unit and image processing unit are analyzed.According to actual need,selection of main components in illumination such as light source,lens and camera is studied.The composition of mechanical structure and motion control mode are analyzed.The feasibility of measurement scheme is verified from the perspective of system composition.This paper studies the auto focusing technology of machine vision detecting the edge of parts.With specific problems in experiment,a method of selecting focus evaluation window based on gray value statistics is proposed.By improving traditional Tenengrad function,a two-step focusing clarity evaluation function is designed and a position correction method is introduced.By improving traditional mountain climbing algorithm and designing a two-step focusing search strategy,the efficiency of auto focusing is improved and the focusing problem in machine vision measurement could be solved.This paper studies the image acquisition and processing methods.With image filtering,binarization,morphological processing,image pre-processing is completed.To solve the problem of insufficient view field of camera,an edge tracing algorithm based on neighborhood counting method is proposed which aimed to acquire all part images of parts by orderly camera tracing and shooting.Based on SURF algorithm,the detection area is optimized and image mosaic with the smallest size is proposed to obtain the complete image of parts.With edge detection and least squares ellipse fitting method,the geometric dimensions of parts in image coordinates are calculated.The image pixel equivalent is calculated by calibration experiment of the camera system.Measuring the geometrical size of the parts with this paper's algorithm,the absolute and relative errors of the measurement results can be controlled within 25?m and 0.5%.Experiments show that the measurement method studied in this paper is effective and has a certain reference and practical significance for machine vision measurement when the view field of camera is insufficient.