Research On The Measurement Technology Of Profile End Face Based On Machine Vision

Profile is a solid straight bar with a certain section shape and size after metal processing.Due to its good mechanical properties and convenient connection methods,there are widely used applications in modern manufacturing and construction industries.At present,the main method in the field of profile measurement is to manually use vernier caliper,inner diameter micrometer and other manual measuring tools to complete the measurement task.With the increase of output and accuracy of profile testing,the traditional manual measurement of low efficiency and low stability of testing are exposed.Based on the research results of high precision measurement of standard parts and profile size measurement characteristics of machine vision in recent years,thesis designs a profile size measurement system based on machine vision for common profiles.The main research contents of thesis are as follows:(1)Aiming at the requirement of high precision profile measurement edge detection,an improved subpixel edge detection algorithm based on polynomial interpolation was proposed.By adding adaptive threshold judgment mechanism and edge refinement,sub-pixel edge detection task is completed,and experiments are designed to verify the superiority of the proposed method compared with the traditional sub-pixel method in profile edge detection task.(2)In order to solve the problem that different types of profile size measurement requirements are changeable,and simple feature combination methods such as straight line and circle can not be used to measure directly,thesis proposes a key dimension measurement algorithm based on template.This method uses distance similarity as the judgment basis,matches the identified features and the features to be measured,and completes key dimension measurement tasks such as the outline size of the profile and the size of the assembly slot size.(3)In order to solve the matching problem between the collected image and the template,and the accuracy and efficiency requirements in the process of profile measurement,a cross-grain profile image classification method is proposed in thesis.The moment features of the collected profile images are used to complete the coarsegrained classification,and the classification accuracy probability is used as the judgment basis for enabling the fine-grained classification method.For the coarsegrained classification results with low accuracy,the fine-grained classification method based on image features with high time consuming,high precision and high computational power will be started.This cross-grain method can balance the precision and efficiency requirements of profile measurement and effectively complete the task of image classification in profile measurement.Finally,the machine vision measurement platform is built,and the experimental verification and result analysis are carried out.According to the size range and measurement accuracy requirements of profiles,the appropriate industrial camera,industrial lens,light source was selected,and the experimental platform was built.The proposed algorithm is verified by experiments and the measured results are analyzed.The results show that for the high-precision aluminum profiles with the diameter of the outer circle less than 100 mm,the average measurement error of the system built in thesis is 0.05 mm,which meets the measurement requirement of profiles smaller than the high-precision class,and verifies that the system built in thesis can better solve the task of high precision profile measurement.