Image Segmentation Algorithm Suitable For Industrial Applications

INDUSTRIAL 4.0 often by visual inspection,analysis of product quality parameters.Image object segmentation refers to the detection of the target object extracted from the image,the image analysis of the leading process of understanding.The accuracy of image analysis,the rationality of the test data or the correctness of the premise.This requires image segmentation must be stable and reliable,and industrial products to detect a target object categories,and different,such as tile surface flatness,chromatic aberration,the side of the defect;bottle filling roundness,defects;sealing of character recognition;corrupted pixel screen,color difference and so on.At present,although there are many kinds of image segmentation methods,but it is required to select the scene of different detection target,to extract the experimental data,to analyze and verify.To this end,the thesis topic research,has the theory value and the practical significance.In this paper,a detailed analysis of the existing segmentation techniques is carried out,and the classification of the algorithms is proposed,which is based on the underlying feature of the image and the rules of the high level model.In each category the segmentation algorithm of multiple image research scholar,detailed analysis of the segmentation principle,the advantages and disadvantages of each algorithm are compared,describes the many people involved in industrial image project,set up each project target segmentation are tested.The experiment results multiple algorithms of industry image results and data,proposed according to the experimental results of time and accuracy data selection an acceptable algorithm and its application in specific projects.For the printing of characters in a character project background gray level difference is large.choose a variety of segmentation threshold based on the comparative experiment;according to the characteristics of a single part of unglazed tiles texture mainly concentrated on the edge,on the selection of testing multiple segmentation method based on edge;the light oil filling characteristics not uniform,select multiple segmentation method based on region of the experiment;for the characteristic of the large pieces of glazed tile brightness and background difference,select the two segmentation method based on clustering of the test;according to the characteristics of oil bottle in the uneven brightness of the box,select a test for the segmentation graph based on the screen;defects and background difference threshold is small and uneven brightness,selection experiments on the two kinds of segmentation methods based on energy functional.The main work of this paper include:through rigorous data formula with analysis of each algorithm specific calculation steps,from the perspective of the theory of comparative advantages and disadvantages of each algorithm,through experimental analysis of each algorithm on real industrial image processing effect,the comparison of each class of algorithm of time efficiency and accuracy.The innovative points of this paper are as follows:according to the low-level features of images,such as threshold,edge and region information and high-level model such as graph theory and energy functional will present image segmentation algorithms are divided into six categories;correct rate of data by the method.The experimental results of time efficiency and image processing to select for specific industrial projects a segmentation algorithm.