Machine Vision Based-2D Measurement Method For Circular Workpiece

In industry,some parameters of products such as size,surface finish,strength and so on,have to be detected in order to ensure the quality of the products.With the development of large-scale production,the traditional manual operation methods will become unstable when the number of products,as well as the detection accuracy and velocity are increased.Machine vision measurement technology has been applied widely in the fields of industrial inspection and quality control for the advantages of high automatization,high detection precision and non-contact measurement.For the need adapting to the development trend of modern industrial measurement,this paper adopts machine vision and image processing technology to detect the diameters of some circular parts,and conducts an in-depth study on developing a 2-D dimension circular parts detection method based on machine vision system under existing experimental conditions.The research work mainly in the following aspects:1)Hardware platform construction.On the basis of analyzing the system framework as well as the performance parameters of hardware,considering the cost-effective and the actual demand,we construct a measuring device,including computer,CCD image sensor,lens,and give the structure and the work flow of the system software.2)Analysis and selection of image preprocessing methods.By comparison of the performance of some commonly used image processing techniques,i.e.,image filtering,segmentation,edge detection and so on,on the processing of workpiece images,the suitable preprocessing methods can be found and were applied into our work.3)In order to detect circles quickly and accurately with less memory and time consuming,this paper presents a method based on artificial fish swarm algorithm(AFSA)for the automatic detection of circular workpiece.Firstly,we use the combination of three non-collinear edge points as candidate circles(i.e.,artificial fish),then calculate the fitness value by means of counting the number of edge points which locate on the candidate circles.Guided by the fitness value,artificial fish can eventually search the optimum set of three edge points by repeating preying,swarming,following behavior and so on.Lastly,we also introduce the circle detection methods based on the basic Hough transform,the random Hough transform(RHT),the genetic algorithm(GA)and give an analysis and comparison with the proposed method.Experimental results over several complicated and noisy images have validated the efficiency of the proposed technique regarding accuracy,speed and robustness.4)System calibration.This paper completes the system calibration and then carries on a large number of experiments to the circular workpiece.The experimental results were analyzed,the reasons of errors and the ways to eliminate them were discussed.The resultsshow that the proposed method can meet the accuracy requirement of general industrial detection.This also provides a way for the specific application of machine vision to the industrial measurement.