Study On Real-time Measurement Method Of High-Speed Revolving Object Based On Image Information

In the process of high-accuracy machining of small parts with high-speed milling machine,because tool size changes caused by temperature changes significantly affect the machining accuracy,it requires real-time measurement and to amend the control parameters.There are many difficulties to carry out the accuracy measurement to the tool geometry of high-speed rotation,this issue discusses and adopts the method that measures the axial dimension for ball-end milling of small diameter(1mm) based on processing technology of image information.This paper mainly carried out the following studies.The working accuracy of the NC milling machine requests measurement system's resolution achieves 1μm,the most simple mean is raising the CCD camera's resolution,but this approach is limitated by the hardware,can not be achieved.So this paper uses software processing technology to realize the highly accuracy localization of image target.In the paper,firstly,it researchs the principle and method of sub-pixel edge detection.Secondly,a number of commonly used sub-pixel edge localization methods are introduced,such as gray moment method,gravity center method, polynomial interpolation and fitting.Principle,the good and bad points of each algorithm are elaborates.Through analysis,it shows that the gray moment method and the center of gravity method are well suited to locate symmetrical edge in principle and has a better anti-noise ability.While polynomial interpolation method is fit for positioning on the asymmetry edge,is bad for the anti-noise ability.The algorithm be needed in this paper,is not only required to has a higher location accuracy,stronger anti-noise ability,but also has better anti-fuzzy performance,better stability.In order to find the algorithm which can meet the requirements,the paper proposes to do simulation experiment on the computer.At first,according to the situation of light irradiation,respectively draw up two kind of ideal arc light,which brightness function subject to symmetrical distribution and non-symmetrical distribution.The real coordinates value of the edge points is known. realizes the gray-level moment method,the center of gravity method,interpolation and fitting with the C++Builder programmable tool.And then the effect by a single factor be simulated.Through the comparison of the true edge and the edge of positioning errors can be made qualitative analysis of algorithm performance.The results of the analysis are as follows.When using algorithm do directly position to the ideal image,it may know the algorithm's own locational accuracy.The experiments show that when using center of gravity method to locate Symmetric edge,the accuracy can achieve the required accuracy of this article,the accuracy of the gray moment method has been greater than the accuracy of this article,although the mean error of the interpolation is very small,but significant error standard deviation,while using center of gravity and the gray methods to locate asymmetric edge,the average error is greater than the maximum measurement accuracy,but the error containes a regular error and it can be amended in stabile edge to meet the requirements of positioning accuracy.After adding noise signal with controllable parameters to the ideal image,locating the image edge using those above algorithms,way,we can analyze each algorithm's anti-noise ability.The best anti-noise ability is the center of gravity method.Followed by gray moment method,the worst is interpolation.In addition,the influence of filtering to the algorithm also has been studied.The method as following, filtering the ideal image with noise signal use 3×3 and 7×7 filter window separately filer three times and one time.Compared with two localization effects in same algorithm,the experiment proves that the former effect is better than the latter.In regard to the anti-fuzzy,the experiment proved that the center of gravity method has less anti-fuzzy ability than the gray-level moment method.However,its positioning errors are also in the scope of measurement error which this article permits.It is usually considered that raising the resolution can improve the algorithm's localization accuracy,but,at the same time,it will also increase the calculated amount.In order to choose the appropriate resolution,two different kinds resolution(1/20 and 1/100) are researched in the paper,The results showed that to improve the resolution can cause the positioning accuracy of the center of gravity method and the gray-level moment method enhance 3-6 times.However,the accuracy of fitting was no significant difference in both cases.in view of the 1/20 resolution has been able to meet the accuracy requirements,therefore,we take the 1/20 resolution.The size of the milling cutter studied in this paper is very small and segment diameter is only 1mm,so the points in the image can be used for sub-pixel analysis is few.Besides,milling head is so short that further limits sub-pixel algorithm positioning accuracy.According to the lens forming images principle,readjusting the image distance and the object distance properly may get magnified milling cutter image.So the paper designs a set of image grabbing device based on above criterion.The actual noise which is irregular,uncontrollable and complicated has a significant impact on the algorithm accuracy.Therefore,it is necessary to analyze the impact of the actual noise on the capacity of the algorithm.By the analysis of several algorithms processing results to actual image.We can see that the best anti-noise ability is the center of gravity method,followed by gray-level moment method,the worst is interpolation which cannot locate normally under the actual noise.Moreover,image acquisition contains variable factors,such as light,high-speed rotating objects and so on.These make the selected algorithm in the paper should not only has a high positioning accuracy,but also has better adaptability in change conditions,which is called Robustness in Image Processing.Regarding to the analysis of Robustness,this paper designs an experiment,in this experiment,a series of image were obtained,these images are either illuminated by variously intensity light source in different angles or illuminated at different time with the identical light source.By comparing the processing results of several algorithms,we drew a conclusion that the center of gravity method has better Robustness as well as the anti-noise aspect than the gray-level moment method.Through the analysis of computer experiments and practical results,the paper,at first,adopts the center of gravity method to locate the arc-shaped milling cutter,and then uses the quadratic curve fitting method to derive the method of vertex coordinates.The imagery processing timelines requests that algorithm processing time must be shorter than the image acquisition time.Under the above requirement,how to realize the shifting window,the center of gravity algorithm,the vertex scheme method and the quadratic curve fitting in hardware are researched in the paper.In the design of the shifting window,each window contains 4 line picture element spots,this may do further equalization to the center of gravity spots.So that the edge which obtains through fitting will more smooth.For the center of gravity algorithm has massive multiply operation,it will take great amount room spaces.What is more serious is that its processing time is longer than the image grabbing time,this does not satisfy the processing request.Therefore,this paper makes improvement to the center of gravity algorithm.Above all,according to the principle of the center of gravity method and handling characteristics,it chooses relative coordinate origin.After that, according to the characteristic of molecule and denominator in the center of gravity method,it designs the additive procedures.Finally,it programs the improved center of gravity method with functional block diagram law,and programs the quadratic curve fitting method to derive the method of vertex coordinates with verilogHDL.Both simulation results conform to requirement of the paper.