Research And Implementation Of Part Size Detection System Based On Machine Vision

Since the “Made in China 2025” strategy has been promulgated,the domestic machinery manufacturing industry is in the booming development.Manufacturing technology is changing rapidly,and the requirements for detection technology are gradually improved.Traditional manual contact detection method can not meet the current detection requirement because of its low speed?low accuracy and poor flexibility.Moreover,the equipment that can realize the precision detection is very expensive and the antiinterference ability is poor.So it can't adapt to the requirements of the on-line inspection of the factory environment cooperated with automatic production line.On the contrary,the detection technology based on machine vision has the characteristics of fast,accurate,flexible and low cost.Therefore,based on the machine vision technology,a part detection system is designed to improve the detection accuracy and speed.This paper selects Opencv image processing library and C# development tool to design this part size detection system.This system has studied the key technology and the calibration of the system in the process of image processing,and realized the sub-pixel location detection of the part size edge.This paper used a representative body part as the research object.According to the reflective characteristics on the surface and size parameters of the workpiece.By comparing and analyzing experiments,we select the best lighting mode which can highlight the features of the workpiece's diameter and height from three common lighting methods.This paper adopted three kinds of filtering methods to deal with the image denoising respectively.Analyze the gray change of the same line of pixels in different filtering modes,and select the most reasonable filtering method to reduce the influence on edge ambiguity.This paper presents an improved method of edge detection of Canny operator with morphological fusion.This method can reduce the detection speed greatly without loss of detection accuracy,and achieve the edge single-pixel positioning accuracy.Moreover,this paper puts forward an improved Zernike sub-pixel edge operator,the operator can automatically obtain step gray optimal value instead of the usual manual selection,which makes the system behaves robustly.In this paper,the dimension measurement principle of the system is described in detail.In order to avoid the problem that "marginal contour point field value" affects the detection precision,Hough is combined with the least squares fitting to improve the detection accuracy of the line fitting.The part size detection system is used to carry out single workpiece repeated measurement and multi-type workpieces repeated measurement experiment,recording dimension data and detection time.The experimental results show that the test system conforms to the preset detection index.In addition,the error analysis is carried out for the experimental results,and the improvement scheme is proposed for the error source.Based on the above analysis,this part detection system meets the current detection requirements and has universal applicability.