Workpiece Positioning Research Of Dispensing Machine Based On Machine Vision

Machine vision means measurement and decision by means of machine instead of human eyes.A machine vision system is one in which,the visual object is converted into image signals via image acquiring device,and then these signals is transmitted to a special image system and analyzed for features extraction.The decision results control the operation of manufacturing equipment.China has been becoming one of the most active developmental countries in machine vision in the world.The application domains for image vision include industry,agriculture,aerospace,public security,traffic,etc.This dissertation involves the industry of dispensing machine.The traditional dispensing machine usually adopts the manual input dispensing method,which has a series of problems such as poor accuracy and low efficiency,and cannot meet the requirements of high-precision and high-efficiency dispensing for modern manufacture.This dissertation takes visual positioning as the research object and introduces machine vision into traditional dispensers to improve the efficiency of dispensing and product quality.The dissertation gives the hardware components of visual positioning system,and discusses the performance parameters and selection schemes of hardware devices.The three-channel images acquired by the hardware device pass through a series of pre-processing operations such as image graying,image filtering and so on.Image graying technology reduces the amount of image data collected in real time.Image filtering weakens the disturbances from external environment.By studying the perspective model of the camera,the relationship between the image coordinate system and the three-dimensional spatial coordinate system is obtained.The distortion type existing in the camera is studied and corrected by means of OpenCV.Due to the actual situation of the dispenser and the camera,a system calibration method suitable for visual dispenser is proposed.In addition,the scale invariant feature transform(Sift)algorithm is deeply studied.Considering that Sift is designed for matching of object based on feature points,a modifying Sift — M-Sift algorithm is proposed for matching of image blocks,in which the image block is the basic unit,and based on the gradient,the amplitude-orientation histogram of the image block is calculated,and hence a 128-dimensional M-Sift feature vector is generated for image matching.Aiming to address the problem of slow speed and poor real-time performance of original matching algorithm in practical dispensing machine,we propose a fast matching and positioning method based on image pyramid by processing the image through the pyramid layering algorithm,and use the affine transformation to enhance the rotation invariant characteristics of the image.It is successfully used in the positioning process of the dispensing machine.Finally,the software system for dispensing machine based on machine vision was developed using the object-oriented design method on VC++ 6.0.Through experiments,it was shown that the development of the system achieved the expected effect of the dispenser.