Welding Automatic Control Systems, Self-calibration Based On 3d Computer Vision

With semiconductor industry being the mainstream industry of many fields, its market becomes more and more large. But ultrasonic weld machine in later working procedure of semiconductor manufacture mostly operated by hand in our country by now,which leads to low work efficiency and unstable production quality, on the other hand fully automatic weld machines need expensive price.Therefore, Basing on the existing manual ultrasonic gold-silk ball weld machines of small and medium-sized enterprises, this subject adds control equipments with computer, control software , and workbench of electrical servo machine etc., which cost a small amount of money and transform the existing equipments into automatic equipments , and greatly improve the qualification rates of productivity and productsThe paper puts emphases on the research of the initial calibration and system pretreatment, so as to ensure the accuracy of the follow-up work while the system operates in real running time. It mainly includes the following four aspects:(1)Finish the automatic transformation of the manual ultrasonic gold-silk ball weld machine.Automatically orient and weld to all workpieces of the same type . And it also includes developing the identification software of the artificial vision and control software of electrical servo in the project.(2)According to the actual conditions of this weld system, the non-linear automatic calibration arithmetic with six parameters is proposed based on the precision and large calculation characteristic of the non-linear calibration arithmetic in two steps with radial arrange restriction and the concision and granulation characteristic of the inchoate calibration arithmetic. At first calculate out the angle θ between two plane coordinates and calculate out two parameter Xc and Yc; Considering radial distortion, optimize and calculate the left three parameters with genetic algorithms.(3) In order to ensure the quality of the images during both light core calculation and subsequent image matching, wavelet technology is used .According to the principle of immovability in optical axis of camera and some characteristic points of theimage,this paper utilize edge detection, Hough transform and least square algorithm to calculate light core Xc and Yc.(4)Using genetic algorithms to optimize and calculate the distorted parameters k and cl,c2,we can get all precise parameters.Considering the actual system condition,the non-linear automatic calibration arithmetic and system pretreatment are compact and speedy, simultaneously it ensures better robustness and precision of the system.The automatic equipments finished by this subject only needs to detect, identify and self-study the same kind of workpieces for one time,and need not manually operation later. It greatly reforms the later semiconductor production status of small and medium-sized enterprises in our country and reveals flourishing market foreground..It is believed that the fruit of this subject is valuable and influential to the research on the camera calibration and vision identification under special condition.