Research On Weld Position Recognition Based On Machine Vision

With the progress of science and technology and the rapid development of industrial technology,people put forward higher requirements on the welding technology.The traditional manual welding method has the disadvantages of low efficiency,high cost and uneven quality,which is far from meeting the requirements of industrial development.Automation and intelligence are the future trends.Therefore,the measurement and tracking of the weld position has become the focus of current research.Machine vision detection technology has become an important development direction of precision measurement technology because of its advantages of non contact,high precision and fast speed.In this paper,the traditional binocular vision measurement algorithm is analyzed and improved for the welding needs of the workpiece.The binocular vision measurement system is set up and the seam recognition program is written.This paper refers to the design experience of a large number of machine vision measurement systems,and designs the overall plan according to the characteristics of the workpiece and the requirements of measurement.The construction of the hardware has been completed,and the suitable light source,industrial camera,lens and so on are selected.The installation and lighting mode of the light source,as well as the characteristics of the industrial camera and lens are analyzed.After that,we studied the image pre-processing technology,such as image filtering,corrosion,expansion,two value and so on,and the image post-processing technology such as pre-weld point recognition and line fitting.We also gave the processing effects of various algorithms for workpiece images.Finally,the traditional binocular vision measurement method is analyzed.At the same time,the binocular measurement algorithm is improved according to the actual situation of the workpiece.The parallel optical axis and the convergent optical axis are used respectively to measure the accuracy.The results are analyzed.In this paper,binocular vision measurement system is built by LabVIEW software.The accuracy of three-dimensional coordinate measurement is tested on the laboratory CNC machine tool,and the workpiece welding test is carried out at the scene.Theexperimental results show that the two algorithms have achieved the desired level of design in terms of dynamic performance and precision.