Image Detection And Analysis Of Weld Forming Based On Vision

With the continuous development of industrial technology,traditional detection methods can not meet the current needs.For the welding field,the application of machine vision technology has become an inevitable trend of development.The traditional weld quality inspection is done manually,which has the disadvantages of low accuracy of measurement results,long time consumption,high labor cost,etc.The application of machine vision technology can avoid the disadvantages of manual measurement.In this paper,machine vision technology is applied to weld quality inspection,in order to obtain accurate weld geometry information intelligently.In terms of light source selection,the active vision method is selected,i.e.external addition of line laser.On the main hardware selection of the system,by comparing the difference of industrial camera chip,operation principle and signal transmission mode,the HIKVISION industrial camera with CMOS chip and GigE signal transmission mode is finally selected.Then,a vision acquisition system and a motion control system are constructed,which are connected to Ethernet and communicate by TCP/IP protocol.The signal is transmitted to the computer image processing software to collect high-quality weld structural light images.Combining VS software and OpenCV image processing library,the weld structural light image is processed.Firstly,the collected RGB images are grayed out.By analyzing the image features,median filtering is used to smooth out noise.OTSU algorithm is applied to thresholding the filtered image to realize image binarization,which is beneficial to separating the background and foreground of the image and preserving more important image morphological information.Mathematical morphology and ROI extraction algorithm are selected to process the binarized image to obtain the region of interest satisfying the requirements.In the process of centerline extraction,this paper proposes an algorithm that combines Hessian matrix and maximum gray method to skeletonize stripes.In the extraction process of structured light stripe feature points,the least square algorithm is firstly used to fit the skeleton in the image in different regions,and then the skeleton is extracted respectively according to the special positions of each key point.Finally,the conversion parameters of actual weld coordinates are obtained by calibration technology,and the image point coordinates are converted into three-dimensional coordinates for measurement.The experimental results show that the geometric information of weld size obtained by the improved algorithm proposed in this paper has the advantages of high accuracy and less time consumption,and meets the industrial requirements.