Quality Inspection Of Laser Welding Based On Machine Vision

Welding is a widely used technology in industrial production and processing.Laser welding is a new type of non-contact welding method.Compared with the traditional welding method,it has higher precision and more efficient.It can realize the intelligent and automatic welding,and can meet the requirements of modern industrial production.Welding finished products need to be tested for quality,can be used after passing the test.The traditional manual inspection method requires a large amount of labor.Applying machine vision to the quality inspection of laser welding can improve the production efficiency and ensure the accuracy of inspection.In this paper,laser spot welding is carried out on the surface of aluminum plate with welding platform.Aiming at three welding quality problems,such as perforation,broken welding and distortion,a laser welding quality detection system based on machine vision is designed.The system consists of three parts: light source system,image acquisition system and image processing system.The light source system and the image acquisition system mainly choose the hardware,and the image processing system is responsible for the image algorithm processing.In this paper,ring LED light source is used to provide illumination for the system,calibrate the selected CMOS camera,and correct the lens distortion.In order to solve the problem of uneven illumination in laser welding image,a new two-dimensional gamma function is used to correct it.For image filtering,a bilateral filtering algorithm is selected.For image segmentation,Canny edge detection is chosen.A dataset was established for the segmented images,and the genetic algorithm was used to optimize the SVM.After training,the classification accuracy of the three samples were 80%,76.67% and 83.33%,respectively.The convolutional neural network is used to train the images of perforation,welding break and distortion.After training,the classification accuracy of the three samples reaches 92%,88% and 92% respectively,which can meet the requirements of quality detection of laser welding.