Computer Vision Methods for Non-destructive Quality Assessment in Additive Manufacturing

Abstract

Increasing availability and popularity of 3D printers cause growing interest in monitoring of additive manufacturing processes as well as quality assessment and classification of 3D printed objects. For this purpose various methods can be used, in some cases dependent on the type of filament, including X-ray tomography and ultrasonic imaging as well as electromagnetic methods e.g. terahertz non-destructive testing. Nevertheless, in many typical low cost solutions, utilising Fused Deposition Modelling (FDM) based technology, the practical application of such methods can be troublesome. Therefore, on-line quality assessment of the 3D printed surfaces using image analysis methods seems to be a good alternative, allowing to detect the quality decrease and stop the printing process or correct the surface in case of minor distortions to save time, energy and material. From aesthetic point of view quality assessment results may be correlated with human perception of surface quality, whereas, considering the physical issues, the presence of some surface distortions may indicate poor mechanical properties of the 3D printed object. The challenging problem of a reliable quality assessment of the 3D printed surfaces and appropriate classification of the manufactured samples can be solved using various computer vision approaches. Interesting results can be obtained assuming the appropriate location of the camera and analysis of the side view of the printed object where the linear patterns representing consecutive layers of the filament can be easily observed, especially for flat surfaces. Some exemplary experimental results of the application of texture analysis with the use of GLCM and Haralick features, Hough transform, similarity based image quality metrics, Fourier analysis and entropy are presented.