3-Dimensional Reconstruction of a Highly Specular or Transparent Cylinder from a Single Image

Abstract

In production lines, high-speed inspection and quality control are necessary to maintain the quality of a product. Automation in quality checking saves time, reduces manual work, and increases the accuracy of the output. Machine Vision is one of the keys to automation. The inspection of highly specular or transparent surface of the object in ambient lighting conditions is the limitation of traditional machine vision concepts. Some applications do not require to inspect all the features of a product in detail. To overcome time constraints, only essential parameters of the manufactured object are measured during the inspection. For symmetric 3D geometric shape objects, only their perimeter and height are measured. In this paper, a simple approach is proposed to reconstruct the 3-dimensional model of a highly specular or transparent cylinder from a single image captured in a calibrated environment. The paper informs about experiments using the proposed technique; heights and the diameters are measured accurately for three different size cylinders. Two of them are made of stainless steel, and one of them is transparent. All experiments are performed in ambient lighting conditions. The cylinders are translated in X and Y directions with respect to the camera. The dimensions of the cylinders are calculated and compared for five different poses to check the effects of camera position on the accuracy of the results. In the end, the results are compared with the actual dimensions to check the accuracy of the system.