State of the art of 3D scanning systems and inspection of textile surfaces

Abstract

The rapid development of hardware and software in the digital image processing field has boosted research in computer vision for applications in industry. The development of new electronic devices and the tendency to decrease their prices makes possible new developments that few decades ago were possible only in the imagination. This is the case of 3D imaging technology which permits to detect failures in industrial products by inspecting aspects on their 3D surface. In search of an optimal solution for scanning textiles we present in this paper a review of existing techniques for digitizing 3D surfaces. Topographic details of textiles can be obtained by digitizing surfaces using laser line triangulation, phase shifting optical triangulation, projected-light, stereo-vision systems and silhouette analysis. Although we are focused on methods that have been used in the textile industry, we also consider potential mechanisms used for other applications. We discuss the advantages and disadvantages of the evaluated methods and state a summary of potential implementations for the textile industry

Automatic Defect Classification on a Production Line

“…This brings the overall error down to 2.33 % on the average (Fig. 20); the error goes to zero for the elongated defects, and almost to zero for all the other classes (Table 3) 20 The classification error for the training (discontinuous line) and test (continuous line) images is reported for a typical set of experiments in which enhanced features, enhanced multi-label boosting and the first two labels are considered over 129 has been miss-classified. As test sets are constituted of 43 images, randomly extracted at each experiment from the whole data set, this means that one insect image was wrongly classified in a few experiments.…”

“…For this reason, systems for automatic defect classification (ADC) have been recently introduced. These are usually based on a hierarchical two-level structure [19,20], in which the first level extracts a set of features from the image of the potential defect and the second level classifies the image and, in case of dangerous defect, triggers a procedure for defect elimination.…”

Fusión de series temporales de nubes de puntos para su exploración en realidad virtual.

“…El problema del registro, aunque viene facilitado por la asistencia que ofrece el software propio del escáner 3D, requiere de la aplicación de algoritmos tales como el clásico "Iterative Closest Point" [4], ICP, o algunas de sus variantes [5].…”

Integrating Laser-induced breakdown spectroscopy and photogrammetry towards 3D element mapping

“…Provided we have a working solution for 2D elemental mapping with LIBS, deploying a 3D visualization solution requires both a convenient 3D modeling technique and a properly designed workflow for integration of the two techniques. Regarding 3D scan techniques, the most common solutions are laser scanning, LIDAR scanning, and photogrammetry [3]. The first two solutions can deliver precision models even at the microscale, making them the go-to solution for highresolution applications [3].…”

“…Regarding 3D scan techniques, the most common solutions are laser scanning, LIDAR scanning, and photogrammetry [3]. The first two solutions can deliver precision models even at the microscale, making them the go-to solution for highresolution applications [3]. Yet, these often require expensive dedicated hardware that drastically increases the cost of the setup.…”