View and sensor planning for multi-sensor surface inspection

Abstract

Modern manufacturing processes enable the precise fabrication of high-value parts with high precision and performance. At the same time, the demand for flexible on-demand production of individual objects is continuously increasing. These requirements can only be met if inspection systems provide appropriate answers. One solution is the use of flexible, multi-sensor setups where multiple optical sensors with different fields of application are combined in one system. However, the challenge is then to assist the user in planning the inspection for individual parts. A manual planning requires an expert knowledge of the performance and functionality of every sensor. Therefore, software assistant systems help the user to objectively select the right sensors for a given inspection task. The planning step becomes still more difficult if the manufactured part has a complex form. The implication is that a sensor's position must also be part of the planning process since it significantly influences the quality of the inspection. This paper describes a view and sensor planning approach for a multi-sensor surface inspection system in the context of optical topography measurements in the micro-and meso-scale range. In order to realize an online processing of the assistant system, a significant part of the calculations are done on the graphics processing unit (GPU).

Viewpoint placement for inspection planning

“…2500 and 1300 viewpoint candidates from approximately 131,000 and 107,000 vertex meshes, respectively. Gronle and Osten [13] agree with [11] on generating one viewpoint candidate per mesh vertex. Also, they agree with [12] that using every vertex is redundant.…”

Surface Inspection Planning Using 3D Visualization

“…So far it has been observed from an optimization point of view, researching various approaches to deal with the complexity of the planning problem. [5], [6], [10], [17], [18] focused onto finding a single optimal solution on how to inspect an object, given its 3D model. In simple terms, how should the acquisition system be designed, relative to the object, in order to inspect it completely.…”

“…A set of primary variables, describing the spatial relations of the camera, object and the illumination within the acquisition system, and a set of secondary variables, such as camera lens, exposure time, sensor resolution, illumination brightness and wavelength, etc. [5], [6], [17] and [18] offered approaches to determine viable camera placement. However, by providing both camera and illumination placement, Mohammadikaji et al [10] were the only ones to succeeded in producing an inspection plan containing all primary variables.…”

“…Overlaying several sets of viewpoints may crowd the space around the object, especially given the fact that viewpoint candidate sets may contain up to several thousands of viewpoints (e.g. [6], [11], [18]). Therefore, for each set of viewpoints it is possible to change its color, thickness and opacity in order to achieve maximal distinction between them, as can be seen from Figure 1.…”

Performance Evaluation of a Robot-Mounted Interferometer for an Industrial Environment

“…Additionally, this approach allows for each part to be inspected after each critical manufacturing step, detecting defects early and ensuring constant quality. Typical approaches include machine integrated sensors [1][2][3] and robot-arm-based inspection using camera vision, laser scanning or 3D pattern projection [4][5][6][7][8][9][10] (also see industrial systems such as: X4 i-Robot, Creaform Cube-R, GOM Atos Scanbox, Mertolog X4), enhanced by multi-sensor and data fusion approaches [11][12][13][14][15][16][17][18]. Allowing for fast and flexible inspection of free-form surfaces, these systems commonly lack the ability to measure sub-millimeter features with high resolution, as the sensors required for this step are typically large, heavy and require laboratory conditions and are therefore used off-line.…”