Ventil 1 / 2022 • Letnik 28 46 OPTIČNI PREGLED IZDELKOV 46 1 Introduction In the modern smart factories, automated contac- tless inspection methods continuously gain ground over human inspection (being subjective) and tac- tile methods (suffering from slow data acquisition and interpretation). Optical inspection can be ap- plied at different stages, from the quality control of freshly produced components to the posterior check of the moving parts’ wear [1]. Within the con- cept of Industry 4.0, not only do smart optical in- spection tools collect images, but they also imple- ment statistical process control and communicate with other devices in real-time. Their implementa- tion in the production line detects flaws and defec- ts, providing information about the quality and the uniformity of the manufacturing process in real-ti- me. The harvested data can be further analyzed to enable predictive maintenance of the tool itself and reduce downtimes [2]. The quality of optical inspection is influenced by the specific and often harsh conditions met in an industrial environment, which may include variati- ons in the illumination, contaminations and vibrati- ons. Light conditions, in particular, are reported to have a remarkable impact on the captured image quality [1]. This impact can be quite complex in the vicinity of flaws and defects. For example, a smo- oth lacquered surface is specular and reflects at an equal and opposite angle to the incident light. However, regions that exhibit flaws, defects and va- riable roughness reflect at additional directions. Depending on the industrial application, different wavelengths λ of light can be utilized. Ultraviolet light (100 nm < λ < 400 nm) is typically chosen for the inspection of glues, thin films and semicon- ductors. White or colored light across the visible spectrum (400 nm < λ < 750 nm) is often applied for metals and objects with multiple color surfaces that require high color rendering images. Further- more, near and far infrared (750 nm < λ < 1 mm) is mostly chosen for optical inspection of semi-tran- sparent materials and natural fibers, as well as for thermal imaging. We hereby present a short overview of measure- ments, performed under different light conditions, in order to detect flaws and defects on the surfa- ces of lacquered metal components. These me- asurements aim to assist in choosing the optimal light conditions for reduced surface reflectance and enhanced image clarity, addressing one – out of many – challenges related to the development of an advanced optical inspection tool EAGLE. The design of this tool is customized for the inspection of metal components produced for the automotive industry [3]. Research Fellow George Cordoyiannis, PhD, Se- nior Lecturer Iris Fink-Grubačević, MSc, both Faculty for Industrial Engineering Novo mesto; Assistant Prof. Tomaž Savšek, PhD, TPV Group d.o.o. & Faculty for Industrial Engineering Novo mesto t esting the imP act of light conditions on image qualitY for oPtical insPection of surface defects George Cordoyiannis, Iris Fink Grubačević, Tomaž Savšek Abstract: Contactless and automated optical inspection tools make headway in comparison to tactile methods in modern factories. In order to develop an advanced optical tool for inspection of metal components pro- duced for the automotive industry, one faces a number of challenges. We hereby present a brief overview of the image quality of surface flaws and defects obtained under different light conditions. Lacquered metal components of simple and complex geometries have been illuminated by light of variable intensity. Certain trends have been revealed regarding the obtained image quality. Keywords: optical inspection, surface defects, illumination Ventil 1 / 2022 • Letnik 28 OPTIČNI PREGLED IZDELKOV 47 2 Experimental setup and samples Light emitting diode (LED) sources are reported as superior to other types of light sources for achie- ving an optimum defect expressivity on flat, cylin- drical or complex geometry reflective metal surfa- ces [1, 4], such as the ones of our interest. Based on the above, a LED array source has been chosen to illuminate the components. It produces warm- -white light with a correlated color temperature of 3000 K. Our measurements have been planned and performed by means of a simple, house-in-built se- tup in order to test the impact of light conditions on the captured image quality around defects. This setup consists of the following parts: (a) a Nikon digital camera equipped with Tamron 90 mm f/2.8 Di Macro Lens, connected to a laptop for a direct viewing of the captured images (b) a Cree X-Lamp CA2550 LED array source (Cree EasyWhite® LEDs, uniform chromaticity profile), driven by an external amplifier - intensity controller, (c) tripods that offer a stable placement and a three-dimensional mo- vement of the camera, LED source and inspected component. This setup serves as a simple approxi- mation of the more sophisticated EAGLE optical tool, where a robotic arm would be used to collect the components from a moving conveyor and place them properly with respect to the camera and the light source. Cylindrical or complex geometry metal compo- nents bearing defects have been provided by TPV Group d.o.o. Measurements have been performed for vertical and oblique placements of the came- ra and the LED source with respect to the surface defects. Images with a resolution of 3696 x 2448 pixels have been captured under different ambient conditions, such as in a dark room (henceforth re- ferred to as “dark ambient”) and in a candent room with Neon light from the ceiling (henceforth refer- red to as “bright ambient”). In these dark or bri- ght ambient conditions, the inspected components have been illuminated by LED light of variable in- tensity, in the range from 50 lm to 2000 lm. The intensity has been derived by accurately measuring the source driving current and then referring to the luminous flux-current chart on steady-state ope- ration. The distance between the LED source and the components has been set to 20 cm and 10 cm when observing defects on the outer and inner su- rfaces, respectively. 3 Results and Discussion Images of defects on black-lacquered external su- rfaces of a cylindrically-shaped metal component have been captured under different light conditi- ons. The presence of low intensity LED improves the image quality obtained in dark ambient. The impact of variable LED illumination on the probed defect morphology is depicted in Figure 1. The two images are obtained by illumination intensities dif- fering by an order of magnitude, 100 lm and 1000 lm, respectively. For 100 lm, a sharper and more de- tailed image is obtained, in comparison to the 1000 lm. In general, images above 500 lm suffer from increased surface reflectance. Further increase of the intensity up to 2000 lm, degrades the image quality derived in both dark and light ambient con- ditions. In the case of a bright ambient, the effect of low intensity LED is milder, whereas a high intensity produces once again strong reflections. These tren- ds have been steadily reproducible for other similar defects on black-lacquered surfaces. Defects on the surface of components with a more complex geometry have been also observed. The LED source and the camera have been placed ei- ther vertically above the inspected area or at oblique positions. Optimum results have been ob- tained for camera and LED source as close as pos- sible to the vertical position above the inspected area. Representative images captured in a dark or a bright ambient, as well as their combination with low intensity LED are shown in Figure 2. This com- Figure 1 : The image of a defect on the external surface of a lacquered metal cylindrical object is obtained in: dark ambient + LED illumination of 100 lm (panel A); dark ambient + LED illumination of 1000 lm (panel B). The 100 lm image is characterized by reduced reflections and enhanced details of the defect morphology. Ventil 1 / 2022 • Letnik 28 48 ponent consists of a black-lacquered metal part with an attached non-polished and less reflective ring. It bears defects in the form of small caviti- es, grooves, bulges, as well as increased surface roughness towards the edges, which are marked by green arrows in one of the panels of Figure 2. In the case of a dark ambient, the LED illuminati- on highlights minor defects and reveals the sur- face roughness, which is hardly visible otherwise. Simultaneously, it weakly increases the reflectan- ce, which does not degrade the image quality up to 200 lm. Higher intensities, in the range from 500 lm to 2000 lm, are accompanied by a signi- ficantly increased reflectance. In the case of a bright ambient, the image clarity exhibits only a minor improvement due to LED illumination, mo- stly in revealing the surface roughness. The three- -dimensional shape of small cavities and grooves is better distinguished in the case of a bright ambient compared to a dark one. Regarding the placement of the camera and LED source, the re- sults are of reasonably good quality for angles up to 30° with respect to the inspected surface. In some cases, the tiny bulges are better visible in case of strong illumination; however, the increa- sed reflections degrade the overall image quality. The detection of defects on the internal surfa- ces of cylindrically-shaped components is more challenging. In this case, the inspected compo- nent has been attached on a clamp (simulating the robotic arm in the configuration of EAGLE tool) between the camera and the LED source that are facing each other as schematically de- picted in Figure 3. There exist no essential dif- ferences between images obtained with and wi- thout low intensity LED in a bright ambient, as it can be seen in the panels of Figure 3 (note that a dark ambient did not produce any good ima- ges in this case). Apart from a parallel position with respect to the object’s cylindrical axis, the camera has been also placed at small inclinations (from 2° to 8°) in an effort to better highlight the surface morphology of specific defects. However, the inclined camera placement did not offer any remarkable improvement. Contrary to the case of defects on the external surfaces, in the case of in- ternal surfaces the LED illumination did not show any visible effects. The aforementioned trends, as revealed in Fi- gures 1, 2 and 3, have been fully reproducible for several types and sizes of defects on diffe- rent components. Although only representative examples have been presented, the overall tren- ds are supported by a large number of measu- rements, thus, they are robust. The robotic arm seen in Figure 3 is used to select one every few components, as they move along the conveyor, and perform a sampling inspection. Via this end- -of-line testing, the EAGLE smart optical tool performs a pass-or-fail check and places the component back to the conveyor or disposes of it in case it is faulty [3]. At the same time, it can send feedback about faulty components to the production line, preventing downtimes and increased costs. However, it is worth noting that the implementation of an optical inspection tool near the production line requires the dumping of vibrations originating from the adjacent indu- strial machinery. OPTIČNI PREGLED IZDELKOV Figure 2 : Images of a complex geometry component obtained in: dark ambient (panel A); dark ambient + LED illumination of 100 lm (panel B); bright ambient (panel C); bright ambient + LED illumination of 100 lm (panel D). The green arrows (panel D) mark the diffe- rent types of flaws and defects: small cavities, bulges, grooves and increased surface roughness. Figure 3 : The configuration of the camera and light source for capturing images of defects on the inner surfaces is depicted (panel A). The presented images have been captured in: bright ambient (panel B); bri- ght ambient + LED of 100 lm (panel C). Ventil 1 / 2022 • Letnik 28 4 Conclusions Flaws and defects on metallic surfaces have been observed under different light conditions by means of a simple, house-in-built setup. The image clari- ty has been evaluated for different conditions, in a dark or bright ambient, as well as with and without LED illumination. Certain trends have been revea- led that could be summarized as follows. First, LED illumination of low intensity (100-200 lm) is bene- ficial in the vast majority of cases; it yields sharp and detailed images of the defects’ shape and morphology, accompanied by a weak and non-di- sturbing reflectance. Second, higher intensity LED illumination (500-2000 lm) is apparently benefici- al for capturing the shape of tiny bulges; however, it persistently causes strong light reflections from the lacquered surfaces and downgrades the overall image quality. Third, the shape and morphology of defects is effectively captured by camera angles up to 30° with respect to the inspected surface. Fo- urth, in the particular case of defects on the inner surfaces, LED illumination at an angle of 180° with respect to the camera did not offer any improve- ment of the image quality. The presented measurements have been performed aiming to assist the design of a non-contact optical tool EAGLE for inspection of metal components that are produced for the automotive industry. For the latter, dimensionally-stable and defect-free com- ponents are of major importance. Car spare parts must fit tightly and prevent any type of mismatch and malfunction, leakage of oil, intake of moisture and development of rust, as well as disturbing sque- aking and rattling noises [5]. At the same time, the detection of defective components via a selective inspection of components along a moving conveyer can give feedback to the production line at an early stage, thus, reducing downtimes and costs. Sources [1] Ramzi, R., Bakar, E.A.: Optical wear inspec- tion of countersink drill bit for drilling oper- ation in aircraft manufacturing and assembly industry: a method, IOP Conference Series: Materials Science and Engineering 370/2018, pp.: 012041(1-8). [2] Noureddine, R., Solvang, W.D., Johannessen, E., Yu, H.: Proactive learning for intelligent maintenance in Industry 4.0, Advanced Man- ufacturing and Automation IX, pp.: 250-257 (Springer Singapore, 2020). [3] Petrič, A., Kurbegović, H., Savšek, T.: Quality challenges and state of the art in designing development requirements for the system EA- GLE, 6. International Conference on the Devel- opment of Industrial Engineering: Opportuni- ties, Potentials, Challenges, Novo mesto, 2021. [4] Li, L., Wang, Z., Pei, F., Wang, X.: Improved il- lumination for vision-based defect inspection of highly reflective metal surface, Chinese Optics Letters 11/2013/2, pp.: 021102(1-4). [5] Bogue, R.: Car manufacturer uses novel laser scanner to reduce time to production, As- sembly Automation 28/2008/2, pp.: 113-114. 49 OPTIČNI PREGLED IZDELKOV Preizkušanje vpliva svetlobnih pogojev na kakovost slike za optični pregled površin- skih napak Razširjen povzetek: Razvoj brezkontaktnih in avtomatiziranih orodij za optični pregled predmetov napreduje v primerjavi s taktil- nimi metodami v sodobnih tovarnah. Da bi razvili napredno optično orodje za pregledovanje kovinskih kom- ponent, se soočamo s številnimi izzivi, med njimi tudi z izbiro optimalnih svetlobnih pogojev. Predstavljamo kratek pregled meritev kakovosti slike površinskih napak in defektov, pridobljenih pri različnih svetlobnih pogojih. Namen izvedenih meritev je pomoč pri razvoju orodja za optične preglede EAGLE, ki se uporablja za pregled kovinskih komponent proizvedenih za avtomobilsko industrijo. Meritve pomagajo pri izbiri opti- malnih svetlobnih pogojev za zmanjšano površinsko odbojnost in večjo ostrost slik. Preprosto ali kompleksno oblikovane lakirane kovinske komponente so bile osvetljene z LED-svetlobo spremenljive jakosti. V kakovosti slike se opazijo določeni trendi. Kar se tiče napak na zunanjih površinah, je LED-osvetlitev nizke intenzivnosti v veliki večini primerov koristna; daje ostre in podrobne slike oblike in morfologije napak, ki jih spremlja šibka in nemoteča odbojnost. V primeru napak na notranjih površinah LED-osvetlitev ni prinesla nobenih izboljšanih rezultatov. Ključne besede: optični pregled, površinske napake, osvetlitev Acknowledgements This research work has been partly funded by the European Union, from the European Regional De- velopment Fund under the Operational Programme for Investments in Growth and Jobs for the pro- gramming period 2014 to 2020, under contract no. C3330-18-952007 (EAGLE). G.C. acknowledges access to the laboratory facility of Uroš Tkalec (Faculty of Medicine, University of Ljubljana).