12/20/2021 IDS Imaging Development Systems GmbH
You always need them and you see them everywhere: in the metalworking industry, on the job site, in almost any toolbox … the good old hand brush. Whether classic all-purpose brushes or brushes for special applications, the German company Lessmann GmbH has practically all of them in its range. The shape is varied, from straight to ergonomic. But they all have one thing in common: the wooden bodies are made exclusively from untreated red beech. Depending on the model, they also have two hooking holes at the end of the handle. The production is fully automatic and highly rational. In order to guarantee its own claim to the excellent quality of hand brush antlers, Lessmann has relied on classic image processing for many years. But now “The German Brush Company” has implemented an image processing system from the Bavarian systems company Simon IBV GmbH which uses rugged IDS industrial cameras and SIMAVIS® H image processing software to detect even the barely noticeable tolerance deviations in a particularly reliable way.
The brush, which is crushed fully automatically at a production rate of 1500 pieces per hour, is extracted from the milling machine by a timed circulation chain with quiver-shaped receptacles and pushed onto a longitudinal conveyor belt. A multi-camera system is installed on the conveyor belt, which checks 2-6 rows hand-brushed timber for defects such as cracks, chips and dimensions. “The testing task is particularly demanding because untreated copper beech varies considerably in color and grain. For example, cracks cannot always be clearly distinguished from dark grains ”, explains Daniel Simon, authorized signatory at SIMON IBV. But the choice of wood species has good reasons: on the one hand, red beech is recommended for the production of hand brushes due to its excellent properties, such as a particular degree of hardness. On the other hand, sustainability plays a major role. Lessmann can obtain brush supplies in the surrounding area and thus both support regional forestry and avoid transport routes.
While the timber passes through production on a conveyor belt, a total of four IDS cameras of the type GigE uEye FA are triggered by an incremental encoder. This sensor reacts to the position of the belt so that any change in the position of the brush body is detected by the movement of the belt. Image capture is offset 2.5mm per camera, so each camera takes a new image every 10mm. The captured images are discarded until the first camera detects that there is a wooden body in the field of view. From this moment, the other three cameras are activated and up to 35 photos are taken per camera. The number of images is limited by camera 1, because it comes out as soon as no more brush body is visible.
The images captured by IDS cameras are preprocessed and composed simultaneously with the image capture. Thus, during the evaluation time, the acquisition of the image as well as the pre-processing of the next brush can already take place. Individual images of the same situation from the four shifted cameras are cropped, scaled, and merged into an overall image by the software. Brush bodies are evaluated with criteria weighted differently for each camera. The weighting is done via the evaluation criteria test sequence. In a first step, coarse geometric factors such as length, width, height, symmetry and shape deviations are evaluated.
Illustration of the shape test: are the exterior dimensions of the brush body within tolerance? Is the brush body asymmetrical or deformed? Are the holes the correct diameter and position?
The position and overlap of the holes in the brush body are checked, followed by a step-by-step surface inspection.
Surface inspection: are dark or colored areas allowed? Are there any rough spots or cracks?
“First, the dark areas are segmented and evaluated according to the setpoint specifications,” explains Matthias Eimer, system integrator at Simon IBV. “After that, divergent discolorations are searched for, identified and evaluated according to the target specifications. »Even the tolerances for the roughness can be defined in the target values and are then evaluated. It is only at the last step of the frame-by-frame evaluation that the cameras look for cracks. Finally, the overall result is formed and merged from the individual assessments of the points of view. The system checks a total of 32 set points, 27 of which alone for adherence to precisely defined tolerance specifications.
Dark spots are segmented and rated based on set point specifications
The used uEye cameras of the FA family are particularly robust and therefore predestined for use in an environment as severe as the brush factory. Camera housings, lens tubes and screw connectors meet the requirements of protection class IP65 / 67. They are also ideally suited for the multi-camera operation required here due to the built-in image memory, as this decouples image capture from image transmission and allows images to be buffered before transmission into this. application. GV-5280FA industrial cameras with GigE Vision firmware are equipped with Sony’s IMX264 2/3 “Global Shutter CMOS sensor, which also offers excellent image quality, high light sensitivity and exceptionally high dynamic range. CMOS cameras used produce almost noise Free, high contrast 5 MP images With exactly these characteristics, the camera model recommended itself for use in demanding brush tests.
“The camera has the right resolution, the Sony sensor is very good and the protection class is met,” explains Daniel Simon, summarizing the selection criteria.
IDS cameras are easy to integrate, true to the company motto “It’s so easy”, as Daniel Simon knows from his many years of cooperation with IDS. The heart of the solution is the software specially developed by the solution provider.
Output of results
SIMAVIS® H is an image processing software with which complete solutions can be quickly assembled. This machine vision software is based on ProVision® (former SIEMENS development) and HALCON, a complete standard software for processing industrial images with an integrated development environment. This allows individual adaptation to the test requirements of wooden brush bodies: “The fault check was programmed by us individually by hand, we have a lot of experience in the area of wooden surfaces. The tolerance thresholds can be adjusted by the many set points depending on the functionality ”, emphasizes Daniel Simon. SIMAVIS® H offers an intuitive user interface for the operating personnel of the finished system, including manual and automatic operation with type management, set point menu, authorization concept, language change, statistics module and much more. Moreover.
“Thanks to the control, automatic further processing is possible. Thanks to the new imaging solution, wood defects can now be detected more reliably. The proportion of defective wood bodies classified as good fell from around 2% to less than 1%. explains Managing Director Jürgen Lessmann. The subsequent filling of the brush bodies with wire is already fully automatic and the packaging of the finished hand brushes can be carried out in the future using robotics. Due to the improved control of the wooden bodies, the previously required manual quality control of the finished hand brush can be omitted.Before packing, which is carried out by a SCARA robot, only a brief visual inspection is required. machine personnel and increases productivity.
In this exemplary application, the use of artificial intelligence also offers additional potential in the future. “AI will probably further improve the inspection result, which will allow further automation of production, as manual inspection operations can be eliminated,” predicts Jürgen Lessmann. And not just for brushes made of good wood: the multi-camera test system can be adapted for countless products and materials.
Request FREE information from the supplier on the products in this item
Process and Control Today is not responsible for the content of articles and images submitted or produced externally. Click here to email us regarding any errors or omissions in this article.