Customer SolutionsInnovative Visual Quality Control of Wood-Wool Slabs Using NI LabVIEW and IMAQ Vision
Author(s):Christian V. Madritsch, Carinthia Tech Institute; Dr. Thomas Klinger, Carinthia Tech Institute
Industry:Industrial Controls/ Devices/ Systems, University/Education
Product:LabVIEW, Vision
The Challenge:Combining different image processing and analysis methods into a visual quality control system, resulting in a reliable and fast detection of heterogeneous and polymorphic faults of wood-wool slabs.
The Solution:Using National Instruments hardware and software products in all phases of system design: NI Vision Assistant to develop the image processing and analysis methods, Vision Builder for Automated Inspection software and IMAQ Vision to include control and decision making functionality, and Compact Vision System to deploy the visual quality control system for the customer.
Image Processing and Analysis Problem Mastered Using NI Vision Assistant Wood-wool lightweight construction slabs (WWS) are made of the natural raw materials wood and cement. They are used as universal construction and insulation materials. During production, three types of faults appear at the surface of the slabs: rough edges, stains/holes, and horizontal/vertical stripes. The faults do not affect the construction or insulation properties but do affect their visual appearance. The appearance of the faults varies and each slab might contain more than one fault type. At the end of the production process, quality control distinguishes between slabs with acceptable faults and slabs that need to be sorted out. The identified fault types are in general heterogeneous and polymorphic, fundamentally requiring different image processing and analysis methods. During the requirements and development phase, we used NI Vision Assistant in combination with photographic images of the slabs and later with a video camera capturing live images to develop basic methods for the processing and analysis. During the implementation phase, we used the NI Vision Assistant results in either Vision Builder for Automated Inspection software or in LabVIEW together with LabVIEW Vision Development Module. The seamless transition between the different tools offered a fast and stable development cycle. At the deployment phase, we used the Compact Vision System on the customer site as a reliable and robust hardware platform. Three Different Image Processing and Analysis Problems Combined The first fault type we studied was rough edges. During production, the edge of a slab might break. A major issue to reliably detect rough edges using image processing is the correct illumination. An illumination of the slab in parallel to the optical axis of the camera does not highlight the edge in the resulting image. Therefore, it is necessary to illuminate the slab at a 45 degree angle relative to the optical axis. Due to the formation of shades, the broken edge is ideally visible in the resulting image. The image processing chain starts with the extraction of the green plane. We selected the green plane due to the high green portion in the light source. The next step is a brightness and contrast adjustment of the image. We need to increase both brightness and contrast to highlight the edge shades. During image analysis, we use the find straight edge (FSE) tool of the Vision Builder for Automated Inspection software to search for the edge. The edges of interest are vertically oriented. Therefore, we configured the FSE tool to search from right to left and vice versa. The decision if the fault is acceptable or not is made using the allowable deviation limits setting of the FSE tool. To detect stains or holes on the slabs, we used a completely different method. We compare the density of different areas of the slab with each other. If there are no significant stains or holes, the density should be close to equal. The density is measured indirectly using the pixel count. The image processing begins with the extraction of the green plane. Afterward, a threshold operation (entropy) followed by a binary inversion is applied to the image. Finally, the image is processed by a lookup table (equalize). The image processing fragments the surface of the slab into objects of small size. In the image analysis, we use the count pixels tool to count the objects in different regions of the image. Using statistical methods (mean value, standard deviation), the output of the pixel count operation is processed and the results indicate the number of stains or holes. Example Method: Detection of Horizontal and Vertical Stripes The front panel of the WWS quality control system detects horizontal and vertical color stripes, and is defined to be the third fault type. As a first step, the image is processed by a convolution filter (smoothing) to reduce noise caused by the camera as well as the impact of small objects (holes), which are not of interest in this case. Due to the fact that color stripes can only appear in horizontal or vertical orientation, a grid of line profiles is placed on the image. Horizontal and vertical line profiles are considered separately in the next steps. It is evident that the line profiles perpendicular to the color stripes cause a higher standard deviation of the line profile values than line profiles parallel to the color stripes. Therefore, the value of the standard deviation average in horizontal direction is subtracted from the standard deviation average in vertical direction, which leads to a value significant for the stripe strength with its absolute value and significant for the stripe direction with its algebraic sign: if the value is significantly greater than zero, the slab has horizontal stripes; if it is significantly smaller than zero, the slab has vertical stripes. A small value around zero (no matter if positive or negative) indicates that the slab color is evenly distributed. Automated Quality Control Saves Money Using this visual control system, the quality control of WWS has been fully automated. The system is extensible and easily adapted to changing requirements. The customer is entirely satisfied and saves up to $30,000 per year. For more information, contact: Christian V. Madritsch Senior Lecturer Carinthia Tech Institute Europastrasse 4 A-9524 Villach Austria, Europe Tel: +43 4242 900 500-2127 Fax: +43 4242 90 500-8 2127 E-mail: cm@cti.ac.at |
