Current trends such as Industry 4.0 and the smart factory are driving forward the automation of production and paving the way for new robotics and manufacturing concepts. According to the VDMA, the German robotics and automation industry generated a record sales volume of EUR 12.2 billion in 2015 - an increase of 7% compared to the previous year. And for 2016, the VDMA is forecasting a further increase of 2% and thus a new sales record of EUR 12.5 billion.

Advancing automation in the industrial environment not only requires innovative robotics solutions, but also the continuous further development of corresponding accompanying technologies. These include handling and transfer systems as well as recognition processes such as industrial image processing. It uses digital images to identify all conceivable objects involved in the process. The image material is fed in by image acquisition devices that are installed at different locations and monitor the production process. The image data is then processed using special software. In milliseconds, the algorithms compare the captured images with the information stored in the database so that even large product series can be inspected quickly and automatically. Image processing methods are therefore essential in order to automate, accelerate and make production processes more efficient across the board.

Fit for Industry 4.0

Industrial image processing methods also check the correct composition of food.

© MVTec

At the same time, industrial image processing must flexibly meet the requirements of digitally networked, integrated production processes - keyword Industry 4.0. It is important here that corresponding applications can be created quickly and in a user-friendly manner. This is becoming increasingly important, especially in the context of new robotics scenarios. One example: factory halls are currently being conquered by a new generation of industrial robots that are significantly smaller, lighter and more flexible than their predecessors and work closely with their human colleagues. In addition, the robots are mobile and can take on changing tasks - for example, they can help out on the assembly line for a short time to carry out the work of a sick employee. For this purpose, the machines must be individually set up and calibrated for each application. User-friendly handling shortens set-up times and reduces effort and costs.

Applications without programming codes

In terms of machine vision solutions, there is a clear trend towards the simple and user-friendly creation of corresponding applications. The 'Merlic' software from MVTec, for example, does not require any in-depth knowledge of image processing and programming, yet still allows the creation of professional machine vision applications.

No in-depth knowledge of image processing or programming is required to work with the 'Merlic' software.

© MVTec

The core of the software solution is a transparent, image-centered user interface that guides the user intuitively through the application. Instead of complex codes, command lines or parameter lists, the focus is on a clear visual representation that resembles a What-You-See-Is-What-You-Get (WYSIWYG) editor. Standard vision tools such as recording, calibration, alignment, measuring, counting, checking, reading, position determination and error detection are also integrated.

Usability is also enhanced by a process called 'Easytouch'. It allows objects to be identified, marked and selected with a single click by simply moving the mouse over an image. This eliminates the need for time-consuming configuration of complex parameters, which in turn simplifies the creation of applications. By using the software, production companies save time and effort when programming machine vision applications and thus increase the usability of their set-up processes. In addition, employees without in-depth programming and image processing skills can create the applications.

In order to increase the usability of machine vision solutions, they must support the customer's processes precisely and address their individual requirements precisely. MVTec uses the user-centered design approach, for example, to ensure that its products are as user-oriented and tailored to the target group as possible: specific pilot customers are involved in the development process right from the start. In close cooperation with five traditional representatives of the peer group, features are developed that take into account the specific workflows and requirements of the target group.

User-Centered-Design

In this way, feedback from customer representatives flows directly into the solution - and the user gets exactly what they need for efficient image processing in their company. The machine vision provider also carries out usability tests with potential users as part of the software development process. Just a few tests are sufficient to obtain meaningful results. As a rule, over 50% of weak points are identified in just three tests.

The integration of functions that increase the degree of automation of processes also improves the manageability of image processing solutions. In the current version 13 of the image processing software 'Halcon', for example, MVTec has integrated mechanisms for better detection of texture errors. This allows product deviations in objects with a texture - for example textiles, leather or carpets - to be reliably identified by means of an image-based comparison with intact materials. What's special about this is that the process is automated and requires no additional programming effort on the part of the user.

Recognize text and characters reliably

Image-based, semantically independent dot-print characters can be reliably recognized.

© MVTec

Another requirement of many companies is the reliable recognition of text and characters (Optical Character Recognition, OCR). User-friendly machine vision solutions are also available here. While OCR systems in the office environment usually have dictionaries that can be used to extract incomplete or difficult-to-read text from the context, in industrial environments it is generally not coherent text passages that need to be recognized, but rather isolated type plates, batch or serial numbers - a challenge for image processing. Every character, whether letter or number, has to be read correctly here; it is not possible to derive it from the semantic context. In addition, it is usually not clear exactly where the characters are located, as the reading algorithm cannot be based on a predefined line spacing. 'Halcon' provides functions for this purpose that accurately locate the text in the captured image. Among other things, characters are automatically merged from individual pixels, which increases the recognition rate and thus usability in the event of unclean printing.

Author: Dr. Wolfgang Eckstein is Managing Director of MVTec Software in Munich.