The VDI 2632 series of standards provides assistance, harmonising communication between production and image processing planners. Inconsistencies can thus be ruled out from the start. Carefully compiled requirement and system specifications play the key role in this process.
Eliminate inconsistencies from the outset
To automate by means of image processing, you must try to rule out every conceivable inconsistency in advance. To ensure that nothing is overlooked the VDI/VDE Technical Committee on Image Processing in Measurement and Automation Technology has drawn up the VDI/VDE/VDMA Guideline 2632, which shows how clear communication in image processing makes projects successful.
The Guideline provides clear and unambiguous concepts and, with regard to image processing, compiles all the aspects to be considered in a full description of requirements so that orderer and contractor understand one another. A precise and complete project description in the requirement specifications enables all project partners to plan time, cost and manpower realistically and thereby plan the project soundly. It may never be possible to rule out surprises entirely in the course of the project, but they can thus be kept to a minimum. In the final analysis, all of this facilitates completion of the image processing project on time, in accordance with the agreed scope of services and at the price agreed.
Part 1 of the VDI/VDE/VDMA Guideline 2632 lists around 400 individual general concepts, informs about objects and scenes and their illumination and states the differences between individual optics and image sensors. It also specifies which hardware and software is required, which disturbances can occur, and names the essential components and properties of an image processing system. And it does so in German and in English side by side.
The aim is, inter alia, to establish a common understanding among the contractual partners. The top priority is to use generally understandable language, given that users of image processing systems mostly come from a wide range of industries and are not always familiar with image processing concepts.
The schedule for the image processing project
To work successfully on an image processing project, a structured project schedule consisting of the following eight steps may be used:
- Estimate the level of difficulty of the task and of the parts to be tested
- Specify the requirements: requirement and system specifications in accordance with VDI/VDE/VDMA Guideline 2632
- Check the feasibility
- Invite and assess offers
- Decide whether to make or buy
- Carry out the task
- Commissioning and acceptance test of the image processing system
- Establish image processing in servicing, maintenance and qualification
Requirement and system specifications
Of the points listed, the third - specifying requirements - is especially important in practice. Part 2 of Guideline 2632 includes guidelines on drawing up requirement and system specifications. The authors provide specific advice on compiling them and on the influencing factors that both parties must bear in mind when drawing up the system specifications. Recommendations of relevance to image processing are also made for the requirement specifications, always with a focus on the system's specific properties.
The Guideline's benefits are clearly apparent:
Information for the orderer - Makes it clear what the provider means by the technical details he provides.
Checklist for the orderer - What information the provider needs.
Checklist for the provider - Subjects that should be discussed at the beginning of the project.
Query details required for professional project management
Acceptance test of classifying machine vision systems
Part 3 of the Guideline deals with the acceptance test for classifying machine vision systems. It covers inter alia features for differentiating between performance indicators, such as detection or false alarm rates and statements on production statistics. It also describes the effects of an image processing system's suitability for specific tasks.
A current subject under the board's discussion is the compilation of sample catalogues that are to contain answers to various questions such as the following: Can test objects or limit samples be selected according to the significance of their features? Is acceptance of the system to be based on specimen parts or on parts from current production? Which sample sizes are required for the acceptance test? How can statistical certainty be assessed?
published in: inVISION issue 3_2017