The most significant area of use for machine vision systems is in line as part of the production process. Online inspection systems acquire images of products at machine speed and are immediately analysed by the system, providing instant results at the speed of the manufacturing process. These on-line systems are also called real-time systems and depending on the application, the result can be a simple pass/fail, or more complex information in production analysis and control applications. Depending where positioned in the production process an inline vision system can simply reject a defective part ensuring defective parts never reach the next stage of the production process where it may create additional cost, or if positioned at the end of the line can prevent defective products reaching the final customer.
In process control applications measurement information from the vision system can provide either feedback to the process upstream, ensuring manufacturing tolerances are kept within specification, or downstream in applications such as robotics, where location size and orientation can accurately guide a robot thus simplifying mechanical handling.
Offline systems in contrast do not provide instantaneous result data and are mainly used for recording information for later analysis. These systems include high-speed recording and analysis of special processes, or data collection in scientific experiments. These are termed offline systems as there is no real-time decision or data transmission needed.
Verifying that a product, assembly or package has been correctly produced is a major application for on-line vision systems. Applications range from pure presence checks (for example ring pulls on cans) to checking that package seals are correctly placed. The verification ofprinting accuracy is another example of this type of application.
Other example applications include:
The accurate measurement of component dimensions ensuring they are within predefined manufacturing tolerances is an extremely important use of vision. For these applications image quality and resolution is as important as the algorithm, as only the combination of subpixel accurate measurement tools with suitable lenses and the correct illumination enable modern vision systems to deliver precise, resilient and repeatable measurements. For measurement applications particular care needs to be taken to ensure machine vibration and product location does not cause incorrect errors.
Flaws such as contamination, scratches, cracks, discolouration or burn marks may be caused by slight variations in the manufacturing process and manifest themselves as minute variations in the form of the item, but can have severe effects. These types of defects usually appear at random intervals and in varying positions on the product, therefore flaw detection tools have to look for changes in colour or texture and sometimes have to compare the differences between the object under inspection and a known 'golden template'.
A few examples are shown here: