Machine Vision based Quality Control in Food IndustryMaster Automation Group Oy specialises in integration of flexible robotics and machine vision. The company has developed flexible robot module including vision based robot guidance system with quality control features. Machine vision material handling combined with conveyor tracking has already been a proven technology for couple of last years. Next logical step in flexibility is to fulfil online quality control needs in high speed picking applications. Master Automation Group Oy has developed robot guidance software, MAG Robot Guidance, which combines multipurpose robotics vision with easy-to-use quality control features. MAG Robot Guidance is PC-based system, which can handle, not only standard resolutions (CCIR, RS-170), but also high-resolution non-standard camera formats (e.g. 1024 x 1300 pixels). This is a great advantage in cases where quality control is involved. Quality inspection needs in food industry can be categorised in shape analyses, colour recognition, label inspection and general grey scale analyses to analyse surface quality. Quality inspection, in general, is very demanding field for vision. Analysing quality of target can be much more complicated task than simple position detection or dimension analysis. It is very easy to set-up tolerances for dimensions and position information, but classification to define product quality based on the information how the products look needs more sophisticated algorithms.
High performance data exchange capabilities and support for different communication protocols trough Ethernet is needed to handle increased amount of measured information. MAG Robot Guidance system has standard interface modules for all the biggest robot brands.
 Picture 1. Structure and data flow of MAG Robot Guidance system Information received from classification module can be simple accept/reject. In many cases targets need to be classified in several categories like A, B, C and D. In cases where amount of details measured to define quality is big (>10), neural based classification algorithms are needed. Master Automation Group has used neural network based classification since 1994 on its applications.
Example applications where quality control could be used
Picture 2. Analysis of Carelian pie based on grey scale correlation.. Upper histogram shows the result of a pie having darker surface area than sample below.
Picture 3. Examples of biscuit classification. In sample A, targets having defects are rejected. In sample B all the targets has been selected without any quality control. In sample C, inspection of the middle jelly dot is included to reject targets outside quality tolerances. Additional software packages are also included, which enable rapid prototyping as well as fine tuning of the robot program and process parameters. The quality inspection functionally utilises most of the common image analysis tools and is capable of detecting various problems including: size and shape variations, assembly defects, printing errors and surface quality flaws.
Master Automation Group Oy is a Finnish company with special expertise in robotics, machine vision and automation. Master Automation Group is the leading supplier for robot guidance applications with global references.
|
