A decision and communication management methodology for embedded multi-smart camera systems, applied to real-time inspection in lamps production

An industrial vision system must perform a specific and unambiguous inspection. The examination result is crucial to determine if the analyzed item can keep on its transformation process into the production line. So, these systems are strategically installed after those process stations that could have generated a defect in the product. To lower production costs, it is always desirable to detect all failures and get the incorrect pieces out of the line. Implementing this kind of systems is not always simple. Occasionally, special failure features make the use of one single camera insufficient. The solution for multi-camera applications may involve the use and synchronization of several algorithms being executed in different cameras. Also, to fulfill the time constraints, a distributed way for image acquisition and processing is necessary. A Decision and Communication Management Methodology (D&CM-Cam) is presented in this paper. The methodology is designed for embedded Multi-smart Camera systems and consists on a sequence of tasks to assure correct communication among the system elements for a correct and fast decision making. The methodology resides on the smart cameras as algorithms, managing the communication and making the final decision by means of dynamic analysis and classifiers. The technique is described by two diagrams: Static and Dynamic. D&CM-Cam is applied to a case study in lamps production by means of a transparent glass tubs inspection system. Here D&CM-Cam is set for 4 parallel smart cameras to find out 2 different failures at about 17 parts per minute.