Industrial image processing applications as orchestration of web services

Contemporary production lines often make extensive use of image recognition to achieve high levels of automation. Shorter product life cycles and smaller lot sizes force companies to adapt their processes very quickly. Still, industrial image processing algorithms typically run on dedicated computer systems next to the cameras, which is rather inflexible and leads to high costs for hardware and maintenance. In this paper, we propose a service-based architecture to tackle these problems. We divide image processing applications into universal building blocks, each of which is implemented as web service. The recognition application is conceived as a composition of these services, orchestrated using the Business Process Execution Language (BPEL). The composed application is provided with a standard industrial interface (OPC UA) for easy integration into the production process. This concept is evaluated by assessing proof-of-concept implementations. Main advantages (among others) are the flexible reuse of existing image processing operations and applications, and the freedom to implement services independently using the preferred language, framework and platform. Challenges are real-time constraints, which can be alleviated by emerging technologies like Time Sensitive Networking (TSN).