Optimized Parallel Software Architecture Design for Industrial Materials Sorting Systems

The introduction of new technologies in the industry makes possible to develop new functionalities in production lines, but it requires to solve some challenges in several areas.One of these challenges is the hardware and software design of the computational system, taking into account both the time requirements of the industrial system and the economic requirements.In the field of sorting and collecting pieces of different composition but with similar appearance, hyperspectral cameras have recently been introduced in the industry.They provide more information than conventional vision cameras, but it must be transmitted, pre-processed and analysed depending on the final application of the system.Obviously, the computing cost increases as the volume of data to be processed does.The system's computing power must be increased if it is unable to process all the information in real time.Distributed software architectures must therefore be designed to efficiently use the computational power of parallel computing systems, since computational power increases primarily by the number of processing elements increment rather than by increasing the power of each processing element itself.In this paper we present a software architecture design that distributes heterogeneous work and accelerates computationally expensive processes to meet the industrial system time requirements.