Description of the prototype Sample Clauses

Description of the prototype. Quality control is a main concern in manufacturing processes. This UC propose the use of a 3D inspection system to control the full output of a given production line. Each part has to be inspected to ensure no errors are present in the produced parts. In order to provide an inspection system that avoids common problems found at using mechanical handling, e.g. hiding surfaces or mechanical adaptation to each manufactured part, a 3D inspection system based on machine vision while the parts are in free fall is proposed. The 3D inspection system is based on a multi-camera configuration along a sphere-like structure that allows the computing system to observe the part under study from every possible angle without any occlusion. Using this set of images, the inspection system reconstructs the 3D shape of the object as well as its corresponding texture. This information is used to compare the reconstructed object with a reference system and to detect any important deviation from the manufacturing parameters. This real-time inspection system requires the timeliness of the image capture process to be very precise while providing a short enough response time to cope with the production line throughput requirements. Starting with a functional version of the object reconstruction system, the aim of this UC is to take advantage of the multicore platforms and the tools provided within the EMC2 project to obtain a highly parallel and scalable version of inspection system. Therefore, a very relevant features in this UC concerning with the EMC2 project are the latency and scalability properties of the prototype, i.e. the object processing performed by the inspection system. Specifically, a performance improvement of the object reconstruction stage can be achieved by multiplying computing cores and implementing a software parallelization of most consuming functions. In this way, latency is reduced by means of intra-node parallel computation (“Algorithm parallelization”), whereas throughput is raised by means on inter-node distributed computation (“System scalability”). Thus, it is important to remark the difference between the two types of parallelism that can be applied to the prototype: • Internal (intra-node): Parallel processing of each image. • External (internode): Processing images in parallel. Source code rewriting for algorithm intra-node parallelization has been carried out for the first prototype application developed in this UC. Then, a comparison between a sequ...