IMPLEMENTATION OF GP-GPU PARALLEL PROCESSING TECHNOLOGIES IN ARTIFICIAL VISION PROCESSES FOR THE SELECTION AND QUALITY CONTROL OF PRODUCTS
- Reference: AEST/2020/030
- Principal investigator: Vicente Morell
- Research Centers: University of Alicante, Multiscan Technologies S.L.
- Duration: January 1st, 2020 – December 31th, 2021
- Financial Entity: Conselleria De Innovación, Universidades, Ciencia Y Sociedad Digital. Generalitat Valenciana
- Funding: 40.000 €
The objectives of the project are the following:
- Understanding of the operation and limitations of all production processes of the company, by the researcher. This will make it possible to identify in which processes it is advisable to introduce these techniques, since they imply an additional cost of resources that may not be profitable for the company.
- Implementation of those processes detected in the first objective as the most appropriate to increase the performance of machines that use GP-GPU technologies. This objective is both the participation in the implementation and the possible training of employees by the company.
- Deployment of the system in the production process of the company. Once the highly parallelized processes with the greatest impact have been implemented, we will proceed to the analysis of the industrial GPGPU devices that would be integrated into the machines, carrying out a cost and performance analysis of such devices. As part of the integration into the final product, it will be monitored if the performance obtained has a significant impact on the production of the machine.
- The last proposed objective is more generic and is the identification by the researcher of those processes where a future collaboration between the researcher and the company is possible.
The objective of this project is to transfer the technological knowledge of Human Robotics to the international company Multiscan Technologies S.L.. This company develops industrial machines to classify and quality control of food products. The software that controls the operation of the machines (called MVS) has been developed and improved over the years but the computational needs of today's machines have created the need for multiple computers on the machines, with a due increase in the complexity and price. The level of parallelism implemented so far in MVS tries to exploit parallelization at the level of the tasks that are necessary in the flow of image analysis. Many of these tasks or phases are highly parallelizable tasks due to the nature of the data to be processed. These machines work with many 2D images with different planes and with increasing resolutions, making the use of GP-GPU (General Purpose - Graphic Process Unit) technologies appear a priori to be an easily integrable resource and due to its computational capacity relationship /cost its use can be decisive in an increasingly competitive market.
Human Robotics Group - University of Alicante
Design and Control
of Robots
Advanced mechanical design of robotic devices and kinematic - dynamic control performance
Contact us
-
Human Robotics
UA Polytechnic School 3
Physics, Systems Engineering and Signal Theory Department
University of Alicante
Ctra San Vicente del Raspeig s/n
San vicente del Raspeig
03690 Alicante, Spain. -
(+34) 965 903 400 Ext. 1094
-
huro@ua.es
-
Week Days : 09:00 – 18:00
Saturday, Sunday : Holiday