Key technologies of smart factory machine vision based on efficient deep network model

Publicado en 3C Empresa – Volume 12, Issue 1 (Ed. 51)

Autores

Fangfang Zhang
Kunfan Wang

Resumen

Abstract

Most of the existing smart space machine vision technologies are oriented to specific applications, which are not conducive to knowledge sharing and reuse. Most smart devices require people to participate in control and cannot actively provide services for people. In response to the above problems, this research proposes a smart factory based on a deep network model, which is capable of data mining and analysis based on a huge database, enabling the factory to have self-learning capabilities. On this basis, tasks such as optimization of energy consumption and automatic judgment of production decisions are completed. Based on the deep network model, the accuracy of the model for image analysis is improved. Increasing the number of hidden layers will cause errors in the neural network and increase the amount of calculation. The appropriate number of neurons can be selected according to the characteristics of the model. When the IoU threshold is taken as 0.75, its performance is improved by 1.23% year-on-year. The residual structure composed of asymmetric multiple convolution kernels not only increases the number of feature extraction layers, but also allows the asymmetric image details to be better preserved. The recognition accuracy of the trained deep network model reaches 99.1%, which is much higher than other detection models, and its average recognition time is 0.175s. In the research of machine vision technology, the smart factory based on the deep network model not only maintains a high recognition accuracy rate, but also meets the real- time requirements of the system.

Artículo

Palabras clave

Keywords

Smart factory; Deep network model; Machine vision; Image processing; Model optimization