عنوان

Computer Vision Based Deep Learning Models for Cyber Physical Systems

Number

TLpq2448475401

.Language of Text, Soundtrack etc

انگلیسی

Title Proper

Computer Vision Based Deep Learning Models for Cyber Physical Systems

General Material Designation

[Thesis]

First Statement of Responsibility

Karim, Muhammad Monjurul

Subsequent Statement of Responsibility

Qin, Ruwen

Name of Publisher, Distributor, etc.

Missouri University of Science and Technology

Date of Publication, Distribution, etc.

2020

Specific Material Designation and Extent of Item

64

Dissertation or thesis details and type of degree

M.S.

Body granting the degree

Missouri University of Science and Technology

Text preceding or following the note

2020

Text of Note

Cyber-Physical Systems (CPSs) are complex systems that integrate physical systems with their counterpart cyber components to form a close loop solution. Due to the ability of deep learning in providing sensor data-based models for analyzing physical systems, it has received increased interest in the CPS community in recent years. However, developing vision data-based deep learning models for CPSs remains critical since the models heavily rely on intensive, tedious efforts of humans to annotate training data. Besides, most of the models have a high tradeoff between quality and computational cost. This research studies deep learning algorithms to achieve affordable and upgradable network architecture which will provide better performance. Two important applications of CPS are studied in this work. In the first study, a Mask Region-based Convolutional Neural Network (Mask R-CNN) was adopted to segment regions of interest from surveillance videos of manufacturing plants. Then, the Mask R-CNN model was modified to have consistent detection results from videos using temporal coherence information of detected objects. This method was extended to the second study, a task of bridge inspection to detect and segment critical structural components. A cellular automata-based pattern recognition algorithm was integrated with the Mask R-CNN model to find the crack propagation rate in the structural components. Decision-makers can make a maintenance decision based on the rate. A discrete event simulation model was also developed to validate the proposed methodology. The work of this research demonstrates approaches to developing and implementing vision data-based deep neural networks to make the CPS more affordable, scalable, and efficient.

Artificial intelligence

Computer science

Systems science

Karim, Muhammad Monjurul

Qin, Ruwen

Electronic name

p

[Thesis]

276903

a

Y