Abstract
In the following chapters of this Thesis, an attempt was made to explain some of the face recognition algorithms and their features and implement such algorithms in Python shell using the OpenCV open source library. The goal of the application is maximum efficiency with the lowest possible cost. As shown below, face recognition using Mechanical Vision is based on two different processes. The first process focuses on whether there is a person in the real-time file in question (in this case live video stream using a web camera). On top of this, the second process focuses on training the model with specific data obtained from a pre-processed dataset and comparing these data with the live stream, in order to judge if the data fits some of the criteria we set, i.e. if a known person appears. In addition, for the implementation of the Dead Man Alarm application, which is used according to the Class regulations in commercial vessels’ machinery space, the algorithm examines and decides whether is there a reason to notify the other crew members in case of, during the inspection of vessel's machinery after normal hours on UMS mode, a crew member becomes alarmingly motionless. Thus, after the brief explanation of the individual parts of mentioned algorithm, there is a detailed presentation of algorithm. In addition, short mention is given to the used hardware, which was chosen to be simple and cost-effective. The results of the algorithm show that it performs well under the adverse environmental conditions prevailing in the vessel's engine room (high temperatures, vibrations of unstable frequency and duration, dangerous gases, poor lighting) by recognizing and identifying a person. Finally, we examine whether sudden immobility of a crew member, i.e. not acknowledgement of the alarm after preset time, poses a reason of concern.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call