The software requirements for the proposed system are as follows, –

1.

Fig. 1

The Raspberry Pi, a mobile device, and the Tensorflow framework with a pre-trained CNN model. TensorFlow's Deep Learning Implementation Using the Enhanced Regional Convolutional Neural Network. the Tensorflow library and the Raspberry Pi-based, pre-trained CNN model

Fig. 2

Mask Detection Using Framework Tensorflow and Pre-Trained CNN Model Based on Raspberry Pi - Acep Ansor, at all TensorFlow is a machine learning and deep neural network research library created by the Google Brain Team. According to (Liping Yuan et al.,2017), In order to identify masks, a convolutional neural network built on top of TensorFlow has been developed. A library of open-source software used for numerical computing that makes use of data flow graphs is known as TensorFlow. It is a method that sends complicated data structures to artificial neural networks so that these networks may do analysis and processing on the data. It has several applications in the field of deep learning, including voice recognition and picture recognition, among others.

Fig. 3

Teachable Machine uses TensorFlow.js, a library for machine learning in JavaScript, to train and run the models you make in the web browser. It is possible for us to teach a computer to identify human photos, sounds, and postures without requiring you to write any machine-learning code. After that, use this template to your own projects, websites, applications, and other endeavours. Transfer learning is the methodology that is used by these models. There is already a neural network that has been trained, and when you construct your own classes, you can imagine that those classes are becoming the very last layer or step of the neural network. This is because there is already a neural network that has been trained. To be more specific, the image and posture models are both learning from previously trained mobile net models, while the sound model is constructed on top of Speech Command Recognizer.

Fig. 4

Operating System

According to (Agus Kurniawan et al.,2018), Raspbian is an operating system-based Debian optimized for the mask detection software and hardware designed for the Raspberry Pi. When the Raspberry Pi board is finally made available to the public, the Raspbian operating system will take its place as the official Raspberry Pi board operating system. Mike Thompson and Peter Green were the ones who first came up with the idea for the open-source operating system known as Raspbian OS. Even though the first versions of the board did not come pre-installed with this operating system (OS), the Raspberry Pi Foundation swiftly developed it, and as a result, every board manufactured after June 2012 is compatible with it.

Hardware platform

The hardware requirements for the proposed system are as follows, – Microprocessor, preferably Raspberry Pi 4 that performs the majority of the computing. – Cameras for image capture.

1) Normal 5 MP CMOS Camera 2) Thermal Camera – Stable power source to power the system at all times of use. – Screen to monitor the inflow of individuals into a particular institution – Secure cloud for storing as well as analysing user data for further enhancements.

Fig. 5

The Raspberry PI is nothing more than a series of tiny computers packaged on a compact board. The circuit board needs a supply voltage of 5 V at all times (with a 1.2V core volt given directly from the input using the BCM2835's internal switching power supply). Raspberry Pi hardware has evolved over time, resulting in variations in CPU type, memory capacity, 3.5mm audio output, HDMI, and 8GB LPDDR4-3200 SDRAM. Additionally, there are now two USB ports, an Ethernet connector, and an HDMI interface.

CPU Main Components

Fig. 6

Instruction Cycle Debugging as Well as Interlocks Floating-point vector representation (VFP) Management of the System Interrupt Management for the Integer Core Loading Store Unit (LSU) and the AMBA Prefetch Unit Memory System Prefetch Unit Memory Debug for the AXI Interface and the Coprocessor Interface Cycles of Instruction and Safety Interlocks Floating-point vector representation (VFP) Management of the System Floating-point vector.

GPU Overview

Fig. 7

One Level Memory System

Fig. 8

MicroTLB is responsible for determining if cache lines are write-back or write-through. Cache is an implementation of Harvard. Pseudo-Random or Round Robin are the two options for cache management strategies, and the RR bit in CP15 register c1 determines which one is used. The MicroTLB analyses each cache line to determine if it is write-back or write-through. Data that is safe and data that is not secure are stored in cache lines.

Scalability

Now that your Raspberry Pi Cluster is self-organizing and sustainable, let's review some crucial phrases. A multiprocessor employs two or more CPUs. Raspberry Pi's Even though it has four processing cores, a quad-core computer is not a multiprocessor.since it only seems to have a single CPU. multiprocessor is a assortment, such as our Raspberry Pi assortment. that functions as an integrated system. Distributed memory systems have local memory banks or a portion of node cores. Message transfers connect processes to a distributed memory system.The relevant code and (private) memory allocation are copies in each process. Messaging may be utilised on a multicore machine or several computers. Below are three nodes (such as this assortment of Raspberry Pi). Every node includes a memory bank and it is interconnected (for us, switching). Our collection is a DMS.

Fig. 9

The Message Passing Interface (MPI) is an industry-standard library that facilitates the passing of messages in a distributed memory model across different work. The most widely used editing paradigm in distributed memory systems is MPI. In the context of a collection, the work done in the header area is responsible for dividing the work and assigning it to a group of employee notes (or employees), who are then tasked with performing a specified series of calculations and returning the results to the header area. The work done in the head area is also in charge of returning findings to the head area.By operating and is mannered, per nod is able to independently and concurrently carry out its own set of computations. MPI is a standard interface that enables programmers to quickly construct programmes that enable different computer systems to interact with one another in this fashion.

Organization

Fig. 10

The following is an analysis of how the whole work works, how the mask detection model training works, how the mask detection test works with image file input, and how the mask detection test works with video input (real-time) The following subsystems make up our solution: 1) Arduino Uno-based temperature measurement subsystem 2) A Raspberry Pi-based computer vision subsystem for mask identification and social distance assessment No. 4 server side 4) A smartphone app for security personnel. All those attempting to enter the building must first pass a contactless temperature check. We use an Arduino Uno with an infrared thermometer (like the MLX906148) or a thermal camera sensor for that (AMG88339 for example). Additionally, it makes use of an ESP8266 WiFi module to connect to Edge servers using the MQTT protocol. If that person's body temperature is greater than usual, the door is locked and a MQTT message with the temperature value and the location where it was captured is delivered to a server. This message is received by the server, processed, and sent to the smartphone.

Fig. 11

programme used by security guards to track potential intruders so they can show up and make sure they don't attempt to enter the premises again. Otherwise, Arduino will send a signal to open the door if the passenger's temperature is normal. Passengers then go on to the next stage of inspection, mask detection. The camera module version 110 revision 3 of the Raspberry Pi single-board computer was used for this task's computer vision subsystem. Security personnel will be alerted through MQTT message if a passenger is not wearing a mask or if it is not cover their nose, in which case they may either offer a mask or issue a warning to depart. If the individual being checked does not wear a mask, the door will be unlocked. Additionally, while inside the structure, Raspberry Pi devices check to see whether social distance is being used correctly or not at certain spots. Similar to this, if social distance is not being used appropriately in any of the rooms, a MQTT message will be sent to the security officers to let them know. On the server side, message processing, event recording, reasoning, and message forwarding are carried out together with the deployment of the MQTT broker and semantic triple store. When communications arrive to edge servers, To determine which security guard should be contacted, they employ semantic annotation and reasoning. Security personnel utilise a simple Android mobile application that visualises data about rule violations and their locations inside buildings after receiving MQTT messages from the server side. Figure 1 provides a summary of the suggested Internet of Things-based solution that attempts to guarantee that COVID-19 safety requirements are correctly implemented inside.

Abbreviations

The following abbreviation are used in this manuscript:

CPU Central Processing Unit

Dsp Digital Signal Processor

RPI Raspberry Pi

FTP