Milestone 2
Milestone 2: Project Plan, Concepts, Design Analysis, and Test Plan
2.1: Project Plan
Roles and Responsibilities
Project Manager: Chris Alessandri
Network Developers: Marc DiGeronimo, Jason Li
Hardware Developers: Elliott Braverman, Michael Banks, Xavion McClary-Fagan
Software
An image processing algorithm will detect what type of mail a user receives. This is run on an AWS cloud network that includes SageMaker Machine Learning stored by S3 Buckets. When the service is run, AWS SNS will email and text the user directly.
Hardware
A sensor detects if the door is open. When it does, the camera will start recording and take pictures of the mail. The image data is then sent to a Raspberry Pi Pico W that will then send the image to the AWS cloud network.
Project Management
Team communication is done via Discord to discuss ideas, proposals, and implementations of any design concepts. A Google Drive folder contains all files and are accessible to the team as well as to the team's Senior Design advisor. The team meets twice a week to discuss their progress and meets with their Senior Design Advisory every other week.
Work Breakdown Structure
Pre-Alpha (September - December 2023)
Milestone 1 (September 1st - October 11th)
Customers, Problems, Needs, and Requirements
Which Internet of Things Device should the team work with?
Upskill in Cloud Computing, Image Processing, and Electrical Design
Milestone 2 (October 12th - November 29th)
Project Plan
Roles + Responsibilities, System Details, Work Breakdown Structure
Concepts - Research Different Technologies (Hardware + Software)
Determine Concepts via Design Analysis (i.e. KT Decision Matrix) with Mathematical Justifications
Product Formulation
Design
Analysis
Test Plan
Team Culture (November 30th - December 13th)
Team Responsibilities
Team Assessments
Presentation
Alpha Stage (December 2023 - April 2024)
Design Verification/Planning (December 14th - January 31st)
Milestone 3 - Alpha Development (February 1st - February 28th)
Alpha Prototyping
Design Performance
Cost Review
Verification Stage (March 1st - March 21st)
Prototype Testing
Modification
Beta Stage (March 22nd - April 24th)
Implement Changes
Poster Development (January 16th - April 10th)
Poster Abstract
Poster Mid-Project Submission (March 20th)
Poster Final Submission (April 10th)
Presentation Development (March 1st - April 26th)
Create Slide Deck
Rehearsal and Revisions
Final Touches
Innovation Expo (April 26th, 2024)
2.2 Concepts
Acceptance Criteria
Basic Functions - The device is fully capable of sending notifications when it has determined mail has been delivered
Battery Life - The device supports a minimum of one month of battery life.
Image Processing - The device is capable of distinguishing between basic letters and small packages, notifying the user which one has been inserted into the mailbox.
No False Positives - The device is equipped with sensors and a camera that only triggers when mailbox activity is detected.
Success Criteria
Our device shall successfully address our target market. It shall provide an additional layer of convenience that people find to be desirable. It shall have the capability to be integrated into peoples’ day to day lives with relatively low risk of failure.
Out of Scope
Audio capabilities such as microphone or speaker usage.
Safeguard capabilities such as locking a mailbox to prevent theft.
Compatibility with other smart home and IoT devices.
Assumptions
We assume our target market can afford this product.
We assume that connectivity and efficiency will not be hindered by the PO box itself, including material, thickness, and other environmental obstacles.
Dependencies
We will not be designing nor manufacturing the components of our project. We depend on third party companies to deliver high-quality and quality-assured components.
We depend on the average size of our target market’s PO boxes. The smaller these are, the smaller our product must be to not hinder the delivery of mail.
Constraints
Our project must have an MVP ready by April 26th, 2024.
Our product can receive reimbursement of up to $400 in costs. This acts as a soft budget that can only be increased if a sponsor is found.
Minimum Viable Product (MVP)
Our minimum viable product consists of sensors attached to Raspberry Pi Pico W and its respective camera module. These shall detect when mail arrives, sending an image of the mailbox via email / SMS. We shall iterate to offer wireless communications via the AWS Cloud environment.
The image processing aspects shall be secondary to network development. Once the network is set up, our team shall focus its attention on image processing algorithms and components. Should we complete this with time to spare, we shall iterate to include additional functionality such as character detection.
2.3: Concept Selection (via KT Analysis)
Preliminary Notes:
Our tables use weight and rating scales of [1-5] rather than [1-10]. This is to better justify our ratings, as the difference between 2 & 3 in a [1-5] scale is more significant than in a [1-10] scale.
Our tables use the same needs and wants. Needs are more general, including hardware, software, and overall compatibility between components. Wants are based on our customer wants & needs from Milestone 1 and are listed below:
Accuracy (5) - Accuracy of detecting mail & determining what mail has arrived.
Accuracy is most important as this is the main goal of our system. We want to make sure that the mail notification gets through to the customer. Otherwise, our whole system will be pointless!
Security (4) - Confidentiality, integrity, and availability (3 tenants of cybersecurity).
Will our system be vulnerable to cybercriminals? This is important as we do not want any data gathered by the device to leak out to the public, such as the personal information of our customers.
Additionally, can this device be used or given a purpose that aids overall system security?
Compactness (4) - Size and scale of our device, as well as how unobtrusive it is.
Compactness is very important as the team wants to integrate the device into mailboxes and PO boxes that have already defined size restraints. Our device should not greatly limit a mailbox’s already limited space.
Energy Efficiency (3) - How much energy our device uses. AKA how often the device must be charged or otherwise maintained.
This is important as the final version of our product shall be implemented in scenarios where power isn’t readily available. For our alpha and beta prototypes however, we intend to use large batteries or connect our prototype to a power source to make sure that our other components work.
Accessibility (3) - Accessibility to all people, regardless of PO box type / location.
Accessibility (or ease-of-use) is very important as it is a factor that will influence customer satisfaction. A very complicated and/or hard to use system will result in customers being irate and not want to use our product. Additionally, overcomplicated systems may not work for certain PO boxes.
Since we have shifted from mailboxes & PO boxes to only PO boxes, this aspect has diminished in its importance. PO boxes overall have fewer variations than mailboxes.
Cost (2) - Cost of constructing our product *(lower value = lower cost)
Our product is limited by various constraints, one of which being our $400 budget. Balancing between component costs is integral, and high costs lead to pricier products. However, with a novel idea and relatively cheap components in mind, we prefer to focus on the project before its price for now.
Justifications of our ratings can be found below each table. We will not be justifying any GO in the Needs section, although we will justify any NO GO.
Justifications:
Accuracy: Accuracy was defined by the clarity and resolution of the resulting images. The ToLuLu and Raspberry Pi Camera modules require light for high resolution, so these would be rated low. The ToLuLu Camera with a light would result in good accuracy if used with a light/flash. The Blink Outdoor Camera is designed for outdoor and security use, so it would be the best in dark/outdoor environments.
Security: The ToLuLu and Rapsberry Pi camera modules are not weather resistant, and are prone to damage if not protected by an enclosure. The Blink has high security due to having a weather resistant enclosure.
Compactness: The ToLuLu camera module is built as an external product and would be difficult to include in a single product, which gives it poor compactness, however the Raspberry Pi camera module is built to be small and compact, which is beneficial for this product. The Blink Outdoor camera is compact, as it already includes an enclosure and technology to stream the camera feed out to an external source. The Raspberry Pi camera would be the most compact.
Energy Efficiency: The ToLuLu and Raspberry Pi cameras use relatively little energy. While the Blink Outdoor camera uses more energy than the other options, it is designed to be used outdoors and has energy conservation measures to last a long time. It also uses replaceable batteries. The Raspberry Pi would be the best choice energy-wise.
Accessibility: Accessibility was defined as how compatible the camera would be with our system. The ToLuLu and Raspberry Pi cameras would be easy to integrate with a Raspberry Pi, however the Blink Outdoor camera would be very difficult, as it is a separate product and is designed around its own system/interface. The Raspberry Pi camera module would be the best choice for accessibility.
Cost: The Blink Outdoor camera is extremely expensive ($100) and would not be cost effective. The ToLuLu cameras are also expensive (~$20) and wouldn’t be cost effective either. The Raspberry Pi camera is cheap (~$5) and would be the best choice cost-wise.
Justifications:
Accuracy: When it comes to accuracy, Light sensors were given the lowest level among all three of the sensors. The reason for this is simple, depending on the mailbox, and the environment the mailbox is placed in, the light levels can experience a wide level of variation that cannot be standardized for all customers. As an example, one user may have a PO box that has little to no openings, so there would be little to no variation in light level without the mailbox being opened. This is the ideal case scenario. However, there may also be other customers that have mail/P.O. boxes that do not block out light as perfectly. This kind of mailbox may experience variations in lighting depending on the external conditions outside the mailbox. This creates potential for false positives, which would be detrimental to the overall accuracy of the product. Pressure sensors were given a higher score than light sensors, but a lower score than motion sensors. This is simply due to the fact that, depending on the pressure sensor, smaller and lighter forms of mail may not be detected due to failing to meet a minimum weight threshold for a positive reading. While this is unlikely, there is still potential for false negatives, which is why pressure sensors were given a slightly lower score than motion sensors. On the other hand, the motion sensors were given the highest score of 5. The reason for this is that motion sensors are far more suited for the task of detecting mailbox activity. Unlike a light sensor that could detect changes in light even when the mailbox hasn’t been opened, a properly functioning motion sensor will not detect any movement unless the mailbox is being accessed. This leads to far more accuracy overall, which is why the motion sensors were given the maximum score of 5.
Security: All of the sensors were given a minimum security score of 0 due to the fact that sensors do not contribute to the overall security of the device.
Compactness: As for compactness, Pressure sensors were given the highest score due to the fact that pressure sensors tend to be much flatter and more disc-like compared to motion and light sensors. Not only that, but pressure sensors do not need to be exposed since it is not monitoring an external condition like light or motion. Instead the pressure sensor would monitor any weight placed on the device itself. Being able to leave a sensor unexposed/within the main device has a positive impact on compactness. Motion sensors and light sensors were given a similar score since they can be close in size, and both require some external exposure, which may lead to them having a separate module that may not be placed as deep within the device.
Energy Efficiency: Based on reviewing multiple retailer sites (such as https://us.faradite.com/products/motion-sensor-360-volt-free) for each sensor, pressure sensors on average seemed to consume the least power. Motion sensors consumed the most, and light sensors were between the two.
Accessibility: When it comes to accessibility, the primary factor that was analyzed was which sensor would allow the device to work with the most functionality across the widest range of circumstances. Upon evaluation, the motion sensor seemed to best meet this criteria. Motion sensors not only offer high accuracy, but they also perform to a similar standard regardless of other conditions such as size and build quality of the mailbox, orientation of the device, conditions outside the mailbox, etc. Light sensors were given the highest score after motion sensors. Due to the fact that light sensors will perform to a similar standard regardless of the orientation of the device and the build quality of the mailbox. However, since external light conditions may lead to false positives, light sensors were given a lower score than motion sensors. This is due to the fact that people who have their mailboxes in a place with large amounts of light variation may not be able to make full use of this product should a light sensor be integrated. The sensor that received the lowest score was the pressure sensor. This is because using a pressure sensor would place numerous constraints on the usage of the device. For example, if a pressure sensor were to be used the device would be unable to be mounted anywhere besides the bottom of the mailbox. This is due to the fact that in order to detect weight, mail must be placed directly on top of the pressure sensor. Using a pressure sensor would also place limitations upon the size of the device. If the device is not thin enough and is mounted on the bottom of the mailbox, it would lower the overall capacity of the mailbox. Also, if the device does not have enough surface area, it may not be able to detect the weight of mail placed everywhere inside the box. If a prospective mail courier places a letter to the left of the device instead of on top of it, no notification would be sent as the device would not be able to detect the weight.
Cost: Based on reviewing multiple retailer sites(such as https://us.faradite.com/products/motion-sensor-360-volt-free) for each sensor, pressure sensors on average seemed to cost the most money. Motion sensors were priced the second highest , and light sensors were the lowest costing overall.
Justifications:
Accuracy: The PIR sensors are efficient for detecting movement from infrared light. All these products use the same PIR technology and would likely perform the same. While they do detect movement, they cannot determine where, or what moved. However, this is not needed as they will be implemented along with a camera.
Security: These products are designed to be incorporated into a larger product so they don’t have any weather resistance. They would require a weather-proof enclosure to improve security.
Compactness: All these products are small and compact, as this technology doesn’t require a lot of room.
Energy Efficiency: PIR are the most efficient energy-wise as they do not output energy for detection purposes.
Accessibility: These sensors are easily accessible with a Raspberry Pi and are designed to be compatible with our system.
Cost: PIR sensors are very cheap and can be found for ~$1. They can also be bought in bulk to reduce prices further.
Justifications:
Accuracy: No Mounting was given a 2, as it can easily be pushed around by incoming mail. Adhesive and Magnet were given slightly higher ratings, as both are at risk of being knocked loose (magnet less so than adhesive). Permanent was given a 5, as there is almost no risk of being knocked loose.
Security: No Mounting was given a 1, as the device could easily be removed from the PO box if broken into. Adhesive and Magnet were given a 3, as they could also be removed, albeit with more difficulty. Permanent was given a 5, as it could not be feasibly removed without significant difficulty.
Compactness & Energy Efficiency: These Wants were not considered, as the method of mounting does not have a direct effect on them. The only one that may apply would be No Mounting, which may get in the way due to its position. However, this was included in the Accuracy section.
Accessibility: Both Adhesive and No Mounting were given 5s, as both methods are extremely simple and are not limited by location. Magnet was given a 4, since adding magnetic strips requires some assembly compared to the previous two. Permanent was given a 1, which is described in the NO GO section below. Additionally, it would require special tools and installation, which would be especially difficult in a cramped PO box.
Cost: These ratings were given according to preliminary research into each method’s cost. Magnet was found to be the most expensive due to the need for magnetic strips. Permanent was found to be slightly less expensive due to the cost of screws. This assumes the customer already owns a drill, otherwise this would be rated as a 1. Adhesive prices (sticky tack, tape, and similar) are lower than the above, giving it a 4, while No Mounting is entirely free, giving it a 5.
NO GO:
No Mounting was given a NO GO in compatibility. This mounting method would be incompatible with the Mounting Orientation section, as it could only be oriented on the bottom of the PO Box. This in turn would make it incompatible with some sensors, especially since it would be prone to getting buried by mail.
Permanent was given a NO GO in hardware. While this mounting method is theoretically the best, it is not feasible in a PO box. PO Boxes are owned not by the customer, but by the building they are settled in. Drilling into private property is not something our customers will be able to do.
Justifications:
Accuracy: Back was given a 1, as mail and packages would tend to bury or block it. This would conflict the product’s entire purpose. Side and Door were given 3s. Side orientation has a similar risk of being blocked, albeit to a lesser extent. Door, while less prone to being blocked, has issues that are listed in the NO GO section below. Top was given a 4, as it is at the least risk of being blocked (although there is still some risk).
Security & Accessibility: These two factors are inversely related, as more accessibility means the device is easier to be tampered with. Door is the most easily accessible, and as such its security is the lowest. It wasn’t given a 1, as the device’s presence could act as a deterrent to criminals. Side and Top were given 3s across the board, as they are not particularly secure nor accessible. Back was given the highest score, as it is quite difficult to access.
Compactness: This Want was considered more so as “unobtrusiveness” for this component. Back and Door both run the risk of obstructing or damaging mail, as it effectively limits the depth of the PO box. Top is somewhat less obtrusive, limiting the height of the box. This can obstruct certain packages though. Side is the least obtrusive, limiting the width of the box. Of the three dimensions, this affects mail the least.
Energy Efficiency & Cost: These Wants were not considered, as the mounting orientation does not have a direct effect on them.
NO GO:
Door was given a NO GO in compatibility. This orientation would cause some sensors to activate upon opening the PO box, rather than when mail is placed inside. This is more akin to Ring, which captures the person accessing the mailbox, rather than our product’s purpose, which is to capture the incoming mail and identify it.
Justifications:
Accuracy: Powering method was given the minimum score for accuracy as it has no bearing on the overall accuracy of the device
Security: Powering method was given the minimum score for security as it has no bearing on the overall accuracy of the device
Compactness: Swappable batteries were given a two for compactness. This is due to the fact that in order to have a swappable battery, there must be some access point for the user to be able to directly access this battery. Unlike custom sized rechargeable batteries that are built into the device and not made directly accessible to the user, swappable batteries have to follow an industry standard. Typically, multiple double A sized batteries would be used for a device of this nature. With the lack of custom battery sizing and necessity of user accessibility, swappable batteries are relatively detrimental when it comes to compactness. Rechargeable batteries received a higher score due to the aforementioned custom sizing of the battery. When rechargeable batteries are integrated into a device they are often optimized for a balance between high capacity and compact form factor. However, rechargeable batteries still occupy space within the device and it must be allocated as such. A rechargeable device must also have a built-in charging port. Due to this, the wired option is the most compact. While there is an external wire, the device itself would not need to have any space within provisioned for batteries
Energy Efficiency: Energy efficiency was given the minimum score as the source of power has little to no bearing on how much energy a device consumes
Accessibility: Swappable batteries were given the highest score in terms of accessibility due to the fact that swappable batteries are easily purchasable, have long shelf life, and can be replaced as soon as they are drained. Swappable batteries lead to lower amounts of downtime compared to rechargeable batteries since there is no charging time associated with swappable batteries. As soon as the batteries fail they can be replaced and the device will resume operation. As for rechargeable batteries, you must remove the device from the mailbox, recharge it, and during that period of charging the device is no longer available for use. Wired Power Delivery was the least accessible as most people do not have the capability to wire a mailbox directly to a power source from their house. In the case of P.O. boxes, wiring a device within a P.O. box may not be allowed or possible especially for apartment owners.
Cost: Devices with rechargeable batteries were given a 1 for manufacturing cost. This low score is due to the fact that devices with rechargeable batteries require custom sized batteries, charging circuits, and additional safety features not present in the alternatives to ensure safety in cases such as overcharging. Devices with swappable batteries would cost less than rechargeable batteries due to the lack of charging circuits and additional safety features. However, due to the space provisioning and manufacturing of battery compartments, it would cost more than a wired device (which received the highest score in terms of cost efficiency).
NO GO: N/A
Justifications:
Accuracy: All modules except the Milk-V receive a 5 for accuracy because they all have the hardware, software, and add-ons to achieve their intended purpose. The platforms are very mature and developed well. The Milk-V scores a 4 because of the incomplete and sometimes problematic software that results from a relatively new platform.
Security: Raspberry Pi 4 receives a 3 for security since it can run much more software meaning it can run more possibly malicious software. It has heavier software and more code means more bugs and more bugs can mean more possible security vulnerabilities. Raspberry Pi, Pico, and Arduino Nano receive a 4 because they run embedded operating systems that are more lightweight and support less software meaning less malicious software. Milk-V is good for security because it is risc-v based rather than ARM based meaning attackers are less familiar with the instruction set and are less likely to craft attacks since it is not as popular as ARM based machines as well. However it receives 4 not 5 because the software can be incomplete and buggy, which can lead to security vulnerabilities.
Compactness: Raspberry Pi 4 receives a 3 because it is a small single board computer but due to the I/O capabilities it takes up more space. This can be a problem for smaller boxes but it may not matter for larger ones but it is something that we need to consider for P.O. boxes. The remaining modules all receive a 4 because they are all designed to be compact modules for embedded applications so it is a great fit for these applications.
Energy Efficiency: Raspberry Pi 4 receives rating of 3 because it has the highest power draw of all of the modules at 2.5 A. While it is efficient in the context of desktop systems it is not as efficient for this application because of the extra functionality that it has. The remaining modules all have an energy efficiency of 4 because they all have much less power draw at 25, 22 and 19 mAh respectively, all close enough in this context.
Accessibility: Raspberry Pi 4 receives a rating of 4 because it has the most software support as well as the most powerful hardware. It can run desktop GNU/Linux with the official OS being Debian based with full access to ARM based packages in apt. Due to this it can run programs that the others cannot. There is also a wide variety of support and availability for both first and third party accessories. The Raspberry Pi 4 also has the most amount of I/O. Raspberry Pi Pico receives a rating of 3 because it does not have the same software capabilities of the Raspberry Pi 4 since it is an embedded machine. It does not have access to a full GNU/Linux desktop. However, it does have the same variety of support and availability for accessories/add-ons and software such as the Pi 4. The Milk-V Duo has a rating of 1. While it does have access to desktop GNU/Linux distributions and has built in computer vision accelerators, the software support and stability is poor. Since it is RISC-V based the software is very buggy and results in a lot of needed tinkering to get things to work. Moreover, the GNU/Linux distros need to support RISC-V which many software does not yet support. We do not think we should be dedicating too much time to get software compatibility set up and would rather focus on other things. Arduino Nano gets a rating of 3 because it is based on ATmega328 so it does not have the same software capabilities as Raspberry Pi 4. It is similar to the support of Raspberry Pi Pico. It also has a good support of accessories and add ons.
Cost: Raspberry Pi 4 receives a rating of 2 for cost because it is the most expensive unit at a cost of $60-$70 for the base module without any charging cables, cases, heatsink, Micro SD card and other components. Raspberry Pi Pico receives a rating of 5 because it is the cheapest at $5-$7, this cost does not reflect charging cables and micro sd card and other components. Milk-V Duo and Arduino Nano receive a rating of 4 as they can both be found at around $14-$17 without additional components.
NO GO: N/A
Justifications:
Accuracy: All image processing software receives an accuracy of 5 because each option has the capability to properly achieve the intended effect but optimization and implementation is on the developer, it is not a limitation of the software.
Security: All options have the same security ratings because all of the code is either compiled or interpreted at C/C++ code. This leads all to the same possibility of memory attacks that C and C++ code is vulnerable to.
Compactness: Compactness is not applicable for these options
Energy Efficiency: Energy efficiency is not applicable for these options
Accessibility: OpenCV Python gets a rating of 4 for accessibility because the group has more collective experience with python programming and it is easier to implement when compared to alternatives. OpenCV C++ gets a rating of 3 because the group has less experience with C++ and it can be more difficult to set up the programming environment for C++. The errors are also more difficult to deal with compared to python. OpenCV python and C++ have good documentation. MATLAB has a rating of 3 because of good documentation and some of the group have experience with MATLAB.
Cost: All options but MATLAB have a cost rating of 5 because they are free. MATLAB has a rating of 4 because it is free for students but after that hefty licensing fees are required.
NO GO: C++ is a no go because it is too complicated for the scope of the project and the other options are more straightforward for implementation.
Justifications:
Accuracy: In terms of accuracy contextually, the working definition leans more towards what’s more likely to be received or read. The problem with a web app is that you have to check, even if it sends you notifications, they aren’t as accessible as notifications on the phone. Phone Apps allow you to send and receive phone notifications that are more frequently checked than other types. One key statistic for measuring that is click-through rates. Texts and push notifications are higher than email notifications, 6%, 3.5—4.6%, and 2.9% on emails, but at least they are all actively going to you as opposed to a web app, which you’d have to check on your own.
Security: All of these are fairly susceptible to security breaches, which goes for anything online. Web apps are always online, you’d also be storing data online that is a vector for attack. Phone apps can have their developer account compromised and have forced updates leading to a compromised system. Emails get hacked all the time, but at least SMS has a different vector of attack mainly from cell service providers, which would be one more layer of abstraction from us.
Compactness: This doesn’t apply in the traditional sense, so the working definition would be how intrusive this would be to you. Or how much work you’d need to do to get running. A web app that you need to check is pretty intrusive in that you’d need to create an account and go out of your way to check it. While a phone app you might not need to check due to notifications, you’d still need to download it. Whereas you likely already have emails or texts already.
Energy Efficiency: N/A
Justifications:
Accuracy: Both are rather accurate, reliable, and available with a low amount of many outages
Security: Both can be rather secure, but they still suffer from cybersecurity vulnerabilities.
Compactness: Compactness is not a factor as AWS and Azure are software.
Energy Efficiency: These scores are equivalent as both of the cloud providers use roughly the same amount of energy.
Accessibility: Accessibility is scored higher for AWS compared to Azure due to AWS’s additional customizability.
Cost: Azure and AWS are close in cost, with Azure costing slightly less. Azure also has the added convenience of notifiying clients when too much data is being used. This feature can save more money as compared to AWS, which allows costs to climb with no alerts or notifications.
Justifications:
Accuracy: LEDs are a lot brighter and allow for visual clarity, helping with sensors and visual inspection
Security: N/A
Compactness: From what could be observed when comparing the brightness of each bulb versus their size, LEDs are the superior option. Many of the available infrared lights are large incandescent bulbs that are not compact at all.
Energy Efficiency: LED’s are extremely efficient in terms of wattage, incandescent infrared bulbs operate in a range of 200-300 watts, whereas LEDs can run at approximately 30 watts.
Accessibility: There are LED lights everywhere now, from bulbs to strips, infrared light seems to mainly have bulbs and heat therapy products.
Cost: LED’s are pretty cheap, depending on arrangements, you can buy individuals, you can buy strips, bulbs, to suit your needs, we might be able to get a infrared bulb for 3 times the price of a LED bulb, but when it comes to other smaller configurations, those can cost hundreds of dollars as they’re marketed as heat therapy products.
Improved Morphological Chart
2.4: Design
System Diagram/Architecture
Process Flowchart
2.5: Analysis
Hardware Specifications
The hardware in our project involve these components:
Raspberry Pi Pico W - microcontroller version of Raspberry Pi similar to an Arduino board. The sensors and camera will be able to integrate with the Raspberry Pi Pico W which will be responsible connecting to the AWS Cloud infrastructure that runs our image processing algorithm.
Infrared Sensor: This IR sensor is passive meaning that it will be able to wait in standby until motion is detected via the opening of the mailbox. This would then trigger the camera to start taking photos of the mail put in the PO box.
Raspberry Pi Camera Module - This Raspberry Pi Camera Module connects directly to and has compatibility with all Raspberry Pi products.
Software Specifications
The software is mainly reliant on these following technologies
Python: programming language that will integrate with OpenCV to be able to detect from the PO box what type of mail and from whom.
OpenCV: open source image processing library that gives the team various ways to perform image processing on the pictures taken from the Raspberry Pi camera module.
Amazon Web Services (AWS): cloud environment that will host all of our processing services ranging from our image processing algorithm via OpenCV and Python
SNS (Simple Notification Service) to notify users that their mail has arrived
Lambda (automation tool that runs scripts when events are triggered)
S3 buckets (database storage units)
Sage Maker (Machine Learning tool).
2.6: Test Plan
The team will first experiment with the minimal amount of components necessary to conduct an MVP test. The System Developers will be working on the image processing algorithm using OpenCV and Python running on a local computer attached to the Raspberry Pi Camera Module. The Network Developers will work on the AWS cloud infrastructure and set up the Simple Notification Service and SageMaker. This should be completed or wrapped up by the end of January.
System Developers will test if edge detection on a package or letter can be detected. Network Developers will test if the AWS SageMaker and Simple Notification Service pushes real-time descriptive notifications to the user's device.
The System Developers will then integrate the rudimentary image processing algorithm to the motion sensor and Raspberry Pi Pico W. If there is extra time, more work can be done on making additional features to the image processing.
The Network Developers will work on porting the image processing algorithm developed by the System Developers to the AWS Cloud infrastructure.
Once these separate processes are satisfactory, all the components and phases will be combined together to produce a whole system.