Objective: Design an image processing system that can detect, target, identify, and discriminate different aircraft using AI.
Background: Currently, modern air-to-air missile guidance systems for all major superpowers, infrared and radar specifically, have methods upon which they can be defeated, such as flares, chaff, radar jamming etc. Image processing can bypass all known countermeasures by tracking on the simple principle that the system can "understand" what it is looking at stay on target.
Methodology: Using machine learning and AI, we would theoretically be able to feed a system a pool of data of aircraft. Of course, in order to create an intelligent application, that we would like to identify aircrafts with accuracy, we would need to feed thousands of sources that the application would need to test and succeed in.
Expected results: We understand designing ordnance has many underlying issues in feasibility and legality, so we only really expect to create the identification system. It would be an iterative development, meaning we would first train the AI to discriminate between aircraft and other potential objects, then if the object is moving, then country of origin, and etc.
Costs: $10,000
Objective: Schools with thousands of students can have a faster way to verify students while in different parts of a building or entering and leaving a building. This can take away the need to take out ID cards, you can walk to a screen and be checked in.
Background: Stevens students often must grovel through their pockets and bags in order to take out their IDs and sign into the library or buildings that they should have easy access to. Therefore, to streamline the process and make it simple for everyone involved, students should be able to have a method in which they can simply walk in.
Methodology: Implementing a facial recognition system using the school's database of Stevens students' faces.
Expected results: To create a system that can successfully scan and verify students faces and let them into buildings.
Costs: $10,000
Objective: As the technology for fingerprint scanning has become more affordable, they have become ubiquitous. As a result, it is possible to develop a rudimentary biometric-lock system using a low-cost scanner and a raspberry pi. This can be developed to unlock a computer, send a physical signal or message.
Background: Common locking mechanisms include 4-digit passcodes that can be easily guessed by malicious software or are physical mechanisms that can be bypassed, such as physical key-pin locks. Lightweight fingerprint locks can overcome the problems of ordinary locking mechanisms by offering security that is difficult to physically bypass or bypass by way of guessing/software. Applications include secure data storage units, biometric safeties on firearms, and door locking mechanisms.
Methodology: Implementing a biometric fingerprinting system using a Raspberry Pi and a low-cost fingerprint scanner. this will be accomplished by registering and storing biometric data, testing whether the system successfully identifies a fingerprint, and whether it can perform an action upon success.
Expected results: to have a scanner correctly identify one user from another, along with sending a signal result that can be used for another action.
Costs: $15 + shipping
Objective: Create a steganographic algorithm that automatically applies either predetermined data (such as a prewritten message) or given automatic metadata (such as location, camera type, time, and date).
Background: Images on their own contain much information, but it is possible for them to be missing certain key points of information as well. Image steganography can be used to include information that would otherwise not be included in an image, such as GPS Coordinates, time, camera model etc. Image steganography methods could be used to include covert messages as well, or even scannable QR codes.
Methodology: Learn the transforms necessary to apply encode basic information/patterns to images (likes using OpenCV). Then build on these transforms to apply more complex patterns/information.
Expected results: To have images that contain embedded information that can be extracted using image processing algorithms.
Costs: No cost for software libraries.