MacProfessionals| Eeny, Meeny, Miney, Sample!

Project Details

Awards & Nominations

MacProfessionals has received the following awards and nominations. Way to go!

Global Nominee

The Challenge | Eeny, Meeny, Miney, Sample!

You are the astronaut/robotic mission lead tasked with bringing valuable specimens from the Moon back to Earth for further study. How will you evaluate lunar samples quickly and effectively before or while still on the mission? How will you differentiate samples of potential scientific value from less interesting material?

SpaceMan

SpaceMan is a system used to find, analyze and collect different type of lunar samples on the surface of the moon without requiring man-power.

SpaceMan

SpaceMan is a group of robots, containing a cluster of sensors, and a capsule which is the nest and power shell of the robots with a trained Artificial Intelligence Image Classifier integrated into it.

SpaceMan is used to find, evaluate and collect lunar samples, if they get approved by the capsule. SpaceMan system reduces the search time for lunar samples by evaluating them according to their physical and chemical structures in real time.

Introduction

When the first people set foot on the moon many years ago, it was an inspiring moment for people around the world. But another kind of explorer is responsible for much of the modern enthusiasm for space exploration. We as youngsters with the contributes of several key people were allowed to conduct our research and think of this idea. The idea is to make a system that finds, evaluates and collects lunar samples. Several situations risk the astronauts’ health and cause inefficiencies while collecting lunar samples but there are also solutions that will be explored in the following paragraphs.

Firstly, the main problem faced by the astronauts in previous researches was that, they were searching for minerals [1] and rocks in a specific small area and they didn’t know if those were valuable. They have brought approximately 2,000 samples of rocks to the earth [1] which were then analyzed and most of them were categorized as useless.

Secondly, the different toxic materials found on the surface of the moon, like silica (SiO2) can only be recognized with scientific experiments and researches so they can cause a lot of damage while they are being carried to the lab [2]. One of the diseases caused by silica is silicosis which causes difficulties in breathing and infections in the lungs which can result in death [2].

Thirdly, the space radiation as we know is many times higher on the moon than it is on earth [3]. As we know the moon does not have an atmosphere which means that there is no other protection for the astronauts except their suit. It may cause cancer and damages in CNS (central nervous system) when exposed to it for a long time [3]

Factors determining the value of a lunar sample

Presence of Minerals
Magnetic Properties
Radioactivity Properties
Presence of Silica
Distinguished colors
Special markings on its surface

We found this by doing research on different articles provided by NASA which are referenced in the end of this page.

Processes determining the value of the sample

As mentioned before, we have a cluster of sensors, and a trained image classifier algorithm. Down below you can see how we combine both of these, to perform fast and real time lunar sample evaluation.

DETERMINING THE PRESENCE OF SILICA

Silica is known as a toxic mineral which is found in a large amount on the surface of moon mixed with the rocks and meteorites, being near this substance can cause serious lung diseases. Silica, as a substance has the property of absorbing infrared light spectrum. IRS or else Infrared sensor is usually used to measure distances but in this situation, we use the sensor for a completely different purpose. As Silica absorbs the IR light it causes an instant drop in the sensor's output data (voltage). This change in data can show us signs of silica in the lunar samples.

Our actual sensor on board our prototype : Click Here To View (Google Drive Image)

Graph showing how IR sensor changes value : Click Here To View (Google Drive Image) (1023 = 5v, on arduino 10 bits analog input read).

DETERMINING RADIOACTIVITY PROPERTIES OF THE SAMPLE[2]

Our robot has an on-board Geiger–Mueller counter, which can detect if a sample contains radioactive properties. But by doing this we had encountered a problem which is that space and the moon's surface contain a lot of ionized particles passing through, which means our sensor can turn false positive results. But we found a solution to minimize this error in data. By combining real live data from the Lunar Reconnaissance Orbiter (LRO) and using a trained pattern recognition algorithm we can see if the results are real. Firstly, by using data from LRO we can know in real-time if there is any solar wind of ionized particles directed to the moon's surface. Secondly, the Geiger tube will always keep on track of radiation no matter if it is analyzing a sample or just moving, then if we apply the data from the sensor consistently, into a trained data pattern recognition algorithm, we can detect unusual radiation activity, therefore, detecting the lunar sample's radioactive properties.

Our actual Geiger–Mueller processor on board our prototype : Click Here To View (Google Drive Image) And it's tube on bottom of the prototype: Click Here To View (Google Drive Image)

DETERMINING THE MAGNETIC PROPERTIES OF THE SAMPLE

On board the robot, there is an electromagnet connected to a relay, this relay will switch on a high powered battery, for the electromagnet to function. Once the electromagnet is in function, there is another proximity IR sensor which will detect if anything has been attached to the electromagnet. And since there is a low amount of gravity on the moon, even if the magnetic force is low, we can observe the rock being affected by the force, if it contains magnetic properties.

Electromagnet on board the prototype : Click Here To View (Google Drive Image). The proximity sensor, which is used to detect if the anything is attached to the electromagnet : Click Here To View (Google Drive Image).

DETERMINING THE PRESENCE OF FLUORESCENT MINERALS[4]

To find the presence of Fluorescence Minerals we can use an Ultra-Violet light. These minerals have the ability to temporarily absorb a small amount of UV light and instant later release a small amount of light of a different wavelength. This change in wavelength causes a temporary color change, the change can be detected with a camera and image recognition algorithm.

Actual picture of the UV LED's in the sensors platform of our prototype : Click Here To View (Google Drive Image). Another picture of where they are placed : Click Here To View (Google Drive Image).

SPECIAL MARKINGS ON ITS SURFACE

Special markings on rocks can indicate origin, type and other things about the rock, which can increase the value of the rock. To perform this scan on the sample, we use a white dimmed LED, to brighten up the image for the camera. Then the camera takes the picture and sends it back to the capsule, where the image processing takes place, by applying bump map algorithm to create the texture ready for another pattern recognition algorithm to analyze the surface texture. And by this we can categorize the rock according to its special markings.

Picture of a rock, with a bump map algorithm applied, that later can be used for pattern recognition AI : Click Here To View (Google Drive Image). On board camera on our prototype : Click Here To View (Google Drive Image).

DISTINGUISHED COLORS

This feature will obviously make a rock more valuable, to do this we again use the same picture, but this time we apply an algorithm which will take the surrounding areas’ RGB colors, and it will get a repeating color spectrum, so if there are unusual RGB values somewhere in the rock, it will immediately acknowledge that the rock has a special color.

Technology we used ON OUR PROtOTYPE MODEL:

Arduino Uno R3 (Image Processing for OV7670 camera)
Arduino Mega Pro 2560 (Main part, sensor data, communications, etc.)
Camera (OV7670)
Ultra Violet LED's
Infrared Sensor (Sharp 2Y0A21) (For Silica presence detection)
Proximity Sensor (HW-006 v1.2) (For detecting if anything gets attached to the electromagnet)
Bluetooth Module (HC-05) (Main link of communication between Commander Application and the Robot)
Geiger-Mueller Counter (d-v1.1 Cajoe)(For detecting radioactive properties of a sample)
Battery shield with 18650 battery (For managing power supply of the system)
Solar panel cell 6V (Solar battery charging)
Power Relay 16A (For switching the electromagnet on and off)
Coil bobbin (Made with Ferrite) (used for creating a magnetic field to test a sample for magnetic properties).

More images of our prototype :

Image 1 (Side picture of our prototype): Click Here To View (Google Drive Image)

Image 2 (Sensor cluster platform): Click Here To View (Google Drive Image)

Image 3 (Bluetooth Module) : Click Here To View (Google Drive Image)

THE BIGGER PICTURE ON TECHNOLOGICAL ASPECT:

More sophisticated sensors and more precise sensors
4K Camera
Thermal Camera
Magnetic hall - effect sensor
Remote movement control system
Mechanical arms
On board CPU (Not micro-controller)
On board AI (On Board Processing)

Artificial Intelligence Image Classifier Usage

For implementing AI, firstly we took around 200 pictures of special rocks mostly from NASA data portals. We used Googles AutoML Vision Image Classifier API to classify whether the rock has visual properties of a mineral. Firstly we trained it, then we tested it and implemented on the Commander Application.

To get more precise results, we tend to increase the dataset of mineral pictures, and classify each of them by percentage on it's specialty for ex; Distinguished color, Markings etc.

THE COMPLETE PROCESS OF SAMPLE ANALYSIS AND MISSION CONCEPT

1. The 'SpaceMan' robots are deployed from the spaceship to the target area, and they are let out to investigate for samples. They can be remotely controlled by the astronauts on board the spaceships without requiring them to exit their spaceship at all.

2. Once a potential sample is detected by the cameras on board the robot, the robot will stop, send a notification to the astronaut and wait for confirmation to start analyzing the sample. Once the confirmation is granted, the robot will then start the analyzing process.

3. Firstly, the robot will "sit" on top of the sample (as seen on the video). Then it will start the process by turning on white dimmed flash lights and taking picture of the rock, saving it and sending it back to the capsule where the image processing will take place. Secondly, it will bombard the rock with UV lights to test for fluorescent properties (fluorescent minerals can indicate the value of a rock) by taking another picture after the process is done. Thirdly, the robot will turn on the electromagnet and will observe if the sample is reacting to the magnetic field. Later, the Geiger-Mueller counter will consistently be checking for radioactive properties (explained briefly on the evaluation section above). At the end the IR Sensor will check the presence of Silica

4. After the analyzing process, the data taken from all the sensors will be sent back to the capsule. In the capsule, the Image Classifier, Pattern Recognition algorithm, Live LRO Data API(For making radioactive sensors data more accurate) and the sensors data will combine to give us a final result about the value of a rock scaling from 0 - 100%.

5. After the results are calculated, the astronaut will get a notification telling them the result, and then they will decide whether to take the rock or not.

6.The robot also consists mechanical arms that can grab the sample, but this arms should be controlled by the astronaut on-board.

The Prototype :

There are two Arduinos on board, one is used for image processing where the OV7670 camera is connected to, it takes pictures every 5 seconds and writes them through serial communications. The other one is where the whole sensor processing and the communication between the commander and the robot via Bluetooth happens. The robot consistently sends the radioactive sensors’ data to the server, but the other sensors wait for the analyze command, once the analyze command is given from the board, they start a new scenario where it first takes a picture, then bombards the rock with UV lighting, takes another picture, checks for silica presence by using the IR sensor, and then finally checks for magnetic properties. At last it sends back the results to the Server. After these steps, the robot will wait for approval from the capsule.

Screenshot of Commander Application (Full source code can be found on our github repo) : Click Here To View (Google Drive Image)

Screenshot of Google's AutoML Vision Image Classifier API (Image dataset can be found on our github repo) : Click Here to View (Google Drive Image)

GITHUB SOURCE :

The link to our github repo : https://github.com/0dayboi/SpaceApps2019

Here you can find everything we coded and developed during the 3 day weekend, including the resources where we found and the whole minerals image dataset that we scraped from different sources to train our Google AutoML Vision Image Classifier API that we use for image classification. Some codes may seem uncompleted due to time limit.