Treasure Box

Build the Circuit:

• Place the motor controller on the Raspberry Pi starting at pin 1 with the orientation as shown in the picture.
• To make it easier to move everything into the box later make your connections to the Pi and motor controller with extra jumper wires. Use M-F wires for the lock and battery box and F-F wires for the button.
• Connect the rotating door lock DC motor to Motor A on the motor controller being sure to connect the red and black wires as shown.
• There is a switch on the battery box. Turn it OFF for now. Connect the battery box to the motor controller. It is critical that you connect the red wire to VIN and the black wire to GND.
• Connect the button (momentary switch) to BOARD pin 40 (i.e. BCM GPIO 21) and BOARD pin 39 (Ground).
• Connect the camera to the camera port and use the helping hands to hold it. It's not shown in the picture to the left because I thought it would be too busy.

Test the Button (Momentary Switch)

• Create a new Python file from the IDLE shell and copy and paste this code. You can click inside the black box area where the code is and use ctrl+a to select all of the code. Use ctrl+c to copy and ctrl+v to paste.
• Notice how this code uses BCM to identify the GPIO pins. We connect the button to BOARD pin 40 but will be identifying it using BCM GPIO 21.
• We don't have to connect a resistor in series with the button because we are using the Pi's internal resistors. That setup is activated in line 9 GPIO.setup "pull_up_down"
• The "while True:" statement sets up an infinite loop where the Pi will be looking and waiting for a button press.
• Run the program and press the button a couple of times to make sure it is working.
• Use ctrl+c to halt the program when you are done so the GPIO pins return to the zero input state.

Test the Rotating Door Lock DC Motor

• Create a new Python file from the IDLE shell and copy and paste this code. Name it 02b-doorLock.py.
• Here you have to think about what position the lock is in initially. Is it locked or unlocked? Will you need to lock the door or unlock it? There are three programs to test the locking mechanism (02a-doorLockTest, 02b-doorLock, and 02c-doorUnLock). You can find all of these files at the end of this page.
• All of the tests rely on three main functions
  • StopLock()
  • Lock(t)
  • UnLock(t)
• 02a-doorLockTest.py will call both Lock() and Unlock() to test the full range of motion of the door lock.
• 02b-doorLock.py will move the lock from the unlocked position to the locked position.
• 02c-doorUnLock.py will move the lock from the locked position to the unlocked position.
• Notice how the Lock(t) and UnLock(t) functions take a timing argument "t". The locks I tested took approximately 5.5 seconds to rotate to their position. You may have to use a different time depending on your setup.
• Before you run the programs remember to turn the switch on the battery box to ON. The battery box is what supplies power to the door lock motor.
• Run the programs to test the functionality of the locking mechanism. Remember to avoid trying to turn the lock to a position it is already in. You could burn out the motor.

Lock the Door Using a Button Press

• Create a new Python file from the IDLE shell and copy and paste this code. Name it 03a-buttonLock.py.
• We combine the button press and door lock code together so that the door locks when we press the button. Don't keep pressing the button because it will keep trying to lock the door even if it is already locked. You could burn out the motor by doing that.
• You can use the 02c-doorUnLock.py program to unlock the door when needed.
• Run the programs to make sure the button press and locking mechanism can work together.
• For fun I created 03b-buttonLock.py. It will lock the door on the first button press and unlock the door on the second button press. It will continue to lock on the odd button press and unlock on the even button press.

Active vs Passive Cooling

• The following Computer Vision programs are incredibly hard computationally for the Raspberry Pi.
• Imagine if we asked you to run as fast as you can for 100 yards. You would probably get a little hot and tired. Now imagine that we asked you to run as fast as you can for 100 miles. You would probably die of heat exhaustion well before you made it the full 100 miles.
• When we do computationally intensive tasks like computer vision we're asking the Pi to run as fast as it can for as long as it takes to solve the problem. We need some way to cool it.
• In the picture to the left you will see some red heat sinks on the CPU (larger) and Ethernet (smaller) controller chips. Unfortunately, heat sinks will not be enough to keep the CPU cool enough while we're running the computer vision programs.
• We will need to use a computer fan to actively cool the heat sink that's on the CPU. We'll use a AA battery pack to run the fan instead of a 9 volt battery.
• When you are running your test programs outside the treasure box you might need to hold the fan over the Pi or set it next to it since we don't have a mount to hold it.

Checking the CPU temperature

• You should check the CPU temperature by typing in a second terminal window
  • vcgencmd measure_temp
• The normal operating temperature of the CPU is around 45-55C. If the CPU hits 82C, you will hit what's know as thermal throttling. The CPU will reduce it's speed from 1200 MHz to 600 MHz to try to save itself. So instead of running the CPU will start walking.

Cat Face Detection Using OpenCV

• Don't forget to connect the Raspberry Pi camera if you haven't done so already. This code is going to use the camera to look for a cat's face.
• We asked you to get a picture of your cats face or download one from google. Have that ready on your phone or have a printed picture. You will need a picture to put in front of the camera.
• In the same terminal window that you were working in the section above that was in the (CV) virtual environment, type in
  • python -m idlelib
• In the python shell that opens create a new file, copy and paste the contents of the code to the left, and call it 04-catFaceDetection.py.
• Notice that the CascadeClassifier is haarcascade_frontalcatface_new.xml. This Haar Cascade gives less false positive detections than the one we were using before to detect cat faces in pictures. You don't want to open the box when there isn't a cat there.
• REMEMBER: You will need to actively cool the CPU with the computer fan while you run this program.
• Run the program. You will see a window appear that is the live camera feed.
• Hold the picture of your cat in front of the camera and see if it draws a rectangle around it's face. You may need to hold it closer or further away from the camera for it to work.
• In a second terminal window check the CPU temperature by typing in
  • vcgencmd measure_temp
• Instead of retyping the command you can use your up arrow key and enter to run the command again to see how the temperature is changing. You might have to do this a couple of times until you are confident that the temperature has stabilized at something below 70C.
• Re-position the fan if you have temperatures higher than 70C.
• If you have problems detecting the cats face, you will need to adjust scaleFactor and minNeighbors.

Almost There (debugging/improving)

• First make sure your door lock is in the locked position.
• Create a new Python file from the IDLE shell and copy and paste this code. Name it 05a-catTreasureBox.py.
• This time you will not see a live feed of the video. Remember when you have all the components in the box you won't have a monitor in there. We can reduce the CPU load by not showing the video feed. However, the camera is still active and capturing video. It's just not shown on the screen.
• Run the program.
• Hold the picture of your cat in front of the camera. The door should unlock. You may need to hold the picture closer or further away from the camera for it to work.
• Remember if you still have problems detecting it's face you will need to adjust scaleFactor and minNeighbors to the numbers you found in the program above.

Program Notes:

• if not isinstance(faces, tuple): When a cat face is detect the faces object becomes a numpy array. In simple terms a coordinate map of a rectangle (x position, y position, width and height). It turns out that when the faces object is empty it's just a tuple. We can use the isinstance() function to identify when a cat face is detected.
• One problem with this program is that it does not know when the door is locked or unlocked. It will try to drive the motor to the unlocked position even if the door is already unlocked.
• I've added a break statement so the program will terminate after the first UnLock() call.
• Notice that I don't have a break statement for the button press so you still have to be careful with that and not try to lock the door when it's already locked.

Cat Facial Recognition Treasure Box

• To let the program know when the door is locked or unlocked I added a variable lockPosition.
• It is initially set to lockPosition = "unlocked" because the lock mechanism has to be unlocked for you to actually close the box.
• You will see a couple of extra "if" statements.
• if lockPosition == "unlocked" is used so that when you press the button the door will lock if lockPosition is equal to unlocked but do nothing if it isn't. We could press on the button all day and won't have to worry about destroying the lock motor!
• if lockPosition == "locked" is used so that if the camera sees a cat's face and the door is already unlocked, the program won't try to keep unlocking it.