Skip to content

Latest commit

 

History

History
388 lines (313 loc) · 19.5 KB

testing.rst

File metadata and controls

388 lines (313 loc) · 19.5 KB

Testing imagenode with various cameras, detectors and sensors

Contents

Overview

imagenode is a part of a distributed computer vision system that runs on multiple computers. The minimal system is one Raspberry Pi running imagenode and one larger computer (such as a Mac) running imagehub. You can see the overall design of such a system in the yin-yang-ranch Github repository Yin Yang Ranch project. imagenode is dependent on having working copies of imageZMQ on both the RPi computer and the display computer. Be sure to set up and test imageZMQ BEFORE you try to install and test imagenode. The instructions for installing and testing imageZMQ are in the imageZMQ GitHub repository

Once you have run all the imageZMQ test programs, the next step is to to test imagenode by running it on a Mac or other computer with a webcam. In all of the tests below, imagenode will be tested using the receive_test.py program running on a Mac (or other display computer). The receive_test.py program running on the Mac acts as a test hub. Testing this way tests both imagenode and imageZMQ together without the need for using a full image hub at the same time. It is best to do substantial testing with this combination before adding imagehub to the mix.

Directory Structure for the Tests

Both imagenode and imageZMQ should be git-cloned to any computer that they will be running on. imageZMQ is pip installable. See the link above to imageZMQ documentation for details. I have done all testing at the user home directory of every computer the programs are used on:

~  # user home directory
+--- imagenode.yaml  # copied from test1.yaml file & edited as needed
|
+--- imagenode    # the git-cloned directory for imagenode
|    +--- docs
|    +--- imagenode
|    |    +--- imagenode.py
|    +--- tests
|    |    +--- test1.yaml
|    |    +--- test2.yaml
|    |    +--- test3.yaml
|    +--- yaml
|         +--- picam-light-test.yaml
|         +--- picam-motion-test.yaml
|         +--- picam-sensor-test.yaml
|         +--- webcam-light-test.yaml
|
+--- imagezmq   # the git-cloned directory for imageZMQ
     +--- docs
     +--- imagezmq
     |    +--- imagezmq.py  # contains the imagezmq classes
     +--- tests

Test 1: Running imagenode and a test hub with both running on a Mac

The first test runs both the sending program imagenode and the receiving receive_test.py program (acting as a test hub) on a Mac (or linux computer) with a webcam. It tests that the imagenode software is installed correctly and that imagenode.yaml file has been copied and edited in a way that works. It uses the webcam on the Mac for testing. The Mac will be displaying the images received from its webcam as well as various Light Detector specific testing windows. It tests the Light Detector which detects "lighted" versus "dark" states in a specified ROI.

  1. Make sure imageZMQ is installed and tested on your Mac or other display computer. The link to the imageZMQ GitHub repository is above.

  2. Clone the imagenode GitHub repository onto your mac in your home directory:

    git clone https://github.com/jeffbass/imagenode.git

  3. Open 2 terminal windows on your Mac. One will be used for running imagenode and the other will be used for running receive_test.py.

  4. In one terminal window, copy imagenode/tests/test1.yaml to imagenode.yaml in the home directory (~) using the command below. This test1.yaml file contains settings that will send a continuous stream of images from the webcam on the Mac while detecting "lighted" and "dark" states. You will not need to edit the yaml file. The name of the file in the home directory must be imagenode.yaml for these tests:

    cp  ~/imagenode/tests/test1.yaml  ~/imagenode.yaml

    You may want to open the imagenode.yaml file in your text editor as you do the rest of the testing below. You may want to change some of the options between test runs of imagenode.py to see how they affect detection of the "lighted" and "dark" states. This test uses the Light Detector, and you can see how the Light Detector is specified in the imaganode.yaml file (which is the file that you just copied from the test1.yaml file in the imagenode/tests directory).

  5. In the same terminal window, change to the ~/imagenode/imagenode directory. You will be running imagenode.py from here in a few more steps:

    cd ~/imagenode/imagenode

  6. In the other terminal window, change to the tests directory in the imagenode repository: ~/imagenode/tests:

    cd ~/imagenode/tests

  7. In this same terminal window, run the program receive_test.py and leave it running:

    workon py3cv3  # my virtualenv name; use yours instead
python receive_test.py

  8. In the first terminal window (in directory ~/imagenode/imagenode), run the imagenode.py program:

    workon py3cv3  # my virtualenv name; use yours instead
python imagenode.py

In about 1 minute, you should see a steam of images from the Mac's webcam appear in OpenCV display windows on the Mac. There are actually several windows stacked on top of each other. Drag them to separate areas of the screen. These windows are:

WebCamTest: the main image window showing whatever the webcam is aimed at
  (presumably yourself). The ROI for detection will be outlined in a blue
  rectangle in the upper right of the windo.
ROI: a smaller window showing the area specified in the imagenode.yaml file
  for detecting light. It appears in its own window in natural color.
Grayscale: a smaller window showing the same area shown by the ROI window,
  but showing in Grayscale.
Mean Pixel Value: a smaller window showing the average pixel intensity value
  from 0 to 255 computed from all the pixels in the Grayscale ROI.
State: a smaller window showing the current state calculated by the Light
  Detector. The text inthe window will say "lighted" or "dark" depending on
  the light intensity of the ROI area.

There will also be 1 or 2 other "mini windows", which are really just the name bars for the windows. The purpose of these "window name bars" is to show what event messages the imagehub would have recorded to the imagehub event log. If you have 2 windows (because you have changed the brightness of the ROI area), one window bar will say "WebCamTest |light | lighted" and the other window bar will show "WebCamTest |light | dark".

If you move a darker object into the camera view of the ROI area, you will see the Mean Pixel Value change and, depending on how light or dark the object is, the State will change from "Lighted" to "Dark".

You can change the option values of the light detector to run experiments with the Light Detector.

  1. Stop the imagenode.py program running in the imagenode terminal window by pressing Ctrl-C. Edit the ~/imagenode.yaml file to change the threshold value to a different value.
  2. Rerun the imagenode.py program and watch what happens.

You can leave the test hub program receive_test.py program running while you stop the imagenode.py program, change the yaml file, and restart the imagenode.py program.

You can experiment with other option setting values as well. You can read about the option settings and get an explanation of the file and adjusting the settings in imagenode Settings and the YAML files.

All of these windows are used for testing. In a production use of imagenode, and imagehub, the event messages and the event related images would be stored in appropriate directories on the imagehub computer. The windows would not be shown on the hub computer because the send_test_images option would be set to False in the imagenode.yaml file.

Press Ctrl-C in each window to end both the test programs.

Test 2: Testing imagenode running on a RPi with a test hub running on a Mac

The second test runs the sending program imagenode on an RPi with a PiCamera and the program receive_test.py (acting as a test hub) on a Mac (or linux computer). The Mac will be displaying the images received from the RPi PiCamera as well as various detector specific testing windows. It tests that the imagenode software is installed correctly on the RPi and that the imagenode.yaml file has been copied and edited in a way that works. It tests the Light Detector which detects "lighted" versus "dark" states in a specified ROI in the field of view of the PiCamera.

  1. Make sure imageZMQ is installed and tested on your Mac or other display computer. The link to the imageZMQ GitHub repository is above.

  2. Make sure imageZMQ is installed and tested on your RPi that has a PiCamera that will be sending images to test the Light Detector. The link to the imageZMQ GitHub repository is above.

  3. Clone the imagenode GitHub repository onto your RPi in the home directory (typically the "pi" user home directory):

    git clone https://github.com/jeffbass/imagenode.git

    Your directory structure on your RPi should be like the directory structure described above.

  4. Open 2 terminal windows on your Mac. One will be used for running imagenode on RPi and the other will be used for running receive_test.py as a test hub on the Mac.

  5. In one terminal window, ssh into the RPi. Copy imagenode/tests/test2.yaml to imagenode.yaml in the home directory (~) using the command below. This test2.yaml file contains settings that will send a continuous stream of images from the PiCamera to the Mac while detecting "lighted" and "dark" states. The name of the file in the home directory must be imagenode.yaml for these tests:

    cp  ~/imagenode/tests/test2.yaml  ~/imagenode.yaml

    Open a text editor in your RPi terminal window. Edit the ~/imagenode.yaml file to change the H1 hub address to point to the TCP address of you Mac that will be acting as a hub.

    While you are editing the H1 hub address in the imagenode.yaml file, you may want set the vflip option to True. I find that in over half of my RPi PiCamera setups, the camera is positioned upside down; it has to do with the way the PiCamera cable connects to the main board. Setting the vflip option to True will cause the image to be vertically flipped.

    You may want to open the imagenode.yaml file in your RPi text editor as you do the rest of the testing below. You may want to change some of the options between test runs of imagenode.py and see how they affect detection of the "lighted" and "dark" states. This test uses the Light Detector, and you can see how the Light Detector is specified in the imaganode.yaml file (that is the file that you just copied from the test2.yaml file in the imagenode/tests directory).

  6. In the same RPi terminal window, change to the ~/imagenode/imagenode directory. You will be running imagenode.py from here in a few more steps:

    cd ~/imagenode/imagenode

  7. In the other terminal window, which is going to be used to run the test hub on the Mac, change to the tests directory in the imagenode repository:

    cd ~/imagenode/tests

  8. In this same Mac terminal window (in the ~/imagenode/tests directory), run the program receive_test.py and leave it running:

    workon py3cv3  # my virtualenv name; use yours instead
python receive_test.py

  9. In the RPi terminal window (in directory ~/imagenode/imagenode), run the imagenode.py program:

    workon py3cv3  # my virtualenv name; use yours instead
python imagenode.py

In about 1 minute, you should see a steam of images from the Mac's webcam appear in OpenCV display windows on the Mac. There are actually several windows stacked on top of each other. Drag them to separate areas of the screen. These windows are the same as the windows described above, except that the name of the main image window will be "PiCameraTest".

There will also be 1 or 2 other "mini windows", which are really just the name bars for the windows. The purpose of these "window name bars" is to show what event messages the imagehub would have recorded to the imagehub event log. If you have 2 windows (because you have changed the brightness of the ROI area), one window bar will say "WebCamTest |light | lighted" and the other window bar will show "WebCamTest |light | dark".

If you move a darker object into the camera view of the ROI area, you will see the Mean Pixel Value change and, depending on how dark the object is, the State will change from "Lighted" to "Dark".

You can change the option values of the light detector to run experiments with the Light Detector.

  1. Stop the imagenode.py program running in the RPi imagenode terminal window by pressing Ctrl-C. Edit the ~/imagenode.yaml file to change the threshold value to a different value.
  2. Rerun the imagenode.py program and watch what happens.

You can experiment with other option setting values as well. You can read about the option settings with an explanation of the file and adjusting the settings in imagenode Settings and the YAML files.

All of these windows are used for testing. In a production use of imagenode, and imagehub, the event messages and the event related images would be stored in appropriate directories on the imagehub computer. The windows would not be shown on the hub computer because the send_test_images option would be set to False in the imagenode.yaml file.

Press Ctrl-C to end the test programs on both the Mac and the RPi.

Test 3: Testing imagenode running a motion detector on the RPi to Mac Hub

The third test runs the sending program imagenode on an RPi with a PiCamera and the receive_test.py (acting as a test hub) on a Mac (or linux computer). It is run exactly the same way as Test 2, above. The Mac will be displaying the images received from the RPi PiCamera as well as several motion detector specific testing windows. Test 3 tests the Motion Detector which detects "moving" versus "still" states in a specified ROI in the field of view of the PiCamera.

To run the motion detector test with the RPi PiCamera sending images and events to the Mac running the test hub, follow all the steps in Test 2, with one change. In step 5, copy imagenode/tests/test3.yaml to imagenode.yaml in the home directory (~) using the command below. This test3.yaml file contains settings that will send a continuous stream of images from the PiCamera to the Mac while detecting "moving" and "still" states. The name of the file in the home directory must be imagenode.yaml for these tests:

cp  ~/imagenode/tests/test3.yaml  ~/imagenode.yaml

After copying the yaml file, edit the file to have it point to your Mac's hub address. Also, change the vflip option if you need to. Run the rest of the steps the same way as Test 2, above. A different set of Motion Detector windows will appear on the Mac display:

PiCameraTest: the main image window showing whatever the PiCamera is aimed at.
  The ROI for motion detection will be outlined in a blue rectangle in the
  upper right of the window.
ROI: a smaller window showing the area specified in the imagenode.yaml file
  for detecting motion. It appears in its own window in natural color.
Grayscale: a smaller window showing the same area shown by the ROI window,
  but showing in Grayscale.
frameDelta: a smaller window showing the same area shown by the ROI window,
  but showing the frameDelta difference, pixel by pixel between the most
  recent image and the average of previous images.
thresholded: a smaller window showing the same area shown by the ROI window,
  but showing the motion areas thresholded so they are all white versus the
  non motion areas being all black.
Area: a smaller window showing the computed area of the contours around pixels
  thresholded as moving.
N Contours: The number of contours in the ROI around thresholded pixels.
State: a smaller window showing the current state calculated by the Light
  Detector. The text in the window will say "moving" or "still" depending on
  the light intensity of the ROI area.

If you wave a hand or an object in the ROI of the PiCamera, you will see it in the various windows and see the thresholded contours in the threshold window. You will also see the values of the calculations and the final state of "moving" or "still". It would be helpful to experiment with different values for the options in the ~/imagenode.yaml file and see what impact it has on the various motion detection windows.

Press Ctrl-C to end the test programs on both the Mac and the RPi.

Test 4: Testing imagenode temperature sensor on the RPi to Mac Hub

The fourth test tests the capability of imagenode to capture and send temperature sensor readings. It also uses the Mac running the program receive_test.py as a test hub as in the previous tests. To run this test you will need a DS18B20 temperature sensor appropriately attached to GPIO pin 4 of the RPi.

Set things up as in Test 2. Then, in the RPi terminal window, copy test3.yaml:

cp  ~/imagenode/tests/test3.yaml  ~/imagenode.yaml

After copying the yaml file, run the hub and RPi programs the same way as in Test 2, above. A small window that is only a window title bar will appear on the Mac display:

RPi |Temp | 72 F

The test4.yaml settings file causes the temperature sensor thread to start and report temperature values once per minute. It will report the temperature once per minute even if the temperature doesn't change because the min_difference is set to 0. If the min_difference had ben set to 1, for example, the temperatures would be reported only if they changed by at least 1 degree. Press Ctrl-C to end the test programs on both the Mac and the RPi. Note that because a timer thread is running to read the sensor probe, it is likely that an exception thread traceback message will be printed to the console after you press Ctrl-C in the RPi terminal window. That's normal.

Testing imagenode running on one or more RPi's using imagehub

After you have tested imagenode with receive_test.py running as a test hub, the next step would be to add a full imagehub program to the mix. In this arrangement, the imagehub program would be started on a Mac or Linux computer. One or more RPi's would have their ~/imagenode.yaml files changed to assign appropriate detectors and point to appropriate hub address. In my production use cases, a single imagehub is able to receive detector event messages and detector event images from 8 RPi's at a time without significantly impacting the framerates of the RPi's. To test the imagenode software with imagehub, git clone and then run the test programs in the imagehub GitHub repository.

Return to main documentation page README.rst