Email Us!

Our Website

Sabercat Robotics Vision Processing Utilities

About

---

This is the main vision repository for Saguaro High School robotics team's computer vision code. We are FRC team 4146, the Sabercats. The objective of the code is to provide tooling for designing vision systems as well as providing a fast in-game vision processing.

The current file tree is as follows:

.
├── arch_light_track.py
├── log.py
├── README.md
├── saber_track.py
├── scripts
│   ├── install_cv_arch.sh
│   ├── install_cv_deb.sh
│   ├── install_deps.cmd
│   └── install_deps.sh
├── settings.json
├── test.sh
├── todo.md
└── track.py

All of the installation scripts are located in /scripts/ arch_light_track.py is a stipped down vision processing program for use on the raspberry pi. track.py is the entry point of the vision tuning application. settings.json can be used as settings for the application. saber_track.py contains the most of the application code.

Quick Example

---

There is only one cli argument now! This is the settings. The arguemnt defines the filename of the .json settings file.

python track.py -s settings

This will load settings.json as the settings file.

Settings

---

Here is an example of the json settings file. These are all of the current possible settings in the application.

{
  "port" : "",
  "filters": "test",
  "hsv": "",
  "original": "True",
  "display": "True",
  "in_file": "",
  "out_file": "",
  "log_name": "debug_log",
  "vidmode" : "0"
}

• port is the ip address of the roborio.
• filters are the filter set names to track, leave spaces between words for multiple object tracking.
• hsv Boolean weather to have hsv window show.
• original Boolean weather to have original window show.
• display Boolean weather to have a display at all.
• in_file string of what to name the input file.
• out_file string name of output file that stores hsv color thresholds.
• log_name name of debug log.
• vidmode int camera number.

Filters

---

The program uses colors to track objects. This means that it may get confused if there is a lot of 'noise' it the image. To tell the program which object it should track, we use filters. Filters are just a color reduction where we create an image from the camera stream where each pixel that falls within the color range will be white, and the pixels that are outside of the defined color range are black. We use sliders to adjust the color filters. In order to simplify the color space, we use HSV. HSV is a colorspace that stands for Hue, Saturation, and Value.

Apply color thresholds with the "filter" value in settings.

"filter": "one_filter two_filter"

The above setting will generate two filter thresholds to track. One called one_filter, the other called two_filter. The following is an example of the setup application.

This generates two color filters. It makes a window and trackbar for each of the filters. As well as displays a labeled bounding box over the object.

If you have an output file in the settings, when you exit the slider positions will be saved. If you have an input file, it will take the given track positions and apply them.

Onboard Robot Example

---

You'll want to adjust the color space in the enviroment you will use it in. This means you need to tweak the hsv values on game day in order to adjust for lighting and unusual noise.

For instance if you are trying to track the goal, you should go to the practice match and adjust the HSV values.

python arch_light_track.py test.json 192.168.1.2

Where test.json is the json output file generated and the ip is the ip of the roborio.

This will generate a track file when you QUIT THE SESSION WITH THE Q KEY If you didn't notice the bold caps, you need to use the q key on exit. You cannot use the keyboard disrupts that command lines have.

If you generated that file without the raspberry pi, you may transfer the file to the scp utility.

Use daemons to make a startup script for your raspberry pi.

Log files

The debug log generated should be located in debug_log.log where log name is the json setting for "log_name": "debug_log",

Versions

---

• 0.0.5 json settings format added, cli settings removed.
• 0.0.4 Switched to json track serialization. Runs on arch now.
• 0.0.3 Added tagging support. Fixed bugs. Cleaned code. Added more documentation.
• 0.0.2 Added multiple filtering capabilities.
• 0.0.1 Added more modes and file I/O. Cleaned code.
• 0.0.0 Base opencv HSV filtering and trackbar.

Current utilities include:

• log.py is a simple logging module for streamlined debugging and records (cv2 independent)
• saber_track.py The main module for video tracking
• track.py A small test script showing off the saber_track module

Installation

Install Arch on the Rapsberry Pi

Essentially, just follow this thread

When installing opencv, login to root with username: root and password: root

Install OpenCV

Windows

The best way to install opencv for python is to get Anaconda. This is a python distribution with its own package manager that is more powerful than pip. This packagemanager is called conda. This allows you to easily install numpy on windows. Numpy for Anaconda. When Anacona is installed, make sure it works. In cmd, run python. This should open up a REPL that will prompt you with >>>. You should then be able to import numpy if this fails, you did something wrong. Run the install_deps script. install_deps.cmd. Now, you should Install opencv 3.0 python.

Linux (Ubuntu based)

You can just run the install scripts. sh insall_cv_deb.sh You might have to play arround with the script a bit. If you find a problem, please file a pull request, or contact me.

Rapsberry Pi (Arch)

You should be logged in as a super user. Type su input your password. Then run install_cv_arch.sh (its in the scripts folder). This should do all of the work for you. Keep in mind, it's untested. Contact me if you have any problems.

Note: If you're having a problem, It might be solved by updating your system. pacman -Syu

Contact