FTC Vision Library

Computer Vision library for FTC based on OpenCV, featuring beacon color and position detection, as well as an easy-to-use VisionOpMode format and many additional detection features planned in the future.

Installing into FIRST Official App (recommended)

Clone FTCVision into a clean directory (outside your robot controller app) using the following command: git clone --depth=1 https://github.com/lasarobotics/ftcvision.
Navigate to the FTCVision directory that you just cloned and copy the ftc-visionlib and opencv-java folders into your existing robot controller app.
Open your robot controller app in Android Studio. Make sure you have the Project mode selected in the project browser window (so you can see all of the files in your project).
Find your settings.gradle file and append the following two lines:
```
include ':opencv-java'
include ':ftc-visionlib'
```
Find the AndroidManifest.xml file in your app's FTCRobotController/src/main folder.
Insert the following uses-permission tag in AndroidManifest.xml just below the <mainfest ...> tag and just before the application tag:
```
<uses-permission android:name="android.permission.CAMERA" android:required="true" />
```
Find the FTCRobotController/build.release.gradle AND TeamCode/build.release.gradle files and insert the following lines under dependencies into both files:
```
compile project(':ftc-visionlib')
compile project(':opencv-java')
```
Update Gradle configuration by pressing the green "Sync Project with Gradle Files" button in the header (this may take a minute)
Copy in Vision opmodes (those that end in VisionSample.java, located in [vision-root]/ftc-robotcontroller/src/main/java/com/qualcomm/ftcrobotcontroller/opmodes) from the FTCVision directory into a folder within the TeamCode directory (where your place any other opmode).
Before running the app for the first time, install the "OpenCV Manager" from the Google Play Store to enable Vision processing.
Run and test the code! Let us know if you encounter any difficulties. Note: depending on your version of the FIRST app, you may need to modify the OpModes slightly (such as adding @Autonomous) in order to get them to work for your version of the SDK.
You can now write your custom VisionOpMode!
(Optional) Add Vision testing app (see pictures of it below!) by copying all files from ftc-cameratest into the root of your project. Then, add include ':ftc-cameratest' to your settings.gradle in the root of your project. To run the camera test app, click the green "Sync Project with Gradle Files" button to update your project, then select ftc-cameratest from the dropdown next to the run button.

Installing via Git Submodule (advanced)

When installing via Git submodule, every person cloning your repo will need to run git submodule init and git subomodule update for every new clone. However, you also get the advantage of not copying all the files yourself and you can update the project to the latest version easily by navigating inside the ftc-vision folder then running git pull.

Inside the root of your project directory, run

git submodule init
git submodule add https://github.com/lasarobotics/ftcvision ftc-vision

Follow the guide "Installing into Existing Project" starting from the third bullet point. Please note that since everything will be in the ftc-vision folder and thus directories will need to be modified. Once you get to the step that modifies settings.gradle, add the following lines:
```
project(':opencv-java').projectDir = new File('ftc-vision/opencv-java')
project(':ftc-visionlib').projectDir = new File('ftc-vision/ftc-visionlib')
project(':ftc-cameratest').projectDir = new File('ftc-vision/ftc-cameratest') <- only if you want to enable the camera testing app
```
You can now write your custom VisionOpMode!

Installing from Scratch (for testing only - deprecated)

Clone FTCVision into a clean directory (outside your robot controller app) using the following command: git clone --depth=1 https://github.com/lasarobotics/ftcvision.
Open the FTCVision project using Android Studio
Copy your OpModes from your robot controller directory into the appropriate directory within ftc-robotcontroller. Then, modify the FtcOpModeRegister appropriately to add your custom OpModes.
Before running the app for the first time, install the "OpenCV Manager" from the Google Play Store to enable Vision processing.
Run and test the code! Let us know if you encounter any difficulties.
You can now write your own VisionOpMode!

Status

This library is complete as of World Championship 2016. If you have any questions or would like to help, send a note to smo-key (contact info on profile) or open an issue. Thank you!

Documentation

Documentation for the stable library is available at http://ftcvision.lasarobotics.org.

Does it work?

Yes! FTCVision can detect a beacon 0.5-4 feet away with 90% accuracy in 0.2 seconds. Here are some pictures. :smiley:

Accuracy Test

Can it detect the beacon?

Old Accuracy Test

Can it detect the beacon?

Distance Test

A test from 8 feet away

Basic Analysis Demo

FAST isn't the greatest

Ambient Analysis (Color and Rotation Detection) Test

A test from 8 feet away

Analysis Methods

FAST vs. COMPLEX

The FAST method analyzes frames at around 5 FPS. It looks for the primary colors in the image and correlates these to locate a beacon.
The COMPLEX method analyzes frames at around 2-4 FPS. It uses statistical analysis to determine the beacon's location.
Additionally, a REALTIME method exists that retrieves frames and analyzes them as fast as possible (up to 15 FPS).

Goals

To make it easy for teams to use the power of OpenCV on the Android platform
Locate the lit target (the thing with two buttons) within the camera viewfield
Move the robot to the lit target, while identifying the color status of the target
Locate the button of the target color and activate it

Progress

Beacon located successfully with automated environmental and orientation tuning.
A competition-proof OpMode scheme created so that the robot controller does not need to be modified to use the app.
Now supports nearly every phone since Android 4.2, including both the ZTE Speed and Moto G.