composer colour check — Sharp AI

Skip to content

composer colour check

To perform a color check using the robot's camera, you should use the Color Sensor processor via the ConceptVisionColorSensor Sample OpMode, which determines the predominant color in a specified Region of Interest (ROI). As explained in the official FTC Docs, this processor allows the camera to detect colors from a much greater distance than traditional physical color sensors. The system uses a "k-means" clustering algorithm to group similar colors within the ROI, identifying the cluster with the most pixels as the predominant color.

To set up this color check, the FTC Docs instruct you to configure two main inputs in the Sample OpMode. First, you define the ROI, which can be set to the entire camera frame, a sub-region using standard image coordinates, or a sub-region using a normalized coordinate system. Second, you provide a list of candidate color "Swatches" (such as RED, BLUE, or YELLOW) that you expect to find, which the software will use to select a match.

When running the OpMode, the Driver Station will display telemetry indicating the matched color inside your specified zone. For more advanced applications, the FTC Docs note that teams can bypass the generic swatch matches and directly extract the raw RGB color values from the ROI to process in their own custom code.

Was this right?

Sources16

1

Color Processing Color Sensor

FTC Docs·BSD 3-Clause

A simple way to use FTC's new OpenCV vision tools is to operate a "Color Sensor". Namely, it can determine the color seen by the robot's camera, in a specified zone.

Below, the small central rectangle is the region being evaluated:

(Figure: Color sensor detection zone)

A key benefit is that the camera can be much further away from the object than, for example, a REV Color Sensor or others like it.

It's still important to accurately point the camera and carefully select the image zone to inspect.

For the above example, OpenCV can provide results like this:

(Figure: RED Detection using Color Sensor)

The following sections describe how to do this, with a Sample OpMode.

2

Color Processing Introduction > Two Processors

FTC Docs·BSD 3-Clause

The new software includes these processors:

Color Sensor - detects the color of a specified zone in the camera's

view

Color Locator - gives detailed info on clusters of a specified color,

in a specified zone

This tutorial has a Color Sensor page, showing how to use the Sample OpMode called ConceptVisionColorSensor.

For the Color Locator processor, the color "clusters" are called Blobs. As listed above, this tutorial offers one page on Color Blob Concepts, and four pages covering the Sample OpModes called ConceptVisionColorLocator_Rectangle and ConceptVisionColorLocator_Circle.

The Sample OpModes are available in FTC Blocks, and in Java for use in OnBot Java or Android Studio. Each programming section of this tutorial has a Blocks tab and a Java tab.

3

Color Locator (Discover)

FTC Docs·BSD 3-Clause

Another way to use FTC's new OpenCV vision tools is to operate a "Color Locator". Namely, it can find a target color that you specify.

As with the Color Sensor tool, you can specify a Region of Interest (ROI). Only that zone of the camera's view will be searched for the target color.

The "target color" is actually a range of numerical color values, for a better chance of finding the desired color.

OpenCV will form "Blobs" of that color. As described in the Concepts page here, a Blob is a contiguous cluster of similar-colored pixels.

Blobs often have a complex, irregular shape or contour, so they are represented here by a best-fit rectangle called "boxFit".

(Figure: Zoomed Color Blob detection)

The target color here is BLUE. The white rectangle is the Region of Interest (ROI), the teal jagged line is the Blob's contour (fully inside the ROI), and the purple rectangle is the boxFit.

The software reports the size, position and orientation of each "boxFit". That data can be evaluated by your OpMode for robot navigation and other actions.

The following sections describe how to do this, with a Sample OpMode.

4

FTC SDK API > org.firstinspires.ftc.vision.opencv > ColorBlobLocatorProcessor

FTC SDK API·BSD 3-Clause

The ColorBlobLocatorProcessor finds "blobs" of a user-specified color in the image. You can restrict the search area to a specified Region of Interest (ROI).

Open FTC SDK API

5

Color Processing Color Sensor > Next Sections

FTC Docs·BSD 3-Clause

Soon you can move ahead to try the Color Locator processor.

But first, learn a few basic concepts at this tutorial's Color Blob Concepts page.

============

Questions, comments and corrections to westsiderobotics@verizon.net

6

Color Processing Introduction > Next Steps

FTC Docs·BSD 3-Clause

Time to get started!

Following this tutorial in order, first try the Sample OpModes for Color Sensor.

Then read about Color Blob Concepts, and try the Color Locator Sample OpModes.

Soon you'll be ready to add one or both features to your Autonomous OpModes -- perhaps even to help automate your TeleOp!

============

Questions, comments and corrections to westsiderobotics@verizon.net

7

Color Blob Concepts > Color Processing

FTC Docs·BSD 3-Clause

Now let's point the camera at an INTO THE DEEP game element called a Sample.

(Figure: Blob from a Blue SAMPLE)

Here the software was told to seek shades of blue. The orange rectangle encloses a Blob of blue color.

But why doesn't the rectangle enclose the entire game piece? The software is processing only a certain Region of Interest or ROI. That's the white rectangle; its size and location are specified by you.

Anything outside the ROI will not be considered part of any Blob that is detected. This can help you avoid detecting (unwanted) background objects.

In the example above, the Blob was actually outlined in teal (blue-green color), very hard to see. Let's try another image:

(Figure: Teal Outline of Blue Blob)

Now the teal outline of the blue Blob can be seen. Its shape is irregular, which can be difficult for your OpMode to evaluate.

8

Color Processing Color Sensor > Advanced Use

FTC Docs·BSD 3-Clause

Some teams may prefer to read and evaluate the actual RGB color values, rather than rely on a generic Swatch result.

RGB is a Color Space that uses three numerical components of Red, Green and Blue. Values range from 0 to 255. For more info, see this tutorial's Color Spaces page.

Extracting the RGB components can be seen in the Telemetry portion of the Sample OpMode. Click the Blocks or Java tab:

Blocks Here are the RGB components of the ROI's predominant color:

(Figure: Finding color by RGB)

Note: the Color Block has a drop-down list that includes Hue, Saturation and Value. Those settings will not work here, to produce components in the HSV Color Space, because the source Block provides only RGB color (its method name is .rgb).

Java Here are the RGB components of the ROI's predominant color:

Color.red(result.rgb)
Color.green(result.rgb)
Color.blue(result.rgb)

For Blocks or Java, those component values can be assigned to numeric variables, with names like ROIRedValue, ROIGreenValue, and ROIBlueValue.

Now your code can process those RGB variables as desired.

9

Color Processing Color Sensor > Modify the Sample

FTC Docs·BSD 3-Clause

This Sample OpMode is designed for the user to select/edit two inputs:

define the Region of Interest (ROI)
list the colors that might be found

For the first input, there are 3 ways to define the ROI:

entire frame
sub-region, defined with standard image coordinates
sub-region, defined with a normalized +/- 1.0 coordinate system

For the second input, you must list the candidate colors from which a result will be selected as a "Match".

Simply choose from the 10 "Swatches": RED, ORANGE, YELLOW, GREEN, CYAN, BLUE, PURPLE, MAGENTA, BLACK, WHITE. For efficiency, add only those Swatches for which you reasonably expect to get a match.

The Blocks and Java OpModes contain detailed comments to guide you through these edits. They are not repeated in this tutorial.

10

Color Locator (Challenge) > Advanced Development

FTC Docs·BSD 3-Clause

Searching for multiple colors is possible by building multiple processors and adding them to the same VisionPortal. This allows different ROIs, for example, that can overlap if desired.

(Figure: Using two processors)

The next ColorLocator page called Color Locator (Round Blobs) covers detection of round objects.

============

Questions, comments and corrections to westsiderobotics@verizon.net

11

Color Locator (Explore)

FTC Docs·BSD 3-Clause

This Explore page shows how to modify the default settings of the ColorLocator Sample OpMode. It assumes you have already followed this tutorial's previous Discover page, to open and test this OpMode.

ColorLocator has only two required inputs:

target color range
Region of Interest (ROI)

Both of these can be specified in multiple ways.

This Explore page covers settings that are already in the Sample:

which contour types to process
whether to draw contours in the preview
pre-processing of the image
camera resolution
post-filter the Blob results
post-sort the Blob results

12

Color Locator (Explore) > Image Pre-Processing

FTC Docs·BSD 3-Clause

The default Sample OpMode purposely blurs the camera's image. This "pre-processing" happens before OpenCV performs Blob formation, thus affecting the contours seen in DS and RC previews.

The effect is very small (default kernel size of 5x5 pixels), but can significantly improve Blob formation, giving more useful results.

Blurring is one of three available image adjustments to improve processing results. You can experiment with these advanced tools, after studying their usage. See links at the section below called More Documentation.

The Blocks and Java Sample OpModes give this description:

Select and read the Blocks or Java section below:

Blocks (Figure: Pre-processor Options)

Java In the Sample OpMode, follow the instructions at about lines 94-107 to set the desired pre-processing at about line 114.

.setBlurSize(int pixels)
.setErodeSize(int pixels)
.setDilateSize(int pixels)

Any of these pre-processing settings can be disabled by setting their pixel value to zero, or by removing the command.

In the FTC processor, any specified erosion is performed before dilation. This removes specular noise, then returns the remaining blobs to a size similar to their original size. (This also will not be on the final.)

13

Color Processing Introduction

FTC Docs·BSD 3-Clause

The FIRST Tech Challenge SDK software (as of v10.1) includes some Color Processing features from OpenCV, a popular and powerful open-source library for vision processing.

Introduced with INTO THE DEEP, these new features will help FIRST Tech Challenge teams identify colors and process color blobs, useful in any game requiring color and shape recognition.

Here's the outline of this tutorial's main pages:

Much credit to developer and Sample OpMode author @gearsincorg (https://github.com/gearsincorg), EasyOpenCV developer @Windwoes (https://github.com/Windwoes), FTC Blocks developer @lizlooney (https://github.com/lizlooney), and the open-source team at OpenCV (https://opencv.org/).

14

Vision Processor Initialization > Color Processing Initialization

FTC Docs·BSD 3-Clause

Step 2 is similar for creating the Color Sensor/Locator Processor, software that evaluates frames streaming in from the camera.

The Builder is going to look a little different depending on if you are using the Color Sensor or the Color Locator. For more information, see the Color Processing Documentation.

15

Color Processing Color Sensor > Sample OpMode > Running the Sample OpMode

FTC Docs·BSD 3-Clause

On the Driver Station:

Select the Autonomous OpMode that you just saved or built.
Turn off the automatic 30-second match timer (green slider).
Touch INIT only.

The OpMode should give Telemetry, stating the main "matched" color inside the Region of Interest.

(Figure: Driver Station Telemetry)

Move the camera around, and watch the Telemetry area on the Driver Station screen. It should state "BLUE" when pointing at a blue object, and likewise should identify other common colors.

It's working! You have a color sensor in your robot camera. Think about how to use this in the FTC Robot Game.

Skip the next two sections if you already know how to use FTC previews.

16

Color Processing Color Sensor > Building the VisionPortal

FTC Docs·BSD 3-Clause

The Sample OpMode first creates a "Predominant Color" Processor using the Builder pattern. This is the same Builder pattern used to create an AprilTag Processor.

The Sample OpMode then creates a VisionPortal, again using a Builder pattern. This includes adding the "Predominant Color" Processor to the VisionPortal.

How does OpenCV determine the "predominant color" of the ROI? An algorithm called "k-means" (https://en.wikipedia.org/wiki/K-means_clustering) determines clusters of similar colors. The color of the cluster with the most pixels is called "predominant" here. (This will NOT be on the final.)