Code:
package $package;
import java.lang.Math;
import java.util.Comparator;
import java.util.Vector;
import com.ni.vision.NIVision;
import com.ni.vision.NIVision.Image;
import com.ni.vision.NIVision.ImageType;
import edu.wpi.first.wpilibj.CameraServer;
import edu.wpi.first.wpilibj.SampleRobot;
import edu.wpi.first.wpilibj.Timer;
import edu.wpi.first.wpilibj.smartdashboard.SmartDashboard;
/**
* Example of finding yellow totes based on color.
* This example utilizes an image file, which you need to copy to the roboRIO
* To use a camera you will have to integrate the appropriate camera details with this example.
* To use a USB camera instead, see the SimpelVision and AdvancedVision examples for details
* on using the USB camera. To use an Axis Camera, see the AxisCamera example for details on
* using an Axis Camera.
*
* Sample omages can be found here: http://wp.wpi.edu/wpilib/2015/01/16/sample-images-for-vision-projects/
*/
public class Robot extends SampleRobot {
//A structure to hold measurements of a particle
public class ParticleReport implements Comparator<ParticleReport>, Comparable<ParticleReport>{
double PercentAreaToImageArea;
double Area;
double ConvexHullArea;
double BoundingRectLeft;
double BoundingRectTop;
double BoundingRectRight;
double BoundingRectBottom;
public int compareTo(ParticleReport r)
{
return (int)(r.Area - this.Area);
}
public int compare(ParticleReport r1, ParticleReport r2)
{
return (int)(r1.Area - r2.Area);
}
};
//Structure to represent the scores for the various tests used for target identification
public class Scores {
double Trapezoid;
double LongAspect;
double ShortAspect;
double AreaToConvexHullArea;
};
//Images
Image frame;
Image binaryFrame;
int imaqError;
//Constants
NIVision.Range TOTE_HUE_RANGE = new NIVision.Range(24, 49); //Default hue range for yellow tote
NIVision.Range TOTE_SAT_RANGE = new NIVision.Range(67, 255); //Default saturation range for yellow tote
NIVision.Range TOTE_VAL_RANGE = new NIVision.Range(49, 255); //Default value range for yellow tote
double AREA_MINIMUM = 0.5; //Default Area minimum for particle as a percentage of total image area
double LONG_RATIO = 2.22; //Tote long side = 26.9 / Tote height = 12.1 = 2.22
double SHORT_RATIO = 1.4; //Tote short side = 16.9 / Tote height = 12.1 = 1.4
double SCORE_MIN = 75.0; //Minimum score to be considered a tote
double VIEW_ANGLE = 49.4; //View angle fo camera, set to Axis m1011 by default, 64 for m1013, 51.7 for 206, 52 for HD3000 square, 60 for HD3000 640x480
NIVision.ParticleFilterCriteria2 criteria[] = new NIVision.ParticleFilterCriteria2[1];
NIVision.ParticleFilterOptions2 filterOptions = new NIVision.ParticleFilterOptions2(0,0,1,1);
Scores scores = new Scores();
public void robotInit() {
// create images
frame = NIVision.imaqCreateImage(ImageType.IMAGE_RGB, 0);
binaryFrame = NIVision.imaqCreateImage(ImageType.IMAGE_U8, 0);
criteria[0] = new NIVision.ParticleFilterCriteria2(NIVision.MeasurementType.MT_AREA_BY_IMAGE_AREA, AREA_MINIMUM, 100.0, 0, 0);
//Put default values to SmartDashboard so fields will appear
SmartDashboard.putNumber("Tote hue min", TOTE_HUE_RANGE.minValue);
SmartDashboard.putNumber("Tote hue max", TOTE_HUE_RANGE.maxValue);
SmartDashboard.putNumber("Tote sat min", TOTE_SAT_RANGE.minValue);
SmartDashboard.putNumber("Tote sat max", TOTE_SAT_RANGE.maxValue);
SmartDashboard.putNumber("Tote val min", TOTE_VAL_RANGE.minValue);
SmartDashboard.putNumber("Tote val max", TOTE_VAL_RANGE.maxValue);
SmartDashboard.putNumber("Area min %", AREA_MINIMUM);
}
public void autonomous() {
while (isAutonomous() && isEnabled())
{
//read file in from disk. For this example to run you need to copy image20.jpg from the SampleImages folder to the
//directory shown below using FTP or SFTP: http://wpilib.screenstepslive.com/s/4485/m/24166/l/282299-roborio-ftp
NIVision.imaqReadFile(frame, "/home/lvuser/SampleImages/image20.jpg");
//Update threshold values from SmartDashboard. For performance reasons it is recommended to remove this after calibration is finished.
TOTE_HUE_RANGE.minValue = (int)SmartDashboard.getNumber("Tote hue min", TOTE_HUE_RANGE.minValue);
TOTE_HUE_RANGE.maxValue = (int)SmartDashboard.getNumber("Tote hue max", TOTE_HUE_RANGE.maxValue);
TOTE_SAT_RANGE.minValue = (int)SmartDashboard.getNumber("Tote sat min", TOTE_SAT_RANGE.minValue);
TOTE_SAT_RANGE.maxValue = (int)SmartDashboard.getNumber("Tote sat max", TOTE_SAT_RANGE.maxValue);
TOTE_VAL_RANGE.minValue = (int)SmartDashboard.getNumber("Tote val min", TOTE_VAL_RANGE.minValue);
TOTE_VAL_RANGE.maxValue = (int)SmartDashboard.getNumber("Tote val max", TOTE_VAL_RANGE.maxValue);
//Threshold the image looking for yellow (tote color)
NIVision.imaqColorThreshold(binaryFrame, frame, 255, NIVision.ColorMode.HSV, TOTE_HUE_RANGE, TOTE_SAT_RANGE, TOTE_VAL_RANGE);
//Send particle count to dashboard
int numParticles = NIVision.imaqCountParticles(binaryFrame, 1);
SmartDashboard.putNumber("Masked particles", numParticles);
//Send masked image to dashboard to assist in tweaking mask.
CameraServer.getInstance().setImage(binaryFrame);
//filter out small particles
float areaMin = (float)SmartDashboard.getNumber("Area min %", AREA_MINIMUM);
criteria[0].lower = areaMin;
imaqError = NIVision.imaqParticleFilter4(binaryFrame, binaryFrame, criteria, filterOptions, null);
//Send particle count after filtering to dashboard
numParticles = NIVision.imaqCountParticles(binaryFrame, 1);
SmartDashboard.putNumber("Filtered particles", numParticles);
if(numParticles > 0)
{
//Measure particles and sort by particle size
Vector<ParticleReport> particles = new Vector<ParticleReport>();
for(int particleIndex = 0; particleIndex < numParticles; particleIndex++)
{
ParticleReport par = new ParticleReport();
par.PercentAreaToImageArea = NIVision.imaqMeasureParticle(binaryFrame, particleIndex, 0, NIVision.MeasurementType.MT_AREA_BY_IMAGE_AREA);
par.Area = NIVision.imaqMeasureParticle(binaryFrame, particleIndex, 0, NIVision.MeasurementType.MT_AREA);
par.ConvexHullArea = NIVision.imaqMeasureParticle(binaryFrame, particleIndex, 0, NIVision.MeasurementType.MT_CONVEX_HULL_AREA);
par.BoundingRectTop = NIVision.imaqMeasureParticle(binaryFrame, particleIndex, 0, NIVision.MeasurementType.MT_BOUNDING_RECT_TOP);
par.BoundingRectLeft = NIVision.imaqMeasureParticle(binaryFrame, particleIndex, 0, NIVision.MeasurementType.MT_BOUNDING_RECT_LEFT);
par.BoundingRectBottom = NIVision.imaqMeasureParticle(binaryFrame, particleIndex, 0, NIVision.MeasurementType.MT_BOUNDING_RECT_BOTTOM);
par.BoundingRectRight = NIVision.imaqMeasureParticle(binaryFrame, particleIndex, 0, NIVision.MeasurementType.MT_BOUNDING_RECT_RIGHT);
particles.add(par);
}
particles.sort(null);
//This example only scores the largest particle. Extending to score all particles and choosing the desired one is left as an exercise
//for the reader. Note that the long and short side scores expect a single tote and will not work for a stack of 2 or more totes.
//Modification of the code to accommodate 2 or more stacked totes is left as an exercise for the reader.
scores.Trapezoid = TrapezoidScore(particles.elementAt(0));
SmartDashboard.putNumber("Trapezoid", scores.Trapezoid);
scores.LongAspect = LongSideScore(particles.elementAt(0));
SmartDashboard.putNumber("Long Aspect", scores.LongAspect);
scores.ShortAspect = ShortSideScore(particles.elementAt(0));
SmartDashboard.putNumber("Short Aspect", scores.ShortAspect);
scores.AreaToConvexHullArea = ConvexHullAreaScore(particles.elementAt(0));
SmartDashboard.putNumber("Convex Hull Area", scores.AreaToConvexHullArea);
boolean isTote = scores.Trapezoid > SCORE_MIN && (scores.LongAspect > SCORE_MIN || scores.ShortAspect > SCORE_MIN) && scores.AreaToConvexHullArea > SCORE_MIN;
boolean isLong = scores.LongAspect > scores.ShortAspect;
//Send distance and tote status to dashboard. The bounding rect, particularly the horizontal center (left - right) may be useful for rotating/driving towards a tote
SmartDashboard.putBoolean("IsTote", isTote);
SmartDashboard.putNumber("Distance", computeDistance(binaryFrame, particles.elementAt(0), isLong));
} else {
SmartDashboard.putBoolean("IsTote", false);
}
Timer.delay(0.005); // wait for a motor update time
}
}
public void operatorControl() {
while(isOperatorControl() && isEnabled()) {
Timer.delay(0.005); // wait for a motor update time
}
}
//Comparator function for sorting particles. Returns true if particle 1 is larger
static boolean CompareParticleSizes(ParticleReport particle1, ParticleReport particle2)
{
//we want descending sort order
return particle1.PercentAreaToImageArea > particle2.PercentAreaToImageArea;
}
/**
* Converts a ratio with ideal value of 1 to a score. The resulting function is piecewise
* linear going from (0,0) to (1,100) to (2,0) and is 0 for all inputs outside the range 0-2
*/
double ratioToScore(double ratio)
{
return (Math.max(0, Math.min(100*(1-Math.abs(1-ratio)), 100)));
}
/**
* Method to score convex hull area. This scores how "complete" the particle is. Particles with large holes will score worse than a filled in shape
*/
double ConvexHullAreaScore(ParticleReport report)
{
return ratioToScore((report.Area/report.ConvexHullArea)*1.18);
}
/**
* Method to score if the particle appears to be a trapezoid. Compares the convex hull (filled in) area to the area of the bounding box.
* The expectation is that the convex hull area is about 95.4% of the bounding box area for an ideal tote.
*/
double TrapezoidScore(ParticleReport report)
{
return ratioToScore(report.ConvexHullArea/((report.BoundingRectRight-report.BoundingRectLeft)*(report.BoundingRectBottom-report.BoundingRectTop)*.954));
}
/**
* Method to score if the aspect ratio of the particle appears to match the long side of a tote.
*/
double LongSideScore(ParticleReport report)
{
return ratioToScore(((report.BoundingRectRight-report.BoundingRectLeft)/(report.BoundingRectBottom-report.BoundingRectTop))/LONG_RATIO);
}
/**
* Method to score if the aspect ratio of the particle appears to match the short side of a tote.
*/
double ShortSideScore(ParticleReport report){
return ratioToScore(((report.BoundingRectRight-report.BoundingRectLeft)/(report.BoundingRectBottom-report.BoundingRectTop))/SHORT_RATIO);
}
/**
* Computes the estimated distance to a target using the width of the particle in the image. For more information and graphics
* showing the math behind this approach see the Vision Processing section of the ScreenStepsLive documentation.
*
* @param image The image to use for measuring the particle estimated rectangle
* @param report The Particle Analysis Report for the particle
* @param isLong Boolean indicating if the target is believed to be the long side of a tote
* @return The estimated distance to the target in feet.
*/
double computeDistance (Image image, ParticleReport report, boolean isLong) {
double normalizedWidth, targetWidth;
NIVision.GetImageSizeResult size;
size = NIVision.imaqGetImageSize(image);
normalizedWidth = 2*(report.BoundingRectRight - report.BoundingRectLeft)/size.width;
targetWidth = isLong ? 26.0 : 16.9;
return targetWidth/(normalizedWidth*12*Math.tan(VIEW_ANGLE*Math.PI/(180*2)));
}
}