/// <summary>
/// Detects light sources on the input image using hysteresis thresholding. Requires input image to be astrometrically reduced.
/// </summary>
/// <param name="Input">Input image.</param>
/// <returns>A list of detections.</returns>
public List <ImageDetection> Detect(Image Input)
{
Detections = new List <DotDetection>();
PositionDependentExtractor <DotDetector> Extractor = DetectSources;
Extractor.Run(this, Input, Parameters);
List <ImageDetection> Mdect = Detections.Select((x) => StandardDetectionFactory.CreateDetection(Input, x.Pixels, x.PixelValues)).ToList();
foreach (ImageDetection m in Mdect)
{
m.SetResetProperty(new PairingProperties()
{
IsDotDetection = true
});
}
return(Mdect);
}
/// <summary>
/// Attempts to recover a detection on a given image, comparing with the entire set of exposures.
/// </summary>
/// <returns><c>true</c>, if detection was recovered, <c>false</c> otherwise.</returns>
/// <param name="DetPos">Position of the detection to recover.</param>
/// <param name="Img">Image on which to recover.</param>
/// <param name="Radius">Maximum radius of the detection.</param>
/// <param name="InputImages">Input images.</param>
/// <param name="Recovered">Recovered detection.</param>
public bool RecoverDetection(Position DetPos, Image Img, double Radius, IEnumerable <Image> InputImages, out ImageDetection Recovered)
{
Recovered = null;
DotDetector.DotDetection dd = Recover(DetPos.PP, Radius, Img);
if (dd.Pixels.Count < MinPix)
{
return(false);
}
int NoiseCnt = 0;
int FixedCnt = 0;
foreach (Image img in InputImages)
{
/* Noise scanner -- same position in pixel coordinates */
DotDetector.DotDetection detN = Recover(dd.Barycenter, Radius, img);
PixelPoint npp = img.Transform.GetPixelPoint(Img.Transform.GetEquatorialPoint(dd.Barycenter));
/* Star scanner -- same position in equatorial coordinates */
DotDetector.DotDetection detF = Recover(npp, Radius, img);
if (detN.Pixels.Count > NoisePixelThreshold * dd.Pixels.Count)
{
NoiseCnt++;
}
if (detF.Flux > StarFluxThreshold * dd.Flux)
{
EquatorialPoint org = DetPos.EP;
EquatorialPoint nw = img.Transform.GetEquatorialPoint(detF.Barycenter);
if ((org ^ nw) < MinMoveArcSec)
{
FixedCnt++;
}
}
}
if (NoiseCnt > CrossMatchRemove)
{
return(false);
}
if (FixedCnt > CrossMatchRemove)
{
return(false);
}
Recovered = StandardDetectionFactory.CreateDetection(Img, dd.Pixels, dd.PixelValues);
return(true);
}
/// <summary>
/// The segment detector function. It calls the RLHT scorer and if line segments are sensed, it calls the LineAnalyzer to find the source blobs.
/// </summary>
/// <param name="Input">Input data.</param>
/// <param name="Position">Position of the input data array in the image.</param>
/// <param name="Data">Bag of algorithm parameters and data.</param>
static void LTD_RLHT(double[,] Input, SchedCore.ImageSegmentPosition Position, LongTrailData Data)
{
/* Extracts the size of the input data */
int Height = Input.GetLength(0), Width = Input.GetLength(1);
double Diagonal = Math.Sqrt(Width * Width + Height * Height);
/* Initialize VPool */
lock (Data.AgData.VPool)
if (Data.AgData.VPool.Constructor == null)
{
Data.AgData.VPool.Constructor = () => new List <Vector>();
}
/* Applies the RLHT algorithm */
Data.AgData.StrongValueFunction = (x) => ThresholdComputer(x, Data, Diagonal);
var Result = RLHT.SmartSkipRLHT(Input, Data.ImageParameters, Data.AgData);
/* Prepare common data for the LineAnalyzer */
bool[,] Mask = new bool[Height, Width];
double SST = Data.SegmentSelectThreshold * Data.Sigma, SDT = Data.SegmentDropThreshold * Data.Sigma;
int MIB = Data.MaxInterblobDistance, SW = Data.ScanWidth, pX = (int)Position.Alignment.X, pY = (int)Position.Alignment.Y;
if (Data.DropCrowdedRegion) /* If the region is too crowded, it's very likely to be some luminous residue - for example star halos */
{
if (Result.StrongPoints.Count > Diagonal) /* There is no deep meaning between this comparison; a reasonable Diagonal seems to correspond to a reasonable number of lines */
{
goto clear_end;
}
}
/* Analyze each possible trail line and store the detections */
foreach (Vector vx in Result.StrongPoints)
{
var z = LineAnalyzer.AnalyzeLine(Input, Mask, Height, Width, vx.X, vx.Y, SST, SDT, MIB, SW, pX, pY);
lock (Data.Results)
Data.Results.AddRange(z.Select((x) => StandardDetectionFactory.CreateDetection(Data.RunningImage, x.Points, x.PointValues)));
}
clear_end:
/* Release resources */
Result.StrongPoints.Clear();
Data.AgData.HTPool.Release();
Data.AgData.VPool.Release();
}
/// <summary>
/// Match detections and merge those that seem to belong to the same object.
/// </summary>
/// <param name="RawDetections">Input set of detections.</param>
/// <param name="MaxDistance">Maximum distance possible between two detections part of the same object.</param>
/// <param name="MixMatch">Number of overlapping pixels before two detections are considered part of the same object.</param>
/// <param name="PSFMatch">Distance between the barycenters of 2 detections before they are considered the same (for external detections mostly).</param>
public static void MatchDetections(List <ImageDetection> RawDetections, double MaxDistance, int MixMatch, double PSFMatch)
{
int i, j;
List <HashSet <PixelPoint> > LHP = RawDetections.Select((x) => x.TryFetchProperty(out ObjectPoints op) ? new HashSet <PixelPoint>(op.PixelPoints) : null).ToList();
List <PairingProperties> PairPropList = RawDetections.Select((x) => x.TryFetchProperty(out PairingProperties Prop) ? Prop : null).ToList();
for (i = 0; i < RawDetections.Count; i++)
{
for (j = i + 1; j < RawDetections.Count; j++)
{
/* Must be two detections captured at the same time */
if (RawDetections[i].Time.Time != RawDetections[j].Time.Time)
{
continue;
}
/* Check distance */
double D0 = (RawDetections[i].Barycenter.PP ^ RawDetections[j].Barycenter.PP);
double D1 = (RawDetections[i].FetchProperty <ObjectSize>().PixelEllipse.SemiaxisMajor + RawDetections[j].FetchProperty <ObjectSize>().PixelEllipse.SemiaxisMajor);
if (D0 - D1 > MaxDistance)
{
continue;
}
HashSet <PixelPoint> PixPi = LHP[i], PixPj = LHP[j];
bool FlagAnyCond = false;
/* If there are MinPix overlapping pixels, merge detections */
if (PixPi != null & PixPj != null)
{
if (PixPi.Overlaps(PixPj))
{
FlagAnyCond = true;
}
if (!FlagAnyCond)
{
/* Detections that are somewhat linear are checked for colinearity with others */
IEnumerable <PixelPoint> Plist = PixPi.Concat(PixPj);
LinearRegression.LinearRegressionParameters pc = LinearRegression.ComputeLinearRegression(Plist);
LinearRegression.LinearRegressionParameters p1 = LinearRegression.ComputeLinearRegression(PixPi);
LinearRegression.LinearRegressionParameters p2 = LinearRegression.ComputeLinearRegression(PixPj);
if (Math.Abs(pc.PearsonR) > Math.Abs(p1.PearsonR) && Math.Abs(pc.PearsonR) > Math.Abs(p2.PearsonR) &&
Math.Abs(pc.PearsonR) < Math.Abs(p1.PearsonR) + Math.Abs(p2.PearsonR))
{
FlagAnyCond = true;
}
}
}
if (D0 < PSFMatch)
{
FlagAnyCond = true;
}
/* If any merging condition is satisfied, merge the detections */
if (FlagAnyCond)
{
if (LHP[i] != null & LHP[j] != null)
{
LHP[i].UnionWith(LHP[j]);
}
else if (LHP[i] == null)
{
LHP[i] = LHP[j];
}
if (PairPropList[i] != null && PairPropList[j] != null)
{
PairPropList[i].Algorithm |= PairPropList[i].Algorithm;
}
LHP.RemoveAt(j);
RawDetections.RemoveAt(j);
PairPropList.RemoveAt(j);
j--;
}
}
}
for (i = 0; i < LHP.Count; i++)
{
if (LHP[i] != null)
{
try
{
RawDetections[i] = StandardDetectionFactory.CreateDetection(RawDetections[i].ParentImage, LHP[i]);
if (PairPropList[i] != null)
{
RawDetections[i].SetResetProperty(PairPropList[i]);
}
}
catch { RawDetections.RemoveAt(i); PairPropList.RemoveAt(i); LHP.RemoveAt(i); i--; }
}
}
}