You are here: Start » Function Reference » Computer Vision » Line Finder » FindMultipleLines

FindMultipleLines

Header:	AVL.h
Namespace:	avl
Module:	MetrologyPro

Finds multiple straight lines within a rectangular scan field, ordered from highest-scoring to lowest.

Applications: Detection of multiple lines within a single scan field.

Syntax

void avl::FindMultipleLines
(
	const avl::Image& inImage,
	const avl::SegmentScanField& inScanField,
	atl::Optional<const avl::CoordinateSystem2D&> inScanFieldAlignment,
	int inScanWidth,
	int inScanStep,
	int inScanLimit,
	const avl::EdgeScanParams& inEdgeScanParams,
	float inPointBlindness,
	float inMaxSlant,
	float inMinLength,
	float inMaxGap,
	int inMinPointCount,
	avl::LineSelection::Type inLineSelection,
	float inLineBlindness,
	avl::EdgeMeasure::Type inEdgeMeasure,
	avl::LineFinderMethod::Type inLineFinderMethod,
	float inMaxPointDeviation,
	float inMaxTurnAngle,
	atl::Optional<avl::LineMEstimator::Type> inOutlierSuppression,
	bool inRefineEndpoints,
	atl::Optional<int> inMaxCount,
	atl::Array<avl::Segment2D>& outLines,
	atl::Array<float>& outLineStrengths,
	atl::Array<avl::Profile>& outResponseProfiles,
	atl::Optional<avl::SegmentScanField&> outAlignedScanField,
	atl::Array<atl::Array<avl::Point2D>>& outSelectedPoints,
	atl::Array<avl::Segment2D>& diagScanLines,
	atl::Array<avl::Point2D>& diagCandidatePoints
)

Parameters

Name	Type	Range	Default	Description
inImage	const Image&			Input image
inScanField	const SegmentScanField&			Region where the scans are performed
inScanFieldAlignment	Optional<const CoordinateSystem2D&>		NIL
inScanWidth	int	1 -	5	How many pixels are sampled perpendicularly at each scan line's positions
inScanStep	int	1 -	10	Pixel distance between the centers of consecutive scan lines
inScanLimit	int	1 -	50	Maximum number of scan lines
inEdgeScanParams	const EdgeScanParams&		EdgeScanParams ( ProfileInterpolation: Quadratic4 SmoothingStdDev: 1.2f MinMagnitude: 5.0f EdgeTransition: Any )	Parameters of 1D edge detection along a single scan line
inPointBlindness	float	0.0 - 1.0	0.3f	Defines a threshold for rejecting edge points that are weaker than a certain portion of the strongest edge point
inMaxSlant	float	0.0 - 45.0	30.0f	Maximum angle deviation of the resulting line from 90 degrees to the scan axis
inMinLength	float	0.0 -	10.0f	Minimum length of the resulting line
inMaxGap	float	0.0 -	50.0f	Maximum gap within a resulting line
inMinPointCount	int	2 -	3	Minimum number of edge points that a resulting line must contain (otherwise it is rejected)
inLineSelection	LineSelection::Type		Fittest	Selection criteria for edge lines
inLineBlindness	float	0.0 - 1.0	0.4f	Defines a threshold for rejecting entire lines that are weaker than a certain portion of the strongest line
inEdgeMeasure	EdgeMeasure::Type		LogMagnitude	Several options for measuring a single edge point
inLineFinderMethod	LineFinderMethod::Type		Hough	Several alternative algorithms for finding lines within the detected cloud of edge points
inMaxPointDeviation	float	0.0 -	2.5f	Maximum pixel distance between a detected line and its supporting edge points
inMaxTurnAngle	float	0.0 - 90.0	10.0f	Maximum turn angle at any interior vertex of the line path
inOutlierSuppression	Optional<LineMEstimator::Type>		TheilSen	A method for removing outlying points
inRefineEndpoints	bool		True	Turns on a method for refining the length of the detected line
inMaxCount	Optional<int>	1 -	100	Maximum number of lines to return
outLines	Array<Segment2D>&
outLineStrengths	Array<float>&
outResponseProfiles	Array<Profile>&			Edge response profiles, one per scan line
outAlignedScanField	Optional<SegmentScanField&>
outSelectedPoints	Array<Array<Point2D>>&			Lists of points associated with each returned line
diagScanLines	Array<Segment2D>&			Positions of the scan lines placed within the scan field
diagCandidatePoints	Array<Point2D>&			Edge points detected on the scan lines, after inPointBlindness filtering

Optional Outputs

The computation of following outputs can be switched off by passing value atl::NIL to these parameters: outAlignedScanField.

Description

The operation finds all straight lines within the inScanField and returns them ordered from the highest-scoring to the lowest. It works like FindSingleLine, but iterates the selection: once a line is accepted, its supporting candidate points are excluded from the competition for subsequent lines, so the results are non-overlapping. Up to inMaxCount lines are returned.

Hints

Activate diagCandidatePoints and diagScanLines during development to verify that candidates appear on the right scan lines before adjusting higher-level parameters.
Adjust inEdgeScanParams.MinMagnitude first: a value that is too high misses real edges; too low fills the result with noise candidates.
Increase inScanStep to speed up execution on large scan fields. Keep it small (5–10 px) when the line is short or its position varies rapidly across the field.
inPointBlindness suppresses weak per-scan-line candidates relative to the strongest response on that line. A value of 0.3 discards candidates below 30 % of the highest value on that scan line.
If a detected segment is shorter than the actual line, try increasing inMaxGap, widening inMaxPointDeviation.
Set inMaxSlant to the maximum expected deviation of the line from the direction perpendicular to the scan field axis. Narrowing this range reduces false positives.
Use inLineSelection = Fittest for the straightest line, Strongest for the highest-contrast line, and Smoothest for gently curved or slightly wavy lines.
Connect inScanFieldAlignment to a coordinate system to automatically adapt the measurement to the object position in each frame.
Set inMaxCount to the exact number of parallel lines expected to prevent spurious detections and unnecessary computation.
Increase inLineBlindness to reject weak secondary lines; decrease it to include fainter candidates.

Examples

FindMultipleLines locating two intersecting lines. The dashed rectangle is the scan field, the blue and yellow dots are candidates selected for two lines, and the green lines are detected lines.

Remarks

Read more about Local Coordinate Systems in Machine Vision Guide: Local Coordinate Systems.

Hardware Acceleration

This operation supports automatic parallelization for multicore and multiprocessor systems.