You are here: Start » Function Reference » Computer Vision » Image Analysis » MeasureObjectWidth

MeasureObjectWidth

Header:	FIL.h
Namespace:	fil
Module:	MetrologyBasic

Measures the width of an object using stripe detection.

Applications: Easy and precise measurement of distances between to straight parallel edges.

Syntax

void fil::MeasureObjectWidth
(
	const fil::Image& inImage,
	const ftl::Array<fil::ScanMap>& inMeasurementMap,
	const fil::StripeScanParams& inStripeScanParams,
	fil::MeasureObjectMethod::Type inMeasureMethod,
	fil::Selection::Type inStripeSelection,
	ftl::Optional<const fil::LocalBlindness&> inLocalBlindness,
	ftl::Optional<fil::MEstimator::Type> inOutlierSuppression,
	int inOutlierCount,
	ftl::Conditional<float>& outObjectWidth,
	ftl::Conditional<fil::Segment2D>& outSegment1,
	ftl::Conditional<fil::Segment2D>& outSegment2,
	ftl::Optional<ftl::Array<ftl::Conditional<fil::Stripe1D>>&> outStripes = ftl::NIL,
	ftl::Optional<ftl::Array<fil::Profile>&> outBrightnessProfiles = ftl::NIL,
	ftl::Optional<ftl::Array<fil::Profile>&> outResponseProfiles = ftl::NIL
)

Parameters

Name	Type	Default	Description
inImage	const Image&		Input image
inMeasurementMap	const Array<ScanMap>&		Input measurement map
inStripeScanParams	const StripeScanParams&		Parameters controlling the object stripe extraction process
inMeasureMethod	MeasureObjectMethod::Type		Method used to measure the object
inStripeSelection	Selection::Type	Selection::Best	Selection mode of edges of the object
inLocalBlindness	Optional<const LocalBlindness&>	NIL	Defines conditions in which weaker edges can be detected in the vicinity of stronger edges
inOutlierSuppression	Optional<MEstimator::Type>	NIL	Selects a method for ignoring incorrectly detected points
inOutlierCount	int		Determines how many outlying points are rejected before the width is measured
outObjectWidth	Conditional<float>&		Width of the object
outSegment1	Conditional<Segment2D>&		First edge of the object
outSegment2	Conditional<Segment2D>&		Second edge of the object
outStripes	Optional<Array<Conditional<Stripe1D>>&>	NIL	Detected stripes
outBrightnessProfiles	Optional<Array<Profile>&>	NIL	Extracted image profiles
outResponseProfiles	Optional<Array<Profile>&>	NIL	Profiles of the edge (derivative) operator response

Optional Outputs

The computation of following outputs can be switched off by passing value ftl::NIL to these parameters: outStripes, outBrightnessProfiles, outResponseProfiles.

Description

The filter measures the width of an object present in an image. Internally, it performs a series of scans with ScanSingleStripe filter using inMeasurementMap and inStripeScanParams. The so obtained points are then used for computing two parallel segments by means of a slightly modified segment fitting routine. The process is supported by inOutlierSuppression parameter. Finally, having the aforementioned stripe widths and fitted segments' direction, the object width can be computed using a selected inMeasureMethod method. The inOutlierCount stripe widths most differing from the median width are not used in this step. The inMeasurementMap should be created with CreateMeasurementMap function. For the filter to work properly, the scan segments do not have to be necessarily perpendicular to the object edges.

Hints

Define inStripeScanParams.StripePolarity to detect a particular edge type, and only that type.
If some points are not found, try decreasing inStripeScanParams.MinMagnitude.
If the object is more narrow than 6 pixels, change inStripeScanParams.ProfileInterpolation to Quadratic3.
Experiment with inOutlierSuppression and inOutlierCount to deal with some amount of incorrectly detected points.
Experiment with various values for inMeasureMethod to obtain best possible precision.

Examples

MeasureObjectWidth performed on the sample image. Green segments are diagScanSegments, the other two are the found object edges.

Errors

List of possible exceptions:

Error type	Description
DomainError	Invalid MeasurementMap in MeasureObjectWidth function. Use CreateMeasurementMap function to create it properly.