Image Processing and Pattern Recognition Jouko Lampinen.

Image Processing and Pattern Recognition

Jouko Lampinen

About this presentation

• In this set of slides we illustrate a bigger problem which uses both morphological operations and other operations that will be introduced soon.

• In most cases we use morphological operations together with other operations.

• The most important reason of using them is speed and non-linear processing.

Image analysis of grain material in concrete production

• Images captured by standard

1200 dpi color scanner • Grain shape inputs

• angularity, flakiness• Grain texture inputs

• Boundary & surface texture• FFT based texture features

• Image Analysis Tool:

Matlab standalone application• Quality Control Tool:

Excel macro package for running

and analyzing the Bayesian

models (to be discussed)

Example of grains (1.6-2.0 mm sieve fraction)

Grain FeaturesGrain Features Measured from the Image

• Area• Major Axis • Minor Axis• Eccentricity• Convex Area• Equivalent Diameter• Solidity• Perimeter• Compactness• Borderline FFT (5 features related to roughness)• Texture 2D FFT (5 features related to surface structure) • Morphological Spectrum (roundness)

Most of these parameters will be presented in next lectures

Object size and shape characterization

• Bounding box (rotated along principal axes)• Ellipsoid determined by the principal axes • Convex hull

Original sand grain image (natural sand)

Thresholded image (natural grains)

Objects filled

Morphological opening (yellow pixels removed)

Labelled objects

Bounding boxes and minor/major axes

Original sand grain image (crushed)

Thresholded image (crushed)

Objects filled

Morphological opening (yellow pixels removed)

Labelled objects

Bounding boxes and minor/major axes

Grain shape analysis: angularity

Sharp angles in grains break under compression

Measurement: simulate the erosion due to Ice Age by morphological erosion

Morphological spectrum:

S(r)

Amount of material removed by circular structure element of radius r

Feature space!!

Example of Morphological Spectra and Angularity

Crushed gravel Natural gravel (manufactured by Ice Age)

Morphological spectrumWe can scientifically

compare various gravels