Boochs Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images Frank Boochs i3mainz, Institute for Spatial Information and Surveying Technology Holzstr. 36, D-55116 Mainz, Germany [email protected]
Jan 01, 2016
Boochs
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
Frank Boochs
i3mainz, Institute for Spatial Information and Surveying Technology
Holzstr. 36, D-55116 Mainz, Germany
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
2Boochs
content
• motivation• conception• aspects of solution• data processing• practical aspects• examples
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
3Boochs
motivation
area of application:image based surveying of spatial objects
needs: determination of orientation values for all images to be used
is based on: image rays to tie and control points
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
4Boochs
motivation
area of application:image based surveying of spatial objects
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
5Boochs
collection of image rays by means of:
computer based detection of artificial signals
manual identification and measurement of natural object points
motivation
signals are:• templates of a certain diametre• used to define a point (centre)• not used with their spatial data (diametre, area)
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
6Boochs
reduction of success due to:
• image noise
motivation
• perspective distortion
left view right view
in particular for arbitrary arranged images:
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
7Boochs
conventional ways of compensation:
• image noise:improvement of S/N ratio (sensor, light,...)
motivation
• distortion:limitation of angular differences– for optical axes of different images– for surface normal and individual optical
axis
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
8Boochs
• quality of image rays is degraded• reduced connection between images• failure to connect two images
compensation leads to increasing effort:
motivation
in cases of reduced success consequences have to be expected:
• increasing number of images• increasing number of signals
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
9Boochs
idea: improvement of connectivity be means of spatial tie objects
conception
sphere:• substitutes target plane• visible from all directions• defines one image ray
body with n faces:• defines multiple target planes• visible from all directions• defines several image rays• introduces additional information (scale, local co-ordinate system, orientation in space)
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
10Boochs
we use of spatial tie objects with several faces
advantages :
• multiple surfaces useable• at least one surface with a
small angle to the optical axis
• allows completely new arrangements
conception
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
11Boochs
essential requirement which has to be fulfilled:
the targets have to be treated and detected as complete spatial objects
• the whole body (= all visible surfaces) has to be detected
• visible surfaces have to be uniquely identified
• all surfaces need to have a fix and known topology
conception
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
12Boochs
type of target chosen: cube
because of its:
symmetry
equal size of faces
aspects of solution
number of faces
angular difference of adjacent faces
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
13Boochs
discrimination of individual surfaces by means of :
coloring
because:
information is pixel based
aspects of solution
sufficient number of separable colors available (red, green, blue, cyan, magenta, yellow)
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
14Boochs
discrimination of individual surfaces by means of :
alternative: active ligthing
because:
independent of ambient light
aspects of solution
easier analysis of colour values
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
more robust
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
15Boochs
further benefit:
simple geometry
simplifies modelling within different stages of analysis
aspects of solution
allows to introduce effective constraints
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
16Boochs
general algorithmic structure:
data processing
correspondence analysis
preliminary orientation values, object co-ordinates
image analysis
target data (co-ordinates, attri-butes, codes,..)
image
image
image
image
triangulation adjusted orientation values, object co-ordinates
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
17Boochs
image analysis steps: data processing
image (part) vectorised edges
• edge detection• vectorisation
attributed edges
• evaluation of color attributes
• verification of hypothesis
• calculation of geometry
final cube
• generation of cube hypothesis
inner cube edges outer edge candidates
cube candidate
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
18Boochs
image analysis steps (active lighting): data processing
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
original saturation brigthness saturation & brightness
segmentation
contour extraction
contour filtering
noise reduction
• edge detection• calculation of
geometry
final cube
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
19Boochs
further steps: data processing
correspondence analysis
simplified by:
• grouping of points• geometry constraints• target knowledge
image triangulation
based on standard algorithms:
actually : CAP
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
20Boochs
practical aspects
color information
• simplifies detection· provides binary code· characterizes shape / relation of sides of targets
• for passive coding: • may vary considerably (direction of illumination, type of light, sensor characteristic and state)
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
• for active coding: • luminosity and homogeneity need to be controlled (overexposure, reaction of sensor)
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
21Boochs
practical aspects
size of targets
• has to be considered with respect to possibly strong variations within image scale
• has to be sufficient
·size of faces
·geometric quality of edges
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
22Boochs
practical aspects
spatial characteristic of targets• provides useful information within
analysis·parallel edges·right angles at adjacent edges·constant edge length
• introduces scale into image block
· each cube allows to define 2 groups of 2 distances in each direction of space
• further distances only for quality check
· distances are scattered over object space together with targets
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
23Boochs
practical aspects
• each target may define a local co-ordinate system-> allows orientation to a local system
without any further preparations within object space
-> allows to estimate orientation values for images in a common system
• already one target provides data for preliminary orientation
benefits from spatial shape of targets (spatial extent)
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
24Boochs
practical aspects
• rays of each target are covering a local image area
-> provides control information for local image geometry
benefits from spatial shape of targets (multitude of image rays)
-> gives redundancy and allows to control quality of each point
• each target is represented by 8 rays
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
25Boochs
practical aspects
benefits from independence of viewing direction
• provides deepest connection of images possible-> all images will be connected having sight onto a
certain target
-> constraints coming from sight onto tie points are reduced
• allows to optimize image arrangements simply from viewpoint of best geometry
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
26Boochs
practical aspects
benefits from independence of viewing direction
-> images with opposite viewing directions might be combined
• allows complete new image arrangements
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
27Boochs
practical aspects
fields of application
• industrial measurements
• forensic tasks
• facility mapping
• architecture
• stereoscopic visualisation (->consumer market)
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
28Boochs
example
test: pair of images with 900 convergency
4 common targets used
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images
27.5.04 Journée TSI "Eclairage structuré pour la vision active"
29Boochs
example
test: pair of images with opposite viewing directions
orientation performed with minimal data (1 target)
Detection of spatial reference objects for automatic orientation of arbitrary arranged metric images