forked from eo4geocourses/PLUS_OBIA-Introduction
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathindex.html
More file actions
1080 lines (961 loc) · 53.9 KB
/
index.html
File metadata and controls
1080 lines (961 loc) · 53.9 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
<!-- Template Authors: Stefan Lang, Eva Missoni - PLUS
Template Creation Date: March-June 2020
IMPORTANT NOTICES FOR USERS/AUTHORS:
SECTIONS TO BE EDITED BY AUTHORS ARE INDICATED AS FOLLOWS: <!-- AUTHORS: DEFINE ... -->
<!-- THE COMMANDS <section> and </section> EMBRACE THE CONTENT FOR EACH SINGLE SLIDE.
ALTERNATIVELY, YOU CAN USE MARKDOWN NOTATION TO CREATE YOUR SLIDES. More information: https://github.com/webpro/reveal-md
-->
<!doctype html>
<html prefix="eo4geo: http://bok.eo4geo.eu/" prefix="dc: http://purl.org/dc/terms/" lang="en" >
<head>
<meta charset="utf-8">
<title>OBIA - Intro</title>
<meta property="dc:title" content= "Object-based Image Analysis - An Introduction" />
<meta property="dc:creator" content="Stefan Lang, University of Salzburg" />
<meta property="dc:publisher" content="University of Salzburg" />
<meta property="dc:subject" content= "OBIA, image segmentation, EO4GEO, Earth observation, spatial image analysis" />
<meta property="dc:abstract" content= "These slides give an introduction to concepts and methods of object-based image analysis (OBIA). The slides have been prepared for and used in the virtual GEOBIA summer school in 2020." />
<meta property="dc:tableOfContents" content= "(1) Why spatial image analysis?; (2) Regions and image objects; (3) Image segmentation;
(4) Knowledge representation; (5) Class modelling" />
<meta property="dc:description" content= "Learning outcomes: Understand the principles of spatial image analysis; discuss specific advantages of OBIA over pixel-based approach; apply OBIA concepts and analyse complex image content; examine different segmentation and object-based classification methods" />
<meta property="dc:contributor" content= "Dirk Tiede, Eva Missoni, Barbara Hofer, Simon Donike" />
<meta property="dc:created" content="2020-08-27" />
<meta property="dc:type" content= "teaching material" />
<meta property="dc:format" content= "html" />
<meta property="dc:language" content= "EN" />
<meta property="dc:SizeOrDuration" content= "45min" />
<meta property="dc:audience" content= "students" />
<meta property="dc:educationLevel" content= "EQF 6" />
<meta property="dc:source" content= "GEOBIA Summer School Material" />
<meta property="dc:rightsHolder" content= "Stefan Lang, University of Salzburg" />
<meta property="dc:license" content= "https://creativecommons.org/licenses/by-sa/4.0/deed.en" />
<link rel="dc:relation" href="eo4geo:IP3-7" />
<link rel="dc:relation" href="eo4geo:IP3" />
<link rel="dc:relation" href="eo4geo:IP3-5" />
<link rel="dc:relation" href="eo4geo:IP3-7-1" />
<link rel="dc:relation" href="eo4geo:IP3-7-2" />
<link rel="dc:relation" href="eo4geo:IP3-7-3" />
<link rel="dc:relation" href="eo4geo:IP3-7-4" />
<link rel="dc:relation" href="eo4geo:IP3-7-4-1" />
<link rel="dc:relation" href="eo4geo:IP3-7-4-2" />
<link rel="dc:relation" href="eo4geo:IP3-7-4-3" />
<link rel="stylesheet" href="internal_restricted/v1.0/css/reveal.css">
<!-- CHOOSE BETWEEN 2 EO4GEO TEMPLATE THEMES:-->
<link rel="stylesheet" href="internal_restricted/v1.0/css/theme/eo4geo.css" id="theme"> <!-- white background design -->
<link rel="stylesheet" href="internal_restricted/v1.0/plugin/markdown/css/theme/simplemenu.css">
<!--link rel="stylesheet" href="internal_restricted/v1.0/css/theme/eo4geo_grey.css" id="theme"--> <!-- anthrazit background design -->
<!-- EO4GEO FONT ROBOTO SLAB -->
<link href='https://fonts.googleapis.com/css?family=Roboto Slab' rel='stylesheet'>
<link rel="stylesheet" href="partnerLogo.css" id="theme">
<link rel="stylesheet" href="customer_content/css/eo4geo_override.css" id="theme">
<base target="_blank"> <!-- All links in the presentation are opened in a seperate tab"-->
</head>
<body>
<div class="reveal" >
<!-- FORMATTING FOR HEADER -->
<div class = "eo4geologo" style="width: 11%;height: auto" ></div>
<div class = "partnerlogo" style="width: 15%;height: auto" ></div>
<div class = "presentation_title" style=" text-align: center " ><small>Object-based Image Analysis (OBIA) - An Introduction</div></small> <!--The title of the presentation appears on each slide-->
<div class="slides">
<!-- AUTHORS: DEFINE CONTENT ON YOUR FIRST SLIDE -->
<section >
<h1>Object-based Image Analysis (OBIA)</h1>
<h2>An introductory course</h2>
<em>Assoc-Prof Dr Stefan LANG | contributors: Assoc-Prof Dr Dirk Tiede</em><p><p><p>
<b>[1]</b> Why spatial image analysis? | <b>[2]</b> Regions and image objects | <b>[3]</b> Image segmentation | <b>[4]</b> Knowledge representation |
<b>[5]</b> Class modelling</p>
<p><small>University of Salzburg, Department of Geoinformatics, 2021/22/23<br></small></p>
<!-- AUTHORS: PUT YOUR PRESENTER NOTES HERE: -->
<aside class="notes">
This course introduces basic concepts of OBIA such as image segmentation, hierarchical representation, class modelling.
</aside>
</section>
<!-- AUTHORS: DEFINE SECTIONS WITHIN A SECTION TO GET VERTICAL SLIDES:-->
<section ID="Chapter 1">
<section>
<h1>[1] Why spatial image analysis?</h1>
</section>
<section>
<h2>Space and spatial image analysis</h2>
<img width=80% data-src="resources/02_0_space.png" alt="space">
<small><a href="https://www.youtube.com/watch?v=c7OO3qCfH9Y">https://www.youtube.com/watch?v=c7OO3qCfH9Y</a></i></small>
<figcaption>View from ISS to Earth. Depending on observation distance and resolution we perceive 'objects' in a certain scale.
Here we see the eye of a hurricane. © S. Lang (PLUS) / Photograph: A. Gerst (ESA)</figcaption>
<aside class="notes">
We assume it is a hurricane, even though we cannot be absolutely sure, it could either be a cyclone or taifun.
And we only know the name of the storm ("Katrina?") when we know the acquisition date.
</aside>
</section>
<section >
<h2 >Spatial image analysis - bridging remote sensing and GIS</h2>
<div class="r-stack" style="position:relative; width:1024px; height:768px; margin:0 auto;" >
<img class="fragment" data-fragment-index="0" src="resources/03_0_conceptualbridge.png" alt="conceptualbridge" style="position:absolute; top:0; left:0" />
<img class="fragment fade-in" data-fragment-index="1" src="resources/03_1_conceptualbridge.png" alt="conceptualbridge" style="position:absolute; top:0; left:0" />
<figcaption>OBIA can be considered a bridging concept between raster representations (spatial continua) und vector representations (spatial discreta). © S Lang (PLUS)</figcaption>
</div>
<aside class="notes">
OBIA bridges concepts from geoinformatics and remote sensing. While the vector model (left) represents geographical objects with crisp boundaries,
the raster model represents pixel brightness as a continous phenomenon (right).
</aside>
</section>
<section >
<h2>Geographical OBIA (GEOBIA) - a paradigm in image analysis </h2>
<img data-src="resources/04_0_GEOBIA.png" alt="GEOBIA paradigm"></a>
<br><figcaption>A significant literature body and a suite of international conference supporting the idea of (GE)OBIA as a paradigm. © S. Lang (PLUS)</figcaption>
</section>
<section >
<h2>GEOBIA - trans-application potential</h2>
<img data-src="resources/05_0_appl.png" alt="GEOBIA applications">
<br><figcaption>In a variety of application domains, OBIA may be used in bridging between geospatial models and
imaged representations. © S. Lang (PLUS)</figcaption>
</section>
<section>
<h2>GEOBIA provides real-world representations</h2>
<img data-src="resources/06_0_RealityModel.png" alt="RealityModel" />
<br><figcaption>OBIA mimicking our conceptual understanding of reality, facilitates image understanding and
complex representations of the real-world. © S. Lang (PLUS)</figcaption>
</section>
<section>
<h2> Space over colour? </h2>
<p>Im image analysis, spatial properties (<b>structure</b>) are as important as spectral properties (<b>colour</b>). In vision, spatial information even typically dominates colour information (Matsuyama and Hwang, 1990)</p>
<figure>
<img data-src="resources/07_0_spaceovercolor.jpg" alt="space over color" />
<figcaption>Once structural features make us understand there are agricultural fields shown on an image, we start exploring the type of fields. © S. Lang (PLUS)</figcaption>
</figure>
</section>
<section>
<h2> Boundaries in images</h2>
<p>Image data do not explictly contain or store <b>boundaries</b>. However, boundaries appear in several scales and are commonly used in human perception for image understanding. Boundaries can be extracted from image data using <b>image segmentation</b> techniques. </p>
<img data-src="resources/08_0_boundaries.png" width=70% height=70% alt="boundaries" />
<figcaption>Different spatial information derived from images and other continous datasets (e.g. DEM) © B. Riedler, S. Lang (PLUS)</figcaption>
</section>
<section>
<h2>Real-world objects on images </h2>
<p><b>Spatial concepts</b> are inherent in human vision - like context and relationships between image components / objects / texels any visual image interpretation <b>relies heavily</b> on these concepts. In automated remote sensing image analysis such approaches exists for many years, but usage and applicability is fluctuating</p>
<figure>
<div class="r-stack" style="position:relative; width:1000px; height:440px; margin:0 auto;" >
<img src="resources/13_0_Chiemsee.png" alt="Chiemsee" />
<figcaption>Real-world features can be characterised by spatial properties and spatial relations. © D. Tiede (PLUS)</figcaption>
</div>
</figure>
<aside class="notes">
Rivers and lakes have similar spectral behaviour but different spatial features (compact vs. elongated). Islands can appear in many different
ways, but are surrounded by water bodies.
</aside>
</section>
<section>
<h2>Spatial properties of objects</h2>
<figure>
<div class="r-stack" style="position:relative; width:1000px; height:440px; margin:0 auto;" >
<img src="resources/14_0_spatialAspects.png" alt="spatialAspects" />
<figcaption>Spatial (i.e. geometrical, textural and contextual) aspects that can be addressed, even on grey-scale image. © S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Extended set of target classes in spatial image analysis </h2>
<p>OBIA allows to address target classes, which are spatially defined. Basically, we can distinguish between two groups, <b>form-defined classes</b> (e.g. lake or river) and <b>spatio-relational classes</b> (context-related objects such as an island, an urban park, or composite objects such as a residential area).</p>
<figure>
<div class="r-stack" style="position:relative; width:1000px; height:440px; margin:0 auto;" >
<img src="resources/12_0_spatialImageAnalysis.png" alt="spatialImageAnalysis" />
<figcaption>OBIA expands the set of target classes as compared to pixel-based analysis. Geometric or spatio-relational classes
can be derived. © S. Lang, D. Tiede (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Overview spatial image analysis - from image to information </h2>
<figure>
<div class="r-stack" style="position:relative; width:1000px; height:440px; margin:0 auto;" >
<img src="resources/12_1_overview_spatial_image_analysis.png" alt="spatialImageAnalysis" />
<figcaption>Spatial image analysis - from image to information. © S. Lang, modified D. Tiede (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Spatial image analysis in AI </h2>
<figure>
<div class="r-stack" style="position:relative; width:1000px; height:440px; margin:0 auto;" >
<img src="resources/12_2_Spatial_Image_Analysis_and_AI.png" alt="spatialImageAnalysis" />
<figcaption>An integrator for AI? (from https://www.geospatialworld.net/blogs/difference-between-ai-machine-learning-and-deep-learning , modified A. Baraldi, D. Tiede (PLUS))</figcaption>
</div>
</figure>
</section>
</section>
<section ID="Chapter 2">
<section>
<h1>[2] Regions and image objects</h1>
</section>
<section>
<h2>Spatial auto-correlation</h2>
<p>According to Tobler's First Law of Geography (which states that "<i>near things are more related than distant ones</i>"), neighbouring pixels tend to be similar. Boundaries can be detected at discontinuities or gradients between these groups of similar pixels (see Segmentation below). Boundaries form objects on distinct levels (scales)</p>
<div class="r-stack" style="position:relative; width:320px; height:339px; margin:0 auto;" >
<img class="fragment fade-out" src="resources/09_0_spatialprincle.png" alt="spatial principle" />
<img class="fragment fade-up" src="resources/09_1_spatialprincle.png" alt="spatial principle" style="position:absolute;top:0; left:0"/>
<figcaption>A simple image matrix is converted to random colours (left). When using graduated grey shades it reveals a familiar pattern (right). © S. Lang, PLUS</figcaption>
</div>
<aside class="notes">
Once recognizing a human head with a face, we instantly perceive different levels: the level of the parts (eyes, beard, forehead), level of the whole (head).
</aside>
</section>
<section>
<h2>Spatial auto-correlation and grouping</h2>
<p>Spatial auto-correlation is a key spatial principle in image analysis. We rely on this principle when grouping pixels according to their similarity.
Without spatial autocorrelation the number of grouping options would be arbitrary. However, there is also ambiguity in the grouping process.</p>
<div class="r-stack" style="position:relative; margin:0 auto;" >
<figure>
<img src="resources/10_0_samesimilar.png" alt="samesimilar" />
<figcaption>By spatial autocorrelation neighboring pixels can be grouped. It is, however, a matter of optimisation and ambivalence ('ill-posed'). © S. Lang, PLUS</figcaption>
</figure>
</div>
<aside class="notes">
It is not about grouping the same values, but grouping similar values. But how can we determine similarity? We will see that similarity corresponds to
homogeneity (of regions), a fundamental principle of image segmentation.
</aside>
</section>
<section>
<h2>Spatial auto-correlation and boundaries</h2>
<figure>
<img src="resources/11_0_varianceRegions.png" alt="varianceRegions" />
<figcaption>Despite the similarity of neighbouring pixels, we eventually encounter gradients. Those gradients we can identify as boundaries. © S. Lang, PLUS</figcaption>
</figure>
</section>
<section>
<h2>Regions</h2>
<div class="ppt-txt" style="position: absolute; top: 100px; left: 10px; height: 150px; width: 400px">
<p>
<ul>
<li style> <strong>Region</strong> = a spatially contiguous area with common attributes or uniform behaviour (internal <strong>‘homogeneity’</strong>) </li> <br>
<li style> <strong>Regionalisation = spatial classification</strong>, i.e. classification under <strong>spatial constraints</strong> (contiguity of an attribute in space) </li> <br>
<li style> Why does it work? Principle of <strong>spatial auto-correlation</strong> (“1st law of Geography”, W Tobler, 1970)</li>
</ul> </p>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 35px; left: 240px; width:800px; height:420px; margin:0 auto;" >
<img src="resources/15_0_regions.png" alt="Regions" />
<figcaption> Different regions according to different DEM-derived properties (© S. Lang, PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Image Objects </h2>
<figure>
<div class="r-stack" style="position:relative; width:1000px; height:440px; margin:0 auto;" >
<img src="resources/16_0_imageObjects.png" alt="imageObjects" />
<figcaption>Image objects - different imagery (© S. Lang, PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Image Objects </h2>
<table>
<tr><td width="50%"><image src="resources/17_0_ImageObjects.png" alt="image_objects.png" align=left> </td>
<td>
<figure>
<image src="resources/17_1_ImageObjects.png" alt="image_objects.png" align=left> </td>
</tr></table>
<figcaption>Image objects - different imagery and applications</figcaption>
</figure>
</section>
<section>
<h2>What means homogeneous?</h2>
<div class="ppt-txt" style="position: absolute; top: 100px; left: 10px; height: 150px; width: 500px">
<p> <i> not a trivial task... </i>
<ul>
<li style> Challenges in generating regions </li> <br>
They need to be spatially coherent <br>
They do not exist a priori <br><br>
<li style> Defining Homogeneity criterion <strong>H</strong> </li> <br>
Variance* of pixels within a region<br>
Spectral distance (SD) of pixels <br>
Vectors in n-dimensional feature space<br>
</ul> <br><br>
<font size=-1>*as one (statistical) measure to operationalize homogeneity </font> </p>
</div>
<figure></figure>
<div class="r-stack" style="position:relative; top: 20px; left: 250px; width:1050px; height:450px; margin:0 auto;" >
<img src="resources/18_0_homogeneity.png" alt="Regions" />
<figcaption>Homogeneity vs spatial constraints (© S. Lang, PLUS)</figcaption>
</div>
</figure>
</section>
</section>
<section ID="Chapter 3">
<section>
<h1>[3] Multispectral Segmentation</h1>
</section>
<section>
<h2>Principles</h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 10px; height: 150px; width: 400px">
<p> Which image objects are we able to perceive?
<ul>
<li> From 252 digital numbers (14 x 18 image matrix) to ‘Abraham Lincoln’ (one individual person) </li>
</ul> <p>
What do images evoke in our visual context and our brain?
<li> Trying to ignore pixels, to identify familiar structures </li>
<li> Object hierarchy (here: exactly 2) </li>
<li> Semantic enrichment </li>
<li> Combination of experience and knowledge </li>
</ul> <p>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 50px; left: 240px; width:800px; height:420px; margin:0 auto;" >
<img src="resources/19_0_segmentation.png" alt="Regions" />
<figcaption>Image segmentation happens in our human brain. © S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Grey-scale Segmentation </h2>
<figure>
<div class="r-stack" style="position:relative; width:1000px; height:440px; margin:0 auto;" >
<img src="resources/20_0_segmentation.png" alt="imageObjects" />
<figcaption>A boundary can be found along gradients between pixel values. © S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Multispectral Segmentation </h2>
<figure>
<div class="r-stack" style="position:relative; width:1000px; height:440px; margin:0 auto;" >
<img src="resources/21_0_segmentation.png" alt="imageObjects" />
<figcaption>The same principle applies in multispectral images. © S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Segmentation Challenges </h2>
<strong>Conceptual boundaries in images </strong> <br>
Fiat boundaries, according to convention, expertise, etc. <br>
See lesson “Class modelling”
<table>
<tr><td width="50%"><image src="resources/22_0_segmentationChallenges.png" alt="image_challenges.png" align=left> </td>
<figure>
<td> <image src="resources/22_1_segmentationChallenges.png" alt="image_challenges.png" align=left> </td>
</tr>
</table>
<figcaption>Some boundaries in images appear distinctively, others are more conceptual. © S. Lang (PLUS)</figcaption>
</figure>
</section>
<section>
<h2>Segmentation Challenges </h2>
<strong>Conceptual boundaries in images </strong> <br>
Complex classes, often defined by spectrally heterogeneous, but functionally homogenous units
Segmentation rarely provides these units directly
Cyclic approach needed (class modelling)
Expert-based (object validity)
<figure>
<table>
<tr><td width="50%"><image src="resources/23_0_segmentationChallenges.png" alt="image_challenges.png" align=left> </td>
<td> <image src="resources/23_1_segmentationChallenges.png" alt="image_challenges.png" align=left> </td>
</tr>
</table>
<figcaption>Semi-automatically delineated complex biotope types. Left: open orchard meadows; Right: mixed arable land and grassland area (© S. Lang, PLUS)</figcaption>
</figure>
</section>
<section>
<h2>Image Segmentation</h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 10px; height: 150px; width: 400px">
<p> <strong> Local pixel neighborhood </strong>
<ul>
<li> Spatial autocorrelation among pixels </li>
<li> neighbouring pixels tend to have similar values </li>
<li> But not everywhere! </li>
</ul> <p>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 35px; left: 240px; width:800px; height:400px; margin:0 auto;" >
<img src="resources/25_0_segmentation.png" alt="Regions" />
<figcaption> Homogeneity and gradients in a small portion of an image matrix. © S. Lang (PLUS) </figcaption>
</div>
</figure>
</section>
<section>
<h2>Segmentation and classification</h2>
<div class="ppt-txt" style="align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 750px">
<p> <strong> Two interrelated methodological pillars (Lang 2008) </strong> <br><br>
(1) segmentation / regionalisation for nested, scaled representations ('object scape')
<ul>
<li> Segmentation of different types of continuous data </li>
<li> Optimizing segmentation (algorithms and parameterization) in terms of usability, theoretical foundation, etc. </li>
<li> Systematic comparison of segmentation results </li>
</ul> <p>
(2) rule-based classifiers based on spectral and geometrical properties as well as spatial relationships
<ul>
<li> Ontology driven classification, advanced sets of target classes </li>
<li> Process of OBIA not necessarily sequential (i.e. as described here), rather sequential, iterative (class modelling) </li>
<li> Smooth transition to methods to further characterise and analyse classification results (e.g. landscape metrics) </li>
</ul> <p>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 100px; left: 500px; width:800px; height:420px; margin:0 auto;" >
<img src="resources/26_0_segmentationclassification.png" alt="segmentation_classification" />
<figcaption>Yin-Yang allegory of segmentation and classification in OBIA. © S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Multi-scale segmentation</h2>
<div class="ppt-txt" style="position: absolute; text-align: center; top: 80px; height: 150px; width: 400">
<p> Segmentation in several <strong>hierarchical </strong> scales <br><br>
scale-specific vs. scale-adapted segmentation
</div>
<figure>
<div class="r-stack" style="position:relative; top: 150px; width:800px; height:420px; margin:0 auto;" >
<img src="resources/27_0_multiscalesegmentation.png" alt="multiscale_segmentation" />
<figcaption>Hierarchical, i.e. multiscale, representation. © S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Multi-scale segmentation</h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 10px; height: 150px; width: 400px"><br>
<p style = "text-align: left"> Representations in various nested levels <br>
Each object 'knows' its neighbours and hierarchical level
</div>
<figure>
<div class="r-stack" style="position:relative; top: 50px; left: 240px; width:800px; height:420px; margin:0 auto;" >
<img src="resources/28_0_multiscalesegmentation.png" alt="multiscale_segmentation" />
<figcaption>Schematic illustration of multi-scale segmentation. © F. Albrecht, S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Multi-scale segmentation </h2>
<figure>
<div class="r-stack" style="position:relative; width:1000px; height:440px; margin:0 auto;" >
<img src="resources/29_0_multiscalesegmentation.png" alt="multiscale_segmentation" />
<figcaption>Multi-scale segmentation in two hierarchical levels. © S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
</section>
<section ID="Chapter 4">
<section>
<h1>[4] Knowledge representation</h1>
</section>
<section>
<h2>Why knowledge representation?</h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 500px">
<p> We all have a concept of “cold”, “warm”, “hot” </p>
<ul>
<li> In remote sensing and image analysis a lot of expert knowledge exists </li>
<li> Such knowledge is implicit or explicit </li>
<li> The latter can be formalised and encoded in rules, (e.g. NDVI > 0.2 => vegetation) </li>
</ul>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 300px; width:800px; height:420px; margin:0 auto;" >
<img src="resources/31_0_knowledgerepresentation.png" alt="knowledge_representation" />
<figcaption>Knowledge is used in remote sensing to convert primary data (images) into GIS-ready secondary data. © S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>What type of knowledge can be used? </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 500px">
<p> Knowledge in image interpretation: visual skills are complemented by explicit knowledge, enriched by experience and training </p>
<p><strong> Structural</strong> knowledge
<ul>
<li> how concepts within a domain are interrelated </li>
<li> links between image objects and real-world geographical features </li>
</ul> <p>
<p><strong>Procedural </strong> knowledge
<ul>
<li> specific computational functions, algorithms </li>
<li> Encoding of structural knowledge, e.g. by a set of rules … or by a set of representative samples </li>
</ul> <p>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 300px; width:800px; height:420px; margin:0 auto;" >
<img src="resources/32_0_knowledge.png" alt="knowledge_representation" />
<figcaption>Structured knowledge (screenshot of a ruleset). © S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Some facts we can take for granted... </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 500px">
<p>
<ul>
<li> Physical laws and principles <i>(e.g. Stefan Boltzmann‘s Law </i>: The hotter a body, the more energy is emitted;
<i>Wien‘s Displacement Law</i>: The hotter a body, the shorter the wavelength of the maximum radiation) </li>
<li> Different land surface types show specific spectral profiles </li>
<li> Spatial autocorrelation </li>
<li> ... </li>
</ul>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 300px; width:800px; height:420px; margin:0 auto;" >
<img src="resources/33_0_facts.png" alt="others" />
<figcaption>Laws and theories we can trust in image analysis. © S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Others we know from experience... </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 520px">
<p>Heuristics describing certain geographical phenomena are encoded in a <strong> set of rules</strong>. Rules can address all kinds of features.</p>
<p>Example <i>soccer field</i> = a special kind of grassland patch with certain size and shape.
<ul>
<li> Grassland appears “red” in a false-colour band combination </li>
<li> How does a football field look like on Sentinel-2? </li>
<li> Object features (color, shape, size, etc.) </li>
<li> Object-to-neighbour relationships </li>
<li> Hierarchical relationships>
</ul>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 50px; left: 320px; width:800px; height:420px; margin:0 auto;" >
<img src="resources/34_0_others.png" alt="others" />
<figcaption>Heuristics of a soccer field translated into a rule-set © S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Fuzzy (rule) set</h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 500px">
<p> We all have a concept of “cold”, “warm”, “hot” <br><br>
<ul>
<li> We discretise a gradual phenomenon by categorizing it </li>
<li> Boundaries between these concepts are not crisp </li>
<li> The category under concern gets more stable with increasing ‘distance’ from boundaries </li>
<li> Fuzzy sets: a way to model / operationalise vague knowledge </li>
</ul> <p>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 50px; left: 320px; width:800px; height:420px; margin:0 auto;" >
<img src="resources/30_0_fuzzysets.png" alt="fuzzy_rule_sets" />
<figcaption>Fuzzyfication of class boundaries (example water temperature). © S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Class hierarchy (in eCognition) </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 500px">
<p> Classes are embedded in a hierarchical heritage scheme <br>
(1) Feature-based inheritance<br>
<ul>
<li> child classes inherit all descriptive features from their parent classes </li>
<li> Not confined to spectral values </li>
</ul>
<p>(2) Semantic inheritance <br>
<ul>
<li> Classes can be grouped semantically </li>
<li> Belong to the same logical parent class </li>
<li> Semantic inheritance may turn into feature-based when additional explicit information is provided </li>
</ul> <p>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 300px; width:800px; height:420px; margin:0 auto;" >
<img src="resources/35_0_classhierarchy.png" alt="class_hierarchy" />
<figcaption>Feature-based and semantic grouping of classes. © S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>What is semantic querying? </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 500px">
<p> Example: <i>Monitoring of water quality in swimming lakes over three seasons </i><br>
Image search criteria: [(0) sensor type], (1) <strong> time window </strong>, (2) <strong> geographical focus </strong> (AOI),
(3) <strong> cloud cover </strong>, (4) <strong> thematic focus </strong>, (5) <strong>monitoring period</strong>.<br>
<ul>
<li> Use standard metadata (location) and cloud mask </li>
<li> <strong>Thematic </strong> focus: requires a fundamental (i.e. low level semantic) concept of the existence and representation of (swimming) lakes. </li>
<li> Temporal aspect: (constancy or change) implies the notion for changes in critical indicators, e.g. by indices in IR reflectance recorded by multi-temporal imagery </li>
</ul> <p>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 50px; left: 300px; width:800px; height:420px; margin:0 auto;" >
<img src="resources/36_0_semanticquerying.png" alt="class_hierarchy" />
<figcaption>Semantic information stacked in a data cube through time © D. Tiede (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Data to information</h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 500px">
<p> How to measure the information in an image?<br>
<ul>
<li> <strong>Histogram</strong> (distribution of values)</li>
<li> Distribution of colours / colour levels </li>
<li> Image <strong>content</strong> (based on classification) </li>
</ul> <p>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 300px; width:600px; height:305px; margin:0 auto;" >
<img src="resources/37_0_histogram.png" alt="histogram" />
<figcaption>Histogram of NIR band of Landsat satellite data. © S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Data to information <i>(cont.)</i> </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 500px">
<p> Information measurement requires information <strong>units</strong><br> (bins, classified pixels, objects, etc.)<br>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 20px; left: 300px; width:800px; height:405px; margin:0 auto;" >
<img src="resources/38_0_information.png" alt="information" />
<figcaption>Information-as-a-thing (measurable) vs information-as-interpretation (semantic). © S. Lang (PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Quantifiable information content </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 30px; width: 500px">
<strong>Shannon's diversity</strong> is a measure of composition (information content).
<ul>
<li> Image <strong>content</strong> can be measured based on classification </li>
<li> Indices: <i>Diversity H, Dominance, Evenness</i> </li>
<li> Key elements: <b>categorised units</b>, number and percentages of classes</li>
</ul> <p>
<img src="resources/39_0_formula.png" width=50% height="50%" alt="information" />
<br><i>where P = percentage, i = class 1 … i</i>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 20px; left: 300px; width:900px; height:505px; margin:0 auto;" >
<img src="resources/39_1_information.png" alt="information" />
<figcaption>Shannon's Diversity measures the information content of a landscape. © S. Lang, PLUS</figcaption>
</div>
</figure>
</section>
<section>
<h2>Knowledge organising systems (KOS) </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 450px">
<p> Structural knowledge can be organized in knowledge organizing systems (KOS) <br>
<ul>
<li> Realised as graphic notations such as semantic nets or frames </li>
<li> Semantic knowledge representation </li>
<li> using inheritance concept, e.g. <i>‘is part of’, ‘is more specific than’, ‘is an instance of’ </i> </li>
</ul>
<p>Semantic net <br>
<ul>
<li> Control over existing connections, once established </li>
<li> Transparency and operability </li>
</ul> <p>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 70px; left: 300px; width:600px; height:300px; margin:0 auto;" >
<img src="resources/40_0_KOS.png" alt="KOS" />
<figcaption>(© S. Lang)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Classification schemes </h2>
<figure>
<div class="r-stack" style="position:relative; width:1000px; height:440px; margin:0 auto;" >
<img src="resources/41_0_classificationschemes.png" alt="classification_schemes" />
<figcaption>Classification schemes (© S. Lang, PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Classification schemes / taxonomies </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 450px">
<br><p> Corine Land Cover (CLC) <br>
<ul>
<li> Standard EU mapping scheme (used by European Environmental Agency EEA) initiated in 1985 (reference year 1990) </li>
<li> Updates produced in 2000, 2006, 2012, 2018 with the latest update (CLC+) under production - integrating the first time OBIA relevant segmentation techniques</li>
<li>CLC+ technical document: https://land.copernicus.eu/user-corner/technical-library/clc-core-consultations-for-the-technical-specifications</li>
</ul>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 300px; width:900px; height:450px; margin:0 auto;" >
<img src="resources/42_0_classificationschemes.png" alt="classification_schemes" />
<figcaption>(© S. Lang, PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Classification schemes </h2>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 0px; width:900px; height:450px; margin:0 auto;" >
<img src="resources/43_0_classificationschemes.png" alt="classification_schemes" />
<figcaption>Corine classification scheme (until 2018, before CLC+</figcaption>
</div><br><br>
CLC consists of an inventory of 44 classes. It uses a Minimum Mapping Unit (MMU) of 25 ha for areal phenomena and a minimum width of 100 m for linear phenomena
</figure>
</section>
<section>
<h2>Classification schemes </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 550px">
<br><p> Land Cover Classification System (FAO-LCCS) <br>
<ul>
<li> capable to record any land cover type, independent of specific applications and/or geographical areas </li>
<li> overcome problems associated with the interpretation of different land cover class definitions (Mucher et al. 2014) </li>
<li> uses a set of independent diagnostic criteria strictly based on vegetation physiognomy and structure rather than establishing land cover classes based on terminology (Kosmidou et al. 2014) </li>
<li> comprehensive land cover characterization, regardless of mapping scale, data collection method or geographic location. </li>
</ul>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 300px; width:800px; height:400px; margin:0 auto;" >
<img src="resources/44_0_clc.png" alt="classification_schemes" />
<figcaption>(© S. Lang, PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Classification schemes </h2>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 0px; width:1000px; height:550px; margin:0 auto;" >
<img src="resources/45_0_classificationschemes.png" alt="classification_schemes" />
<figcaption>(© S. Lang, PLUS)</figcaption>
</div><br>
</figure>
</section>
<section>
<h2>Spectral libraries </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 500px">
<p> <br>Spectral libraries represent spectral reflectance and physical properties of different land cover types <br>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 20px; left: 300px; width:600px; height:350px; margin:0 auto;" >
<img src="resources/46_0_spectrallibraries.png" alt="spectral_libraries" />
<i>simplified model!</i><br>
<figcaption>(© S. Lang, PLUS)</figcaption>
</div>
</figure>
</section>
</section>
<section id="Chapter 5">
<section>
<h1>[5] Class modelling</h1>
</section>
<section>
<h2>Composite objects</h2>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 0px; width:1000px; height:550px; margin:0 auto;" >
<img src="resources/47_0_compositeobjects.png" alt="composite_objects" />
<figcaption>(© S. Lang, PLUS)</figcaption>
</div><br>
</figure>
</section>
<section>
<h2>Composite objects </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 550px">
<br><p> Level of landscape elements <br>
<ul>
<li> Spectrally homogenous, correspond to agricultural fields, road segments, blocks of houses etc. </li>
</ul>
Composite objects
<ul>
<li> Contain elements (building blocks)</li>
<li> Are spectrally heterogeneous, but functionally homogenous </li>
<li> Geons (Lang et al. 2014)</li>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 300px; width:800px; height:400px; margin:0 auto;" >
<img src="resources/48_0_compositeobjects.png" alt="composite_objects" />
<figcaption>(© S. Lang, PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Class modelling </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 550px">
<br><p> <strong>Ambition: to delineate and classify composite objects in a complex real-world scene </strong> <br>
<ul>
<li> To represent the scene in at least in two hierarchical levels </li>
<li> To delineate homogenous elementary units on lower level(s) </li>
<li> To establish lateral and vertical relationships among objects (object-relationship modelling, ORM) </li>
<li> To characterise the composition of target classes based on the arrangement of elementary units </li>
</ul>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 370px; width:800px; height:400px; margin:0 auto;" >
<img src="resources/49_0_classmodelling.png" alt="classification_schemes" /><br>
<figcaption>(© Lang & Tiede 2015, PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Class modelling</h2>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 0px; width:1000px; height:550px; margin:0 auto;" >
<img src="resources/50_0_classmodelling.png" alt="class_modelling" />
<figcaption>(© Lang & Tiede 2015, PLUS)</figcaption>
</div><br>
</figure>
</section>
<section>
<h2>Class modelling </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 550px">
<br><p> A strategy to approach composite classes by applying multi-scale segmentation and spatial modelling <br>
<ul>
<li> The target classes represent functionally homogenous units – often spectrally heterogeneous –, as being composed by sub-units</li>
<li> The delineation of target units may be ambiguous (but inter-subjectively agreeable – thus non-arbitrary) </li>
<li> The spatial properties of composite classes (or objects) can be described and their structural arrangements of building blocks (elements) can be characterised </li>
</ul>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 370px; width:800px; height:400px; margin:0 auto;" >
<img src="resources/51_0_classmodelling.png" alt="classification_schemes" /><br>
<figcaption>(© Tiede et al. 2010)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Class modelling</h2>
<div class="r-stack" style="position:relative; top: 30px; left: 0px; width:1000px; height:550px; margin:0 auto;" >
<img src="resources/52_0_classmodelling.png" alt="class_modelling" />
<figcaption>(figures: © D. Tiede)</figcaption>
</div><br>
</section>
<section>
<h2>Multiscale class modelling </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 550px">
<br><p> “Multiscale” is implicit in the term class modelling <br>
<ul>
<li> A complex class (composite object) consists of several functional parts in a specific arrangement, such as: </li>
<li> ‘mixed arable land’ composed by a mix of agricultural and grassland fields </li>
<li> ‘residential area’ composed by family houses, gardens, shopping malls, etc. </li>
<li> Complex classes are hierarchically structured (‘body plan’) </li>
</ul>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 370px; width:900px; height:500px; margin:0 auto;" >
<img src="resources/53_0_multiscalemodelling.png" alt="multiscale_class_modelling" /><br>
<figcaption>(© D. Tiede)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Multiscale class modelling</h2>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 0px; width:1000px; height:550px; margin:0 auto;" >
<img src="resources/54_0_multiscalemodelling.png" alt="multiscale_class_modelling" />
<figcaption>(© S. Lang, PLUS)</figcaption>
</div><br>
</figure>
</section>
<section>
<h2>Multiscale class modelling </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 550px">
<br><p> <strong>Spectral signatures of elements </strong><br>
<ul>
<li> Spectral profiles are created by charting the object mean of representative class samples </li>
<li> Profiles can be translated in rules </li>
<li> Fuzzy rules can be applied to translate uncertainty in class descriptions </li>
</ul>
<img src="resources/55_0_multiscalemodelling.png" alt="multiscale_class_modelling" />
</div>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 320px; width:900px; height:500px; margin:0 auto;" >
<img src="resources/55_1_multiscalemodelling.png" alt="multiscale_class_modelling" /><br>
<figcaption>Spectral signatures</figcaption>
</div>
</figure>
</section>
<section>
<h2>Multiscale class modelling </h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 550px">
<br><p> Defining an object’s body-plan<br>
<ul>
<li> A higher-level objects consists of … </li>
<li> x% of object type A </li>
<li> y% of object type B </li>
<li> z% of … </li>
</ul>
<br><p> Critical reconditions<br>
<ul>
<li> Elementary units need to be classified (‘Lego blocks’) </li>
<li> Outline of composite object needs to be delineated </li>
</ul>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 30px; left: 320px; width:900px; height:500px; margin:0 auto;" >
<img src="resources/56_0_multiscalemodelling.png" alt="multiscale_class_modelling" /><br>
<figcaption>(© S. Lang, PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Multiscale class modelling</h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 550px">
<br><p> Specify set of target classes<br>
<ul>
<li> Define class descriptors </li>
<li> Including non-target classes </li>
<li> Perform classification </li>
</ul><br>
<img src="resources/57_0_multiscalemodelling.png" alt="multiscale_class_modelling" />
</div>
<figure>
<div class="r-stack" style="position:relative; top: 0px; left: 320px; width:900px; height:500px; margin:0 auto;" >
<img src="resources/57_1_multiscalemodelling.png" alt="multiscale_class_modelling" /><br>
<figcaption>Target classes and definition with fuzzy rules (© S. Lang, PLUS)</figcaption>
</div>
</figure>
</section>
<section>
<h2>Multiscale class modelling</h2>
<div class="ppt-txt" style="; align: left; position: absolute; top: 100px; left: 0px; height: 150px; width: 450px">
<br><p>Composite object (yellow outline) consists of several components (sub-objects)<br>
<ul>
<li> Create customized feature </li>
<li> Specify relative area of classified sub-objects </li>
</ul>
</div>
<figure>
<div class="r-stack" style="position:relative; top: 0px; left: 320px; width:900px; height:500px; margin:0 auto;" >
<img src="resources/58_0_multiscalemodelling.png" alt="multiscale_class_modelling" /><br>
<figcaption>Composite objects (© S. Lang, PLUS)</figcaption>
</div>
</figure>
</section>