august 20, 2015 mapping croplands using landsat data with generalized classifier over large areas...
TRANSCRIPT
August 20, 2015
Mapping croplands using Landsat data with generalized classifier over large
areas
Aparna Phalke and Mutlu Ozdogan
Center for Sustainability and the Global Environment (SAGE)Nelson Institute for Environmental Studies
University of Wisconsin - Madison
Updates on following• Stratification of study area
• Developing New LDA models at three different levels according to new strata
• Scaling of data
• Area Based Accuracy
Zone Core_Scene Random_Scenezone1 177p39r (Egypt) 171p39r
182p38r(EgyptR1/Libya) 177p40r 194p36r
198p38r201p38r
zone2 174p35r (Syria) 171p35r172p34r (Turkey) 175p33r
179p33r(bulgariaR3)zone3 200p33r (Spain) 196p35r
201p37r (Morocco) 202p34r193p35r(Algeria) 200p31r192p35r(Tunisia)
zone4 181p26r (Ukraine) 174p28r176p26r173p25r
zone5 182p30r (Bulgaria) 190p26r187p27r(Hungary) 183p27r182p29r(Romania) 185p28r
184p30rzone6 199p26r(France) 196p26r
202p24r (UK) 197p27r200p27r207p22r
zone7 189p24r(Poland) 187p19r193p24r(Germany) 182p24r
182p21r(ukraineR1/Russia and some part Belarus) 183p25r
185p24r186p22r188p25r192p23r196p23r194p25r175p23r
zone8 203p33r(Portugal) 191p27r192p29r(Italy) 189p31r
196p29r(algeriaR1/france) 192p30r200p29r202p31r201p32193p27r
Total eight zones covering 23 core footprints and 36 random footprints
New own LDA model coefficients
mean sd max min var range count slope elevation cutoffpointegypt(177p39r) 4.87 -13.24 -2.87 -10.48 50.08 -2.40 0.00 0.02 0.01 -1.97libya(182p38r) 41.46 47.23 -19.04 -8.65 -14.85 -16.08 0.01 0.08 0.00 0.17syria(174p35r) 2.37 14.97 -2.58 5.34 -31.72 -3.34 0.00 0.07 0.00 0.33
turkey_west(172p34r) 8.72 -36.56 0.62 0.87 58.22 0.45 0.01 0.06 0.00 2.12
turkey_east(179p33r) 7.12 -3.17 -2.08 3.71 -7.21 -2.61 0.01 0.10 0.00 1.87
spain 15.59 -14.50 -1.25 0.37 36.72 -1.30 0.00 0.11 0.00 2.38morocco 9.12 -3.07 -4.53 -5.44 9.46 -3.07 0.02 0.09 0.00 -0.54algeria 8.67 -40.55 -1.23 -7.12 105.42 -0.37 0.01 0.14 0.00 -0.40tunisia 10.01 -47.08 0.07 -2.10 127.86 0.26 -0.01 0.14 0.00 -0.06
ukraine (181p26r) 14.55 -22.66 -5.33 -1.92 40.20 -0.57 0.03 -0.04 0.00 -1.65176p26r173p25r
bulgaria (182p30r) 14.14 -19.81 -5.16 -1.96 53.15 -1.57 0.00 0.09 0.00 -0.84hungary(187p27r) 13.60 -56.00 -0.88 -9.10 91.64 1.21 0.00 0.10 0.00 -1.85romania(182p29r) 12.56 -43.12 -2.01 -4.57 88.51 -0.46 0.02 0.18 0.00 -0.55france(199p26r) 8.79 -12.25 -2.81 -2.73 12.72 -1.45 0.21 0.11 -0.01 0.21
UK(202p24r) 2.28 -25.32 -2.67 -5.76 17.98 2.02 -0.06 0.19 0.00 -4.65poland(189p24r) 13.21 -34.20 -4.69 -14.80 62.44 -1.47 -0.01 0.63 0.00 -2.94
germany(193p24r) 13.49 -45.13 -2.95 -14.29 70.55 -0.04 0.00 0.18 0.00 -3.34Russia and some
part Belarus(182p21r)
15.25 -29.06 -11.44 -9.68 170.91 -7.38 0.13 0.30 0.02 -1.12
portugal(203p33r) 11.21 39.14 -3.96 5.29 -155.80 -4.47 0.02 0.14 0.00 3.45italy(192p29r) 6.01 -7.18 -4.07 -4.11 -3.46 -1.80 0.02 0.08 0.00 -1.29
franceORalgeriaR1(196p29r) 14.05 -11.51 -5.30 -5.81 11.29 -2.04 -0.04 0.04 0.00 -0.07
New zonal LDA model coefficients
mean sd max min var range count slope elevation cutoffpoint
zone1 4.819058 -6.02105 -4.3991 -7.36253 58.08601 -4.67452 0.018625 0.040549 0.002289 -1.45194
zone2 4.750754 -22.1928 -2.93649 -3.47778 89.60278 -3.50623 0.023859 0.099738 0.000543 -1.00297
zone3 8.569337 -8.63518 -1.94773 -0.7795 20.47229 -1.85149 0.022177 0.133289 0.0005 1.698421
zone5 12.17899 -28.8478 -3.87646 -3.20911 65.33539 -0.98515 0.022339 0.128569 0.000497 -0.44505
zone6 8.837099 -16.128 -4.35398 -6.26906 12.97489 -0.26407 0.00113 0.174757 -0.00569 -2.55152
zone7 18.62642 -4.18397 -7.01027 -15.1031 5.688592 -3.82965 -0.09017 0.250604 0.001585 -2.56448
zone8 10.25779 -1.15233 -5.14634 -2.3711 3.32551 -3.55784 0.048322 0.063277 0.000152 0.514091
New segment-based (Own-based accuracy)Overall
accuracyproducer_accuracy_
crop.i.producer_accuracy_noncrop.i.
user_accuracy_crop.i.
user_accuracy_noncrop.i.
egypt(177p39r) 84.93% 90.61% 69.65% 89.03% 72.95%
libya(182p38r) 87.55% 44.63% 95.52% 64.57% 90.30%
syria(174p35r) 76.77% 75.74% 78.07% 78.26% 75.40%
turkey_west(172p34r)
87.32% 87.59% 86.91% 92.66% 78.73%
turkey_east(179p33r)
78.30% 70.94% 85.32% 81.91% 75.68%
spain 83.23% 82.76% 84.27% 91.19% 71.18%morocco 77.68% 64.57% 90.00% 85.65% 73.15%algeria 74.93% 62.86% 87.42% 83.57% 69.64%tunisia 72.45% 66.92% 78.71% 77.69% 68.14%
ukraine (181p26r) 70.08% 69.43% 70.68% 65.57% 74.12%
176p26r173p25rbulgaria
(182p30r)79.36% 77.96% 80.98% 80.24% 78.60%
hungary(187p27r) 71.04% 83.49% 55.65% 70.08% 72.94%
romania(182p29r) 78.52% 85.54% 67.58% 80.68% 74.56%
france(199p26r) 76.77% 87.39% 65.45% 73.16% 82.71%
UK(202p24r) 63.24% 71.46% 50.23% 69.62% 52.37%
poland(189p24r) 72.25% 89.73% 40.73% 73.29% 68.54%
germany(193p24r)
74.28% 78.23% 70.89% 70.59% 78.37%
Russia and some part
Belarus(182p21r)67.12% 21.01% 94.25% 68.12% 67.01%
portugal(203p33r) 76.98% 59.00% 88.64% 76.77% 77.09%
italy(192p29r) 81.08% 80.35% 81.89% 80.07% 82.04%
franceORalgeriaR1(196p29r)
76.81% 53.25% 89.49% 72.54% 78.32%
zoneegypt(177p39
r)86.07%
libya(182p38r)
85.61%
syria(174p35r)
73.78%
turkey_west(172p34r)
84.60%
turkey_east(179p33r)
75.84%
spain 83.98%morocco 75.89%algeria 67.74%tunisia 69.80%
ukraine (181p26r)176p26r173p25r
bulgaria (182p30r)
77.30%
hungary(187p27r)
73.01%
romania(182p29r)
80.00%
france(199p26r)
71.87%
UK(202p24r) 57.81%
poland(189p24r)
73.18%
germany(193p24r)
68.07%
Russia and some part
Belarus(182p21r)
83.25%
portugal(203p33r)
62.88%
italy(192p29r)
72.99%
franceORalgeriaR1(196p29
r)77.94%
Area Based AccuracyTraditional pixel based approach accuracy:
where π is overall accuracy, Ci is equal to 1 or 0 if the validation sample unit i is correctly classified (yes and no, respectively), and n is the number of validation units collected.
The accuracy of a map created using OBIA should be computed using:
where N is the total number of segments in the image, and Si is the area of a single sample unit i.
Reference: Meghan MacLean, Russell Congalton, “Map accuracy assessment issues when using an object based oriented approach” Dissertation, University of New Hampshire.
Thank you