The V-I-S Model: Quantifying the Urban Environment - Remote Sensing of Urban and Suburban Areas - page 98

Fig. 6.12 Fraction images produced by the 4-endmember model utilizing vegetation, impervious,
dark soil, and shade. Brighter areas indicate higher abundance while darker areas indicate lower
abundance (Rashed et al. 2001 )

Table 6.5 A comparison between overall classiication accuracy and kappa accuracy for the three
classiications (Rashed et al. 2001 )

Maximum
likelihood Decision tree based on SMA Minimum distance

Overall accuracy 89.52 64.51 52.69

Kappa coefficient 0.88 0.59 0.45

• Multivariate calibration by an Lwin and Maritz ( 1980 ) technique which uses all

six reflective bands for sub-pixel identification that sum to 100 for each pixel
Display of four fractions in a single color image •

• A challenge to the conventional wisdom of “ground truth” wherein the burden

of RS classification is to “match” some other source of presumed correctness