Table 2.

Metrics used for the evaluation of segmentation.

Metrics	Formula	Explanations	Authors
Over-segmentation (OSeg)	$\mathit{OSeg}=1-\frac{\mathit{area}(x_i\cap y_j)}{\mathit{area}(x_i)}$xi – reference objects yj – evaluated objects	Range [0,1] OSeg = 0 → perfect segmentation	Clinton et al. (2010)
Under-segmentation (USeg)	$\mathit{USeg}=1-\frac{\mathit{area}(x_i\cap y_j)}{\mathit{area}(y_i)}$	Range [0,1] USeg = 0 → perfect segmentation	Clinton et al. (2010)
Root mean square (Dij)	$D_{\mathit{ij}}=\sqrt{\frac{{\mathit{OSeg}}_{\mathit{ij}}^2+{\mathit{USeg}}_{\mathit{ij}}^2}{2}}$	Range [0,1]; 0-perfect match	Levine and Nazif, 1985, Weidner, 2008
Area fit index (AFI)	$\mathit{AFI}=\frac{\mathit{area}(x_i)-\mathit{area}(y_i)}{\mathit{area}(x_i)}$	AFI = 0.0 → perfect overlap	Lucieer and Stein (2002)
Quality rate (Qr)	$\mathit{Qr}=\frac{\text{area}(X_i\cap y_i)}{\text{area}(X_i\cup y_i)}$	Range [0,1]; Qr 1 → perfect match	Winter (2000)