An interactive flower image recognition system
An interactive flower image recognition system
An interactive flower image recognition system
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
66 Multimed Tools Appl (2011) 53:53–73in the database, notated dist i , is measured by the weighted Euclidean distance defined asfollows:dist i ¼P NfwðkÞjf i ðkÞk¼1P Nfk¼1wðkÞf ðkÞj; ð27Þwhere f i (k) isthek-th feature value of the i-th <strong>image</strong>, f(k) isthek-th feature value of the input<strong>image</strong>, and w(k) is the weight associated with the k-th feature value. The variable N f determineswhich type of features is used for <strong>flower</strong> <strong>image</strong> <strong>recognition</strong>. If only the features extracted fromthe <strong>flower</strong> region are used for <strong>recognition</strong> purpose, set N f =12; if features extracted from boththe <strong>flower</strong> region and the pistil/stamen area are used for <strong>recognition</strong> purpose, set N f =18. In thispaper, the normalized top five <strong>recognition</strong> rate, NAR5(k), is used as the weight associated withthe k-th feature value (1 ≤ k ≤ 18). That is, w(k) =NAR5(k). NAR5(k) is derived by normalizingthe top five <strong>recognition</strong> rate associated with the k-th feature value, AR5(k):wðkÞ ¼NAR5ðkÞ ¼P N fAR5ðkÞj¼1AR5ðjÞ 100;ð28Þwhere AR5(k) denotes the <strong>recognition</strong> rate when the k-th feature value is individually used for<strong>flower</strong> <strong>recognition</strong> and the <strong>recognition</strong> result will be regarded as accurate if at least one of thespecies in the top five candidates is identical to the input one. Table 1 shows the top five<strong>recognition</strong> rate (AR5(k)) and the weight (w(k)) associated with each feature value for the first<strong>flower</strong> <strong>image</strong> database. Note that these initial weights are distinct for different databases. BasedTable 1 Top five <strong>recognition</strong>rate AR5(k) and weight w(k)associated with each feature valuefor the first <strong>flower</strong> <strong>image</strong>databaseFeature AR5(k) (%) w(k)f(1) : CF 1 66.67 6.91f(2) : CF 2 68.39 7.08f(3) : CF 3 58.62 6.07f(4) : CF 4 63.51 6.58f(5) : CF 5 66.95 6.93f(6) : CF 6 43.97 4.55f(7) : CF 7 61.49 6.37f(8) : CF 8 57.47 5.95f(9) : CF 9 45.11 4.67f(10) : SF 1 58.05 6.01f(11) : SF 2 39.94 4.14f(12) : SF 3 60.63 6.28f(13) : PCF 1 59.77 6.19f(14) : PCF 2 59.77 6.19f(15) : PCF 3 54.31 5.63f(16) : PSF 1 34.48 3.57f(17) : PSF 2 29.60 3.07f(18) : PSF 3 36.78 3.81
Change language