A First Course in FUZZY and NEURAL CONTROL

8.4. IDENTIFICATION OF TRASH IN COTTON 275

Classification results of test samples using ANFIS The classification of
various trash objects in test sample #1 (TA114) is illustrated in Table 8.10.

Table 8.10. Classification results of trash objects in test sample #1

(TA114) using ANFIS: 232-partition

Pat-

tern

Blob

Id

Area Solidity Edif

Net

output

Actual

trash

Class-

ified

trash

1 1 22 0.9565 0.1222 0.7952 Pepper Pepper

2 3 100 0.9346 0.1172 0.8089 Pepper Pepper

3 4 17 1.0000 0.0283 0.7927 Pepper Pepper

4 6 28 1.0000 0.0952 0.8116 Pepper Pepper

5 7 124 0.9688 0.2085 0.7305 Pepper Pepper

6 9 113 0.9658 0.0395 0.8081 Pepper Pepper

7 10 13 0.8125 0.0867 0.9166 Pepper Pepper

8 13 84 0.7568 0.3180 1.7399 Stick Pepper

9 16 491 0.5696 0.1861 0.2777 Bark Bark

10 20 11 1.0000 0.2292 0.7963 Pepper Pepper

11 22 46 0.6479 0.0621 1.3404 Pepper Pepper

12 23 128 0.9343 0.1389 0.7993 Pepper Pepper

13 26 11 1.0000 0.0000 0.7815 Pepper Pepper

14 28 155 0.8908 0.2325 0.7408 Pepper Pepper

15 31 42 0.9545 0.0562 0.8004 Pepper Pepper

16 34 16 0.8421 0.0381 0.8223 Pepper Pepper

17 35 49 0.6622 0.0900 1.3191 Pepper Pepper

18 36 17 1.0000 0.1133 0.8045 Pepper Pepper

19 47 31 0.7209 0.0870 1.1874 Pepper Pepper

20 50 34 0.9444 0.0984 0.7982 Pepper Pepper

21 51 126 0.8873 0.2447 0.7873 Pepper Pepper

22 55 111 0.9407 0.2265 0.7727 Pepper Pepper

23 56 536 0.5714 0.2496 0.4998 Bark Stick

24 57 70 0.9091 0.0696 0.8050 Pepper Pepper

25 58 75 0.9494 0.0393 0.8065 Pepper Pepper

26 60 11 1.0000 0.0764 0.7952 Pepper Pepper

27 66 19 0.8261 0.2060 0.8072 Pepper Pepper

28 68 23 0.9200 0.1095 0.7966 Pepper Pepper

29 71 19 0.9500 0.0452 0.7913 Pepper Pepper

30 75 22 0.8148 0.3951 0.9282 Pepper Pepper

31 76 13 0.8667 0.0217 0.7838 Pepper Pepper

The classification results summarized in Table 8.11 for thefive test samples
are unsatisfactory for bark, stick, andleaf objects in the samples. Even though
the classification rate is 91.8367%, the classification of the individual trash types
is poor ñ especially for bark, stick, and leaf objects. Due to the presence of pep-
perinhighnumbersinthecottonsamples,andsinceallthreeclassifiers classify