STUDI KOMPARASI KINERJA JARINGAN SARAF TIRUAN DAN FUZZY UNTUK PENGENALAN JENIS BUNGA BERDASARKAN FITUR WARNA

ISMI FITRIYANI, NIM. 09650036 (2014) STUDI KOMPARASI KINERJA JARINGAN SARAF TIRUAN DAN FUZZY UNTUK PENGENALAN JENIS BUNGA BERDASARKAN FITUR WARNA. Skripsi thesis, UIN SUNAN KALIJAGA.

Preview	Text (STUDI KOMPARASI KINERJA JARINGAN SARAF TIRUAN DAN FUZZY UNTUK PENGENALAN JENIS BUNGA BERDASARKAN FITUR WARNA) BAB I, V, DAFTAR PUSTAKA.pdf Download (4MB) | Preview
	Text (STUDI KOMPARASI KINERJA JARINGAN SARAF TIRUAN DAN FUZZY UNTUK PENGENALAN JENIS BUNGA BERDASARKAN FITUR WARNA) BAB II, III, IV.pdf Restricted to Repository staff only Download (6MB)

Abstract

There are so many flowering plants in this world, every flower has its own characteristic, such as colors. Human has the ability to recognize object, while machine recognizing ability has different process. The way to find its process, experiments for machine recognizing process is needed using artificial neural network and fuzzy. This experiments using 60 flower images (20 each for balloon, cucumber, and cypress vine flower), the flower images are in .jpg extension and 50x50 pixel size after preprocessing process (cropping and resizing). Then, after the feature extraction process, we can get the color feature for recognition process using Artificial Neural Network (ANN) and Fuzzy. Identification is using two methods, Backpropagation Neural Network (BPNN) and Fuzzy Mamdani Inference System for comparing the results of the accuracy of both methods. Backpropagation Neural Network shows 93,19% result of recognition. The optimal value for the mean squared error parameter is 0.001, learning rate 0,05, momentum 0,7, and epoch 1000. By using the optimal parameter for several ANN architecture, recongnition rate of 100% were obtained, but the lowest mean squared error generated by the one hidden layer with 16 nodes architecture. In fuzzy Mamdani’s reasoning experiments recognition rate of 88,19% were obtained by using 12 curve models (3 models each for 4 curve types). Bell curve gives the best result of 94,44% recognition rate.

Share this knowledge with your friends :

Actions (login required)

View Item