IntisariTelah dilakukan penelitian dan pembuatan program komputer berbasis jaringan syaraf tiruan backpropagation yang bertujuan untuk mengklasifikasikan kualitas biji jagung berdasarkan pada pola distribusi intensitas RGB citra digital biji jagung tersebut. Dalam penelitian ini, kualitas biji jagung terklasifikasi dalam 4 kelompok, yaitu biji busuk, biji berjamur, biji normal, dan biji rusak. Jumlah sampel yang digunakan adalah 120 sampel untuk pelatihan dan 80 sampel untuk pengujian. Urutan tahapan penelitian adalah sebagai berikut: filterisasi citra digital dengan median filter, ekualisasi dengan histogram adaptif, ekstraksi indeks warna RGB untuk ketiga intensitas warna RGB, dan penghitungan mean dan standard deviasi untuk masing-masing indeks warna RGB tersebut. Selanjutnya dari pola mean dan standard deviasi ketiga indeks warna RGB dapat digunakan untuk mengenali kualitas biji jagung menggunakan metode jaringan syaraf tiruan backpropagation. Dalam penelitian ini, jaringan syaraf menggunakan fungsi aktivasi log-sigmoid dan dapat mengenali pola secara optimal bila digunakan 1500 iterasi, 500 neuron, 4 hidden layer, 4 output layer, dan laju pembelajaran 0,01. Hasil penelitian memperlihatkan bahwa jaringan syaraf tiruan yang telah dibuat ternyata mempunyai tingkat akurasi rata-rata sebesar 100% pada proses pelatihan dan sebesar 73,75% pada proses pengujian. ABSTRACTThe research and manufacture of computer programs based backpropagation neural network which aims to classify the quality of corn seeds based on RGB intensity distribution pattern of digital image of the corn seeds have been done. In this research, the quality of corn seeds classified into 4 groups, namely rotten seeds, moldy seeds, normal seeds and damaged seeds. The number of samples used is 120 samples for training and 80 samples for testing. The order of the stages of the research are as follows : filtering the digital image with the median filter, equalization with adaptive histogram, extracting RGB color index for the three RGB color intensity, and calculating the mean and standard deviation for each of the RGB color index. Furthermore, from the pattern of the mean and standard deviation of the three RGB color index can be used to identify the quality of corn seeds using backpropagation neural network method. In this research, the neural network using the log-sigmoid activation function and can recognize patterns optimally when used 1500 iterations, 500 neurons, 4 hidden layer, the output layer 4, and a learning rate of 0.01. The results showed that the neural network which has been made apparently has an average accuracy rate of 100% on the training process and amounted to 73.75% on the testing process.KATA KUNCI: Corn seeds quality, RGB color index, The mean and standard deviation pattern, Backpropagation neural network
The traditional sorting of fruit maturity can be done by seeing the color of the fruit’s skin. Manual sorting will take a long time and the results are subjective. This paper presents the results of maturing cantaloupe fruit based on the color of the fruit skin using a digital image of the fruit skin. The research objective is to classify the maturity of cantaloupe fruit using the Naive Bayes Classifier method and compare the results with similar studies using the Learning Vector Quantization (LVQ) Artificial Neural Network method. This study used the image of a raw and mature cantaloupe rind of 15 images each. A total of 16 images are grouped into training data for the training process and 14 other images are grouped into test data for the testing process. The results showed that the accuracy of training and testing using the Naive Bayes Classifier method was 68.75% and 57.14%, respectively. The accuracy of the training and testing of the Naive Bayes Classifier method turns out to be lower compared to the LVQ Artificial Neural Network method.
IntisariPaper ini menyajikan hasil analisis sebuah desain perangkat lunak berbasis jaringan syaraf tiruan backpropagation untuk mengklasifikasikan citra rontgen paru-paru. Citra rontgen paru-paru yang akan diklasifikasi harus melalui pemrosesan awal terlebih dahulu untuk membuang informasi yang tidak dibutuhkan agar kualitas citra dapat ditingkatkan. Untuk keperluan klasifikasi citra rontgen paru-paru ke dalam kelompok tertentu, digunakan proses ekstraksi fitur histogram pada citra rontgen paru-paru tersebut. Setelah sistem perangkat lunak dibuat, dua tahapan penting berikutnya adalah pelatihan dan pengujian pada sistem perangkat lunak tersebut. Dalam penelitian ini, desain perangkat lunak hanya dibatasi untuk dapat mengklasifikasikan citra rontgen ke dalam tiga kelompok yaitu citra rontgen paru-paru normal, citra rontgen paru-paru terkena kanker, dan citra rontgen paru-paru terkena efusi. Hasil pengujian memperlihatkan bahwa performansi sistem perangkat lunak yang telah dibuat dengan parameter epoch 500, error 0.001, learning rate 0.1 dan jumlah neuron 2500 ternyata memiliki tingkat akurasi sebesar 65%. ABSTRACTThis paper presents the results of analysis of a software design based backpropagation neural network to classify the X-ray image of the lungs. X-ray image of the lungs should be classified through initial processing prior to discard unneeded information that the image quality can be improved. For the purposes of classification of lung X-ray image into a particular group, use the histogram feature extraction process in X-ray image of the lungs . Once the software system is made, the next important step is two training and testing on the software system. In this study, the design of the software is limited to X-ray image can classify into three groups, namely X-ray image of a normal lung, X-ray image of the lung cancer , and the X-ray image of lungs affected by effusion. The test results show that the performance of software systems that have been created with 500 epoch parameters, error 0001, 0.1 learning rate and the number of neurons in 2500 turned out to have an accuracy rate of 65% .KATA KUNCI: image classification, backpropagation artificial neural network, feature extraction histogram, grayscale image of the lungs I. PENDAHULUANPengolahan citra dapat diaplikasikan dalam bidang medis untuk membantu dokter dalam mendiagnosa suatu penyakit. Pengolahan citra digital dalam bidang medis mengalami kemajuan penting ketika ditemukan metode tomografi terkomputerisasi (computerized tomography/ CT) pada tahun 1970-an [1]. Kini teknologi tomografi sudah berkembang pesat sekali. Beberapa contoh penerapan pengolahan citra untuk diagnosis penyakit, misalnya adalah untuk keperluan identifikasi tumor atau kanker rahim, identifikasi penyakit paru- * E-MAIL: a bustomi@physics.its.ac.id paru, identifikasi penyakit hati, identifikasi penyakit tulang, dan lain sebagainya [2].Makalah ini akan memaparkan sebuah aplikasi pengolahan citra untuk membantu dokter dalam menentukan jenis penyakit paru-paru. Dalam makalah ini peng...
The technique for increasing digital image resolution from low-resolution image to high-resolution image based on digital image processing is called the super-resolution technique. In this paper, a super-resolution technique is presented using a two-dimensional bi-cubic interpolation method in the spatial domain. The order of the super resolution method applied is as follows: (1) selecting ten images as samples, (2) decrease the sample image resolution to one-fourth of the original resolution by deleting three quarters of the pixel number, (3) increasing the image resolution of a quarter of the part becomes like the initial resolution using bi-cubic interpolation for three quarters of the additional new pixels, (4) testing this bi-cubic interpolated image with the same pixel-sized initial image, (5) using parameters: average value, minimum value, maximum value and standard deviation value as a comparison parameter between bi-cubic interpolated images and the same pixel-sized initial image. The results obtained from the super-resolution technique using spatial bi-cubic interpolation are: (1) The average error value of the bi-cubic interpolation method in image objects in this study is between 4% to 10% or still quite low, (2) Bi-cubic interpolation methods can work well on square pixel-sized images (m = n) compared to non-square pixel-sized images, (3) Bi-cubic interpolation turns out to produce an array of image pixel values that mirror symmetry against the main diagonal lines of the image before being interpolated.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.