Jannah, Miftahul (2018) Identifikasi Mutu Jagung Menggunakan Fitur Warna Dan Tekstur Berbasis Pengolahan Citra Digital Dan Algoritma K-Nearest Neighbor (K-Nn). Sarjana thesis, Universitas Brawijaya.
Abstract
Jagung merupakan komoditas tanaman pangan yang sebagian besar dimanfaatkan sebagai bahan baku pakan. Permasalahan saat ini, mutu jagung petani memiliki kadar air tinggi. Untuk pemenuhan jagung lokal, perusahaan harus memasok melalui pedagang. Penentuan mutu jagung di tingkat petani didasarkan pada lama pengeringan yang dapat mempengaruhi harga beli jagung. Metode ini memiliki kelemahan yaitu rendahnya efisiensi, objektifitas dan tingkat konsistensi serta dapat menimbulkan konflik antara pedagang dan petani. Karenanya, diperlukan metode identifikasi mutu jagung yang baik dan akurat berbasis pengolahan citra digital. Kemampuan pengolahan citra digital yang canggih memungkinkan proses identifikasi mutu komoditas pertanian menjadi lebih efektif dan efisien. Tujuan penelitian ini untuk merancang sistem identifikasi mutu dan varietas jagung berdasarkan fitur warna dan tekstur serta menentukan tingkat akurasi metode K-Nearest Neighbor (K-NN) dalam menduga mutu jagung di tingkat petani. Pada penelitian ini, standar mutu jagung didasarkan PERMENDAGRI No.27/M-DAG/PER/5/2017 tentang penetapan harga acuan pembelian jagung di petani. Varietas jagung yang digunakan yaitu Pertiwi-3 (P-3) dan Pertiwi-6 (P-6). Sistem identifikasi mutu jagung dirancang menggunakan input 7 fitur (H,S,V, contrast, correlation, energy, homogeneity) dan algoritma K-NN sebagai classifiers. Sistem ini mengklasifikasikan mutu jagung ke dalam 10 kategori mutu, yakni Kategori 1 (P-3 maks kadar air 15%), Kategori 2 (P-3 maks kadar air 20%), Kategori 3 (P-3 maks kadar air 25%), Kategori 4 (P-3 maks kadar air 30%), Kategori 5 (P-3 maks kadar air 35%), Kategori 6 (P-6 maks kadar air 15%), Kategori 7 viii (P-6 maks kadar air 20%), Kategori 8 (P-6 maks kadar air 25%), Kategori 9 (P-6 maks kadar air 30%), dan Kategori 10 (P-6 maks kadar air 35%). Hasil keluaran sistem berupa varietas, kadar air serta harga jagung. Berdasarkan hasil penelitian yang dilakukan, rasio pengujian yang digunakan 70%:30%, jumlah data latih yang digunakan sebanyak 350 citra dan data uji sebanyak 150 citra. Hasil penelitian menggunakan variasi nilai k (tetangga terdekat) 1,3,5,7,9 serta metode jarak perhitungan euclidean dan Cityblock menunjukkan bahwa sistem mampu mengidentifikasi mutu dan varietas jagung dengan baik. Akurasi tertinggi diperoleh saat nilai k=5 menggunakan Cityblock distance sebesar 90,00%.
English Abstract
Corn is a food commodity that most widely used for feed raw material. The current issues, the quality of corn at the farmers has high moisture content. Thus, feed industry should supply through wholesalers to absorb local corn. Quality determination for corn at farm-level based on drying time may affect the purchase price of corn. This method has disadvantages include low efficiency, objectivity, consistency and even remains a problem between the wholesaler and the farmer. Therefore, a good and accurate method for corn quality identification based digital image processing is required. Sophistication of digital image processing allow the quality identification process of agricultural commodities be more effective and efficient. This research aimed at designing an identification system of corn quality and varieties based on color and texture features, also determining the accuracy level of K-Nearest Neighbor (K-NN) for estimating quality of corn at farm-level. In this study, the quality standard of corn used under the Ministerial Regulation No. 27/M-DAG/PER/5/2017 on purchase price determination for corn at farm-level. Corn varieties used such as Pertiwi-3 (P-3) and Pertiwi-6 (P-6). Corn quality identification system designed with 7 features input (H,S,V, contrast, correlation, energy, homogeneity) and K-NN algorithm as classifiers. This system classify corn into 10 quality categories, namely Category 1 (P-3 moisture max 15%), Category 2 (P-3 moisture max 20%), Category 3 (P-3 moisture max 25%), Category 4 (P-3 moisture max 30%), Category 5 (P-3 moisture max 35%), Category 6 (P-6 moisture max 15%), Category 7 (P-6 moisture max 20%), Category 8 (P-6 moisture max 25%), Category 9 (P-6 moisture max 30%), and Category x 10 (P-6 moisture max 35%). Output of the system include category of corn quality, moisture content, varieties and purchase price of corn. Based on this study using testing ratio 70%:30%, there are 350 total dataset used for training and 150 images used for testing set. The results using variation value of k (nearest neighbor)1,3,5,7,9 with euclidean and cityblock distance method show that system is able to identify quality and varieties of corn well. The highest accuracy obtained when k=5 using Cityblock distance equal to 90,00%.
| Item Type: | Thesis (Sarjana) |
|---|---|
| Identification Number: | SKR/FTP/2018/211/051805358 |
| Uncontrolled Keywords: | harga jagung, K-NN, pengolahan citra digital, warna dan tekstur / color and texture, corn price, digital image processing, K-NN, |
| Subjects: | 600 Technology (Applied sciences) > 664 Food technology > 664.7 Grains, other seeds, their derived products |
| Divisions: | Fakultas Teknologi Pertanian > Teknologi Industri Pertanian |
| Depositing User: | Endang Susworini |
| Date Deposited: | 29 Oct 2019 07:34 |
| Last Modified: | 21 Nov 2024 04:46 |
| URI: | http://repository.ub.ac.id/id/eprint/165694 |
![[thumbnail of Miftahul Jannah.pdf]](https://repository.ub.ac.id/style/images/fileicons/text.png) |
Text
Miftahul Jannah.pdf Restricted to Registered users only Download (4MB) |
Actions (login required)
 |
View Item |