Electronic Theses and Dissertation

Salah satu produk turunan kelapa yang populer di Aceh adalah u neulhe. Kualitas uneulhe sangat dipengaruhi oleh proses pengolahan, terutama suhu dan waktu penyangraian,yang saat ini masih dinilai secara subjektif oleh masyarakat setempat. Penelitian ini pentingdilakukan guna menghadirkan metode yang lebih objektif dan akurat dengan memanfaatkanteknologi pengolahan citra digital dan kecerdasan buatan untuk mengidentifikasi tingkatkematangan u neulhe. Penelitian ini bertujuan untuk menyelidiki implementasi dan potensipenggunaan software MVTech Halcon berbasis citra digital serta metode K-NearestNeighbor (K-NN) dalam menciptakan sistem yang mampu mengelompokkan tingkatkematangan u neulhe berdasarkan warna secara lebih efisien dan akurat.Penelitian ini dilakukan di Laboratorium Teknik Pasca Panen Universitas SyiahKuala. Sebanyak 450 sampel u neulheu disiapkan dari dua perlakuan utama, yaitu lamapengeringan (0, 2, dan 4 jam) serta lama penyangraian (3, 6, 9, 12, dan 15 menit). Sampelini dibagi menjadi data training (315 sampel) dan data testing (135 sampel) untuk pengujianmodel klasifikasi menggunakan metode K-NN. Proses penelitian melibatkan pembuatancitra digital u neulheu dengan bantuan perangkat keras dan perangkat lunak MVTechHalcon. Tahapan meliputi akuisisi citra menggunakan kamera dan ring light, pre-processingdengan image enhancement dan cropping, segmentasi citra menggunakan metode otsu, sertaekstraksi citra yang dikonversi dari model warna RGB ke Lab. Data citra digunakan dalamklasifikasi dengan algoritma K-NN yang membandingkan jarak Euclidean antara datatraining dan data testing. Pengujian hasil klasifikasi dilakukan dengan menghitung precision,recall, dan F-measure untuk mengevaluasi kinerja model klasifikasi.Hasil penelitian menunjukkan bahwa klasifikasi tingkat kematangan kelapa sangraidapat dilakukan dengan menggunakan pengolahan citra dengan metode K-NN. Tingkatakurasi data testing menggunakan metode K-Nearest Neighbor yang didapat pada penelitianini untuk undercook, best, dan overcook yaitu 82.22 %, 86,67 % dan 100%. Tingkatkeberasilan metode K-NN dalam mengklasifikasikan tingkat kematangan kelapa sangraiyaitu 89,63%. Untuk mencapai hasil klasifikasi yang lebih akurat, diperlukan analisis yanglebih mendalam dengan mempertimbangkan fitur-fitur lain seperti tekstur dan distribusiwarna, bukan hanya bergantung pada nilai warna RGB atau HSV. Penelitian tentangklasifikasi tingkat kematangan kelapa sangrai dapat dilakukan dengan menambahkan jumlahdata pelatihan dan pengujian guna mendukung peningkatan akurasi.

One of the popular coconut derivative products in Aceh is u neulhe. The quality of u neulhe is greatly influenced by the processing process, especially the temperature and roasting time, which are currently still subjectively assessed by the local community. This research is important to be conducted in order to present a more objective and accurate method by utilizing digital image processing technology and artificial intelligence to identify the ripeness level of u neulhe. This study aims to investigate the implementation and potential use of MVTech Halcon software based on digital imagery and the K-Nearest Neighbor (K-NN) method in creating a system that is able to group the ripeness level of u neulhe based on color more efficiently and accurately. This research was conducted at the Post-Harvest Engineering Laboratory of Syiah Kuala University. A total of 450 u neulhe samples were prepared from two main treatments, namely drying time (0, 2, and 4 hours) and roasting time (3, 6, 9, 12, and 15 minutes). The samples were divided into training data (315 samples) and testing data (135 samples) for testing the classification model using the K-NN method. The research process involved creating digital images of u neulheu with the help of MVTechHalcon hardware and software. The stages include image acquisition using a camera and ring light, pre-processing with image enhancement and cropping, image segmentation using the Otsu method, and image extraction converted from the RGB color model to Lab. Image data is used in classification with the K-NN algorithm which compares the Euclidean distance between the training data and testing data. Classification results testing is done by calculating precision, recall, and F-measure to evaluate the performance of the classification model. The results showed that the classification of the level of ripeness of roasted coconut can be done using image processing with the K-NN method. The level of accuracy of testing data using the K-Nearest Neighbor method obtained in this study for undercook, best, and overcook is 82.22%, 86.67% and 100%. The success rate of the K-NN method in classifying the level of ripeness of roasted coconut is 89.63%. To achieve more accurate classification results, a more in-depth analysis is needed by considering other features such as texture and color distribution, not just relying on RGB or HSV color values. Research on the classification of the level of ripeness of roasted coconut can be done by adding the amount of training and testing data to support increased accuracy.