Pengaruh Ukuran Cropping Terhadap Tingkat Akurasi Pengenalan Jenis Bibit Jeruk Melalui Citra Urat Daun dengan Metode JST Menggunakan Fitur GLCM

Muhammad Zulfikar (Universitas Indo Global Mandiri)
Gasim Gasim (Universitas Indo Global Mandiri)
Zaid Romegar Mair (Universitas Indo Global Mandiri)

Jurnal Nasional Komputasi dan Teknologi Informasi
Vol 8, No 5 (2025): Oktober 2025

Abstrak − Penelitian ini mengevaluasi pengaruh variasi ukuran cropping citra terhadap tingkat akurasi pengenalan jenis bibit jeruk berdasarkan tekstur urat daun menggunakan metode Jaringan Saraf Tiruan (JST) dengan fitur Gray Level Co-occurrence Matrix (GLCM). Citra daun dari lima varietas jeruk (manis, nipis, lemon, bali, sunkist) diolah dengan lima ukuran cropping (160×120, 320×240, 400×300, 480×360, dan 640×480 piksel) melalui konversi ke grayscale, ekstraksi fitur GLCM (kontras, korelasi, homogenitas, energi), dan klasifikasi menggunakan JST. Total 1000 citra (800 latih, 200 uji) digunakan dalam penelitian. Hasil menunjukkan bahwa ukuran cropping 480×360 piksel menghasilkan akurasi tertinggi sebesar 76,0%, sedangkan ukuran 160×120 piksel menghasilkan akurasi terendah (54,0%). Resolusi cropping menengah terbukti optimal dalam mempertahankan detail tekstur tanpa menambah kompleksitas komputasi. Penelitian ini menegaskan pentingnya pemilihan ukuran cropping dalam meningkatkan efektivitas sistem pengenalan citra berbasis JST dan GLCM.Kata Kunci: Pengenalan Jenis Bibit Jeruk; Ukuran Cropping; GLCM; JST; Tekstur Urat Daun; Abstract − This study evaluates the effect of varying image cropping sizes on the accuracy of citrus seedling type recognition based on leaf vein texture using the Artificial Neural Network (ANN) method with Gray Level Co-occurrence Matrix (GLCM) features. Leaf images from five citrus varieties (sweet orange, lime, lemon, pomelo, and sunkist) were processed with five cropping sizes (160×120, 320×240, 400×300, 480×360, and 640×480 pixels) through grayscale conversion, GLCM feature extraction (contrast, correlation, homogeneity, energy), and classification using ANN. A total of 1000 images (800 training, 200 testing) were used. Results show that the 480×360 pixel cropping size achieved the highest accuracy of 76.0%, while the 160×120 pixel size yielded the lowest accuracy of 54.0%. Medium-resolution cropping proved optimal in retaining texture details without increasing computational complexity. This study highlights the importance of selecting appropriate cropping sizes to enhance the effectiveness of image-based recognition systems using ANN and GLCM.Keywords: Citrus Seedling Recognition; Cropping Size; GLCM; ANN; Leaf Vein Texture;