Biela, Stania Salsa (2026) Perancangan Model Prediksi Produktivitas Padi Berbasis Convolutional Neural Network (CNN) dan Particle Swarm Optimization (PSO). Diploma thesis, Institut Teknologi Sepuluh Nopember.
|
Text
2042221138_Undergraduate-Thesis.pdf - Accepted Version Restricted to Repository staff only Download (4MB) | Request a copy |
Abstract
Produktifitas padi memiliki peran strategis dalam ketahanan pangan nasional. Namun, Produktivitas padi merupakan salah satu parameter penting dalam mendukung pengelolaan pertanian yang efektif. Prediksi produktivitas berdasarkan citra tanaman dapat menjadi alternatif untuk memperoleh estimasi hasil panen secara lebih cepat dan efisien. Penelitian ini bertujuan merancang model prediksi produktivitas padi berbasis Convolutional Neural Network (CNN) dengan optimasi Particle Swarm Optimization (PSO). Dua pendekatan model digunakan, yaitu CNN Custom dan MobileNetV2, yang dibandingkan sebelum dan setelah optimasi. Data penelitian berupa citra tanaman padi pada fase pematangan yang dipasangkan dengan data hasil panen dari setiap ubin berukuran 1 m × 1 m. Kinerja model dievaluasi menggunakan Root Mean Square Error (RMSE) dan koefisien determinasi (R²). Hasil pengujian menunjukkan bahwa PSO meningkatkan kinerja kedua model. Pada CNN Custom, nilai R² meningkat dari 0,2231 menjadi 0,3411, sedangkan RMSE menurun dari 0,1041 kg/m² menjadi 0,0958 kg/m². Pada MobileNetV2, nilai R² meningkat dari 0,4660 menjadi 0,6745, sementara RMSE menurun dari 0,1414 kg/m² menjadi 0,1104 kg/m². Berdasarkan hasil tersebut, MobileNetV2–PSO menghasilkan nilai R² tertinggi dan dipilih untuk implementasi real-time inference menggunakan NVIDIA Jetson Orin Nano. Penelitian ini menunjukkan optimasi PSO dapat meningkatkan kinerja model CNN dalam memprediksi produktivitas padi berdasarkan citra tanaman.
======================================================================================================================================
Rice productivity is an important parameter in supporting effective agricultural management. Productivity prediction based on crop images can provide an alternative for obtaining harvest estimates more quickly and efficiently. This study aims to develop a rice productivity prediction model based on a Convolutional Neural Network (CNN) optimized using Particle Swarm Optimization (PSO). Two model approaches were used, namely Custom CNN and MobileNetV2, which were compared before and after optimization. The dataset consisted of rice crop images during the maturation phase paired with harvest data from each 1 m × 1 m observation plot. Model performance was evaluated using Root Mean Square Error (RMSE) and the coefficient of determination (R²). The test results showed that PSO improved the performance of both models. For Custom CNN, the R² value increased from 0.2231 to 0.3411, while RMSE decreased from 0.1041 kg/m² to 0.0958 kg/m². For MobileNetV2, the R² value increased from 0.4660 to 0.6745, while RMSE decreased from 0.1414 kg/m² to 0.1104 kg/m². Based on these results, MobileNetV2–PSO achieved the highest R² value and was selected for real-time inference implementation using the NVIDIA Jetson Orin Nano. The results indicate that PSO optimization can improve the performance of CNN models for predicting rice productivity from crop images.
| Item Type: | Thesis (Diploma) |
|---|---|
| Uncontrolled Keywords: | CNN, MobileNetV2, PSO, produktivitas padi, citra tanaman, prediksi. CNN, MobileNetV2, PSO, rice productivity, crop images, prediction. |
| Subjects: | T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques. Image analysis--Data processing. |
| Divisions: | Faculty of Vocational > Instrumentation Engineering |
| Depositing User: | Stania Salsa Biela |
| Date Deposited: | 15 Sep 2026 02:48 |
| Last Modified: | 15 Sep 2026 02:48 |
| URI: | http://repository.its.ac.id/id/eprint/144481 |
Actions (login required)
![]() |
View Item |
