Implementasi eCognition Object-Based Image Analysis untuk Sensus Pokok Kelapa Sawit Otomatis menggunakan Citra Satelit

Abstract

Sensus pohon kelapa sawit secara manual memerlukan waktu dan biaya yang relatif besar sehingga diperlukan metode yang lebih efisien. Penelitian ini bertujuan mengimplementasikan object-based image analysis (OBIA) menggunakan perangkat lunak eCognition untuk sensus populasi kelapa sawit otomatis berbasis citra satelit. Kinerja metode dievaluasi menggunakan confusion matrix, sedangkan efisiensi operasional dianalisis berdasarkan waktu dan biaya. Hasil penelitian menunjukkan bahwa metode OBIA menghasilkan accuracy sebesar 90,13%, precision 92,19%, recall 97,05%, dan F1-score 94,77%. Inventarisasi terhadap 8.380 pohon kelapa sawit pada sepuluh blok memerlukan waktu 352,01 menit, sehingga menghasilkan efisiensi waktu sebesar 91,62% dibandingkan metode manual. Biaya operasional metode OBIA menghasilkan efisiensi biaya sebesar 89,56%. Hasil penelitian menunjukkan bahwa implementasi OBIA menggunakan eCognition mampu menghasilkan inventarisasi pohon kelapa sawit yang akurat serta meningkatkan efisiensi waktu dan biaya operasional.

Manual oil palm inventory requires considerable time and operational cost, highlighting the need for a more efficient approach. This study aimed to implement object-based image analysis (OBIA) using eCognition software for automated oil palm population census based on satellite imagery. Method performance was evaluated using a confusion matrix, while operational efficiency was assessed through time and cost analyses. The OBIA method achieved an accuracy of 90.13%, precision of 92.19%, recall of 97.05%, and an F1-score of 94.77%. The inventory of 8,380 oil palm trees across ten plantation blocks required 352.01 minutes, resulting in 91.62% time efficiency compared with the manual census. Operational costs were reduced into 89.56% efficiency. These results demonstrate that OBIA using eCognition provides accurate oil palm inventory while substantially improving operational efficiency.