Automatic oil palm unstripped bunch (USB) counting system based on faster RCNN and object tracking

• Main Authors: Wahyu, Sapto Aji, Kamarul Hawari, Ghazali, Son Ali, Akbar
• Format: Conference or Workshop Item
• Language: English; English
• Published: IEEE 2021
• Subjects: TK Electrical engineering. Electronics Nuclear engineering
• Online Access: http://umpir.ump.edu.my/id/eprint/41993/1/Automatic%20Oil%20Palm%20Unstripped%20Bunch_ABST.pdf; http://umpir.ump.edu.my/id/eprint/41993/2/Automatic%20Oil%20Palm%20Unstripped%20Bunch.pdf; http://umpir.ump.edu.my/id/eprint/41993/; https://doi.org/10.1109/IC2SE52832.2021.9792068

USB Palm Oil Counting (Unstripped Bunch) is important in oil palm processing mills. Information on the number of USBs is essential because it shows the level of efficiency of the palm oil processing plant. Due to its complexity, palm oil mills rely solely on manual calculations, which is inefficient workforce waste. Challenging aspects such as partial occlusion, overlap, and even different perspectives limit the use of traditional computer vision techniques. In recent years, deep learning has become increasingly popular for computer vision applications due to its superior performance over conventional methods. This paper proposes a deep learning solution to solve USB computing problems in palm oil mills. Our proposed automated USB counter system consists of an object detector built on the RCNN Faster architecture and an object tracker made in the euclidean distance. Our proposed system identifies and counts USBs with an average accuracy of 71.5% in testing.