Automated Detection of Loose Palm Fruit for Quality Inspection

Abstract

Palm oil plays a significant role in driving the global economy due to the versatile nature of its fruit, which has led to increasing demand worldwide. However, several processes such as quality inspection and documentation—are still performed manually. This project aims to enhance the effectiveness of quality inspection procedures. It addresses current limitations by implementing object detection and recognition methods to support the sorting process. The system utilizes Artificial Intelligence (AI) to improve the efficiency of palm fruit quality evaluation. It is specifically designed to detect and classify loose palm fruits based on their ripeness level. The method used in creating the project is implementation of Huskylens that is used to detect and recognize the palm fruit ripeness level by its colour. The project utilizes the Firebase console for documentation and control, while the ESP32 acts as the central processing unit. Results show that detection and recognizing accuracy could be influenced by the conveyor speed where accuracy percentage increases as conveyor speed decreases. The results show, at 22V the conveyor moves at 0.075 m/s, yielding an average accuracy of 70%. However, the project is constrained by the camera’s field of view and resolution, which may affect its object detection capability. Additionally, the detection speed depends on the processing capabilities of the microcontroller used. This system is intended to benefit small-scale entrepreneurs and organizations who require a reliable and affordable solution to improve their workflow efficiency.