Authors

Abstract

Traditional manual cutting of seed potatoes is a labor-intensive, time-consuming, and inconsistent process that limits large-scale agricultural productivity. To address these challenges, this study aimed to develop and validate an automated, high-speed robotic system for precise cutting angle estimation. We propose SPADE (Smart Potato Angle Decision Engine), an innovative framework integrating deep learning and machine learning algorithms. The SPADE framework is implemented in three stages. First, a custom detection model, termed BUD-YOLO, was developed for the high-precision identification of potato eyes. Second, the K-means algorithm was employed to partition the spatial coordinates of the detected eyes into two distinct clusters. Finally, a Support Vector Machine (SVM) determined the optimal cutting plane by identifying the maximum-margin hyperplane between these two clusters. The proposed SPADE framework was implemented and tested on a custom-built robotic platform with a sample of 100 potatoes. The system achieved an 85% cutting qualification rate with an average processing time of 2.5 seconds per potato, a speed approximately 2-3 times faster than traditional manual labor (5-9 seconds per potato). This study successfully demonstrates an end-to-end solution for the automated cutting of seed potatoes. The developed SPADE framework not only achieves a competitive qualification rate but also significantly enhances production throughput, offering substantial practical value for the advancement of intelligent agricultural equipment.

Abstract in English

传统的手工切割马铃薯种子是一种劳动密集型、耗时且不一致的过程，限制了大规模农业生产力。为解决这些挑战，本研究旨在开发并验证一个能够进行精确切割角度估算的自动化、高速机器人系统。我们提出SPADE（智能马铃薯角度决策引擎），这是一个创新框架，整合了深度学习和机器学习算法。SPADE框架分为三个阶段。首先，开发了一个定制的检测模型，称为BUD-YOLO，用于高精度识别马铃薯芽眼。其次，使用K-means算法将检测到的芽眼的空间坐标分为两个独立的聚类。最后，支持向量机（SVM）通过识别这两个聚类之间的最大边际超平面来确定最佳切割平面。提出的SPADE框架在定制的机器人平台上使用100个马铃薯样本进行了实施和测试。整个系统实现了85%的切割合格率，平均处理时间为每个马铃薯2.5秒。这一处理速度约为传统人工劳动的2-3倍（每个马铃薯5-9秒）。本研究成功展示了马铃薯种子自动切割的端到端解决方案。所开发的SPADE框架不仅实现了具有竞争力的通过率，更重要的是显著提升了生产效率，为智能农业装备的发展提供了重要的实践价值。