Ma Chan, Ying Yibin, Xie Lijuan
School of Biosystems Engineering and Food Science, Zhejiang University, 866 Yuhangtang Road, Hangzhou 310058, PR China; The National Key Laboratory of Agricultural Equipment Technology, Hangzhou, Zhejiang 310058, PR China.
School of Biosystems Engineering and Food Science, Zhejiang University, 866 Yuhangtang Road, Hangzhou 310058, PR China; The National Key Laboratory of Agricultural Equipment Technology, Hangzhou, Zhejiang 310058, PR China; Key Laboratory of on-Site Processing Equipment for Agricultural Products, Ministry of Agriculture and Rural Affairs, Hangzhou, Zhejiang 310058, PR China.
Food Chem. 2025 Mar 1;467:142282. doi: 10.1016/j.foodchem.2024.142282. Epub 2024 Nov 29.
Precise measurement of firmness was crucial for determining optimal harvesting times, implementing rational storage strategies and minimizing avoidable waste. Current technologies for assessing peach firmness struggled to balance high precision and non-destructive methods, while demonstrating high sensitivity to environmental disturbances, thereby limiting their application to production line. Future various scenarios in agriculture would increasingly rely on robotic arms, yet existing firmness assessment technologies were not compatible with these automated systems. Additionally, monitoring contact force was essential for flexible operation of the robotic arms. This work introduced a visuo-tactile sensor equipped with markers to capable of measuring peach firmness and monitoring contact force simultaneously during a single contact process, making it suitable for robotic arm applications. The contact was operated by the texture analyzer to simulate the fruit grasping process by a robotic arm. Utilizing deep neural networks and machine learning-based techniques to process high-precision geometric images collected by an internal camera, the visuo-tactile sensor achieved non-destructive measurements of peach firmness and contact force. For firmness measurement in the test set, the sensor achieved coefficient of determination (R) of 0.878 and a root mean square error (RMSE) of 0.732. For contact force detection, the R was 0.942, and RMSE was 1.115 in the test set. The results showed visuo-tactile sensor was feasible for non-destructive detection of peach firmness and contact force, and has a broad application prospect in the field of agricultural robotics.
精确测量果实硬度对于确定最佳采收时间、实施合理的贮藏策略以及减少可避免的浪费至关重要。当前用于评估桃子硬度的技术难以在高精度和非破坏性方法之间取得平衡，同时对环境干扰表现出高度敏感性，从而限制了它们在生产线上的应用。未来农业中的各种场景将越来越依赖机械臂，但现有的硬度评估技术与这些自动化系统不兼容。此外，监测接触力对于机械臂的灵活操作至关重要。这项工作引入了一种配备标记的视觉触觉传感器，能够在单次接触过程中同时测量桃子硬度和监测接触力，使其适用于机械臂应用。通过纹理分析仪操作接触过程，以模拟机械臂抓取果实的过程。利用深度神经网络和基于机器学习的技术处理内部摄像头收集的高精度几何图像，视觉触觉传感器实现了对桃子硬度和接触力的无损测量。在测试集中进行硬度测量时，该传感器的决定系数（R）为0.878，均方根误差（RMSE）为0.732。对于接触力检测，测试集中的R为0.942，RMSE为1.115。结果表明，视觉触觉传感器对于桃子硬度和接触力的无损检测是可行的，在农业机器人领域具有广阔的应用前景。
