Abstract：
Autonomous field robots are being developed for picking of fruit, where each fruit needs to be individually graded and handled. There is therefore a need for rapid and non-destructive sensing to measure critical fruit quality parameters. In this article we report how total soluble solids (TSS), a measure for total sugar content, can be measured in strawberries in the field by non-contact near-infrared (NIR) interaction spectroscopy. A specially designed prototype system working in the wavelength range 760-1080 nm was tested for this purpose. This novel instrument was compared with a commercial handheld NIR reflection instrument working in the range 900-1600 nm. The instruments were calibrated in the lab using data collected from 200 strawberries of two varieties and tested in a strawberry field on 50 berries in 2022 and 100 berries in 2023. Both systems performed well during calibration with root mean square errors of cross validation for TSS around 0.49 % and 0.57 %, for interaction and reflection, respectively. For prediction of TSS in new berries in 2023, the interaction system was superior, with a prediction error of 1.0 % versus 8.1 % for the reflection system, most likely because interaction probes deeper into the berries. The results suggest that interaction measurements of average TSS are more robust and would most likely require less calibration maintenance compared to reflection measurements. The non-contact feature is important since it reduces the spread of diseases and physical damage to the berries.
摘要：
正在开发用于采摘水果的自主野外机器人，每个水果都需要单独分级和处理。因此,需要快速且非破坏性的感测来测量关键的水果品质参数。在本文中，我们报告了总可溶性固体(TSS)测量总糖含量，可以通过非接触式近红外（NIR）相互作用光谱法在田间的草莓中进行测量。为此，测试了在760-1080nm波长范围内工作的专门设计的原型系统。将这种新颖的仪器与在900-1600nm范围内工作的商业手持式NIR反射仪器进行了比较。这些仪器在实验室中使用从两个品种的200个草莓中收集的数据进行了校准，并在2022年的50个浆果和2023年的100个浆果上在草莓田中进行了测试。两种系统在校准期间表现良好，TSS交叉验证的均方根误差约为0.49%和0.57%，为了互动和反思，分别。对于2023年新浆果中的TSS预测，相互作用系统优越，反射系统的预测误差为1.0%，而反射系统的预测误差为8.1%，很可能是因为相互作用深入浆果。结果表明，与反射测量相比，平均TSS的相互作用测量更可靠，并且很可能需要更少的校准维护。非接触功能很重要，因为它可以减少疾病的传播和对浆果的物理损伤。
