Abstract During storage, the sweetness of kiwifruit increases and its firmness decreases quickly. Since the sweetness is linked to consumers’ taste preference, and the firmness is regarded as the primary indicator of kiwifruit eating ripeness, detecting the sweetness and firmness are welcome by kiwifruit sellers, consumers and processors. To make the detected results intuitive and acceptable to more people, this study aimed to present a portable, low-cost, sensor-based grade detector on sweetness and firmness of kiwifruit. The detector consisted of a control/processing unit, a LED panel and driver unit, a light signal detection and amplifier unit, an input/output unit and a battery. Micro-controller of STM32 was used to do data collection and processing. The LED panel with nine LEDs at peak wavelengths of 800, 810, 850, 880, 900, 940, 970, 1000 and 1100 nm were used as light source. The reflected light from the kiwifruit was sensed by a Si PIN photodiode. The software of the detector was programmed by using C language. Kiwifruit varieties of 'Xuxiang' and 'Huayou' were used as samples, and their sweetness and firmness were divided into four grades. The model of support vector machine was established to determine kiwifruit grades. The validation experiment showed that the identification accuracies of the grade detector were 82.0% for sweetness and 74.0% for firmness of ‘Xuxiang’ variety, and 85.5% and 91.5% for sweetness and firmness of ‘Huayou’ variety, respectively. The detection time for one sample was less than 3 s.

中文翻译：

---

一种便携式、低成本、基于传感器的猕猴桃甜度和硬度等级检测器

摘要 在贮藏过程中，猕猴桃的甜度增加，硬度迅速下降。由于甜度与消费者的口味偏好有关，而硬度被视为猕猴桃食用成熟度的主要指标，因此检测甜度和硬度受到猕猴桃销售商、消费者和加工商的欢迎。为了使检测结果更直观并为更多人所接受，本研究旨在提出一种便携式、低成本、基于传感器的猕猴桃甜度和硬度等级检测器。探测器由控制/处理单元、LED面板和驱动单元、光信号检测和放大单元、输入/输出单元和电池组成。STM32单片机用于数据采集和处理。带有 9 个 LED 的 LED 面板，峰值波长分别为 800、810、850、880、900、940、970、1000 和 1100 nm 用作光源。奇异果的反射光由 Si PIN 光电二极管检测。探测器软件采用C语言编写。以'旭香'和'花油'猕猴桃品种为样品，将其甜度和硬度分为四个等级。建立支持向量机模型确定猕猴桃等级。验证实验表明，分级检测仪对'徐香'品种甜度和硬度的鉴定准确率为82.0%，对'花油'品种的甜度和硬度分别为85.5%和91.5%。一个样品的检测时间小于3 s。探测器软件采用C语言编写。以'旭香'和'花油'猕猴桃品种为样品，将其甜度和硬度分为四个等级。建立支持向量机模型确定猕猴桃等级。验证实验表明，分级检测仪对'徐香'品种甜度和硬度的鉴定准确率为82.0%，对'花油'品种的甜度和硬度分别为85.5%和91.5%。一个样品的检测时间小于3 s。探测器软件采用C语言编写。以'旭香'和'花油'猕猴桃品种为样品，将其甜度和硬度分为四个等级。建立支持向量机模型确定猕猴桃等级。验证实验表明，分级检测仪对'徐香'品种甜度和硬度的鉴定准确率为82.0%，对'花油'品种的甜度和硬度分别为85.5%和91.5%。一个样品的检测时间小于3 s。验证实验表明，分级检测仪对'徐香'品种甜度和硬度的鉴定准确率为82.0%，对'花油'品种的甜度和硬度分别为85.5%和91.5%。一个样品的检测时间小于3 s。验证实验表明，分级检测仪对'徐香'品种甜度和硬度的鉴定准确率为82.0%，对'花油'品种的甜度和硬度分别为85.5%和91.5%。一个样品的检测时间小于3 s。