Apparatus Development For Detecting Anthocyanin Content In Eggplant Peel Based On Multi-spectral Technology

Eggplant is one of the most widely planted eggplant vegetables in China.Anthocyanins rich in eggplant pericarp have stable antioxidant activity,which can help eggplant better adapt to environmental stresses such as low temperature and drought,and anthocyanins are also important raw materials for natural anthocyanin extraction in industrial applications.The content of anthocyanin is an important indicator for breeding selection of eggplant,estimating the purity,estimating the nutritional value and estimating extraction value.At present,the detection of anthocyanins in eggplant peel is mainly based on wet chemical methods such as high performance liquid chromatography（HPLC）and p H difference method,which are time-consuming and laborious and require high professionalism of operators,thus these mathods cannot meet the demand for real-time,rapid and accurate detection.Therefore,this paper studied the detection method of anthocyanin content of eggplant peel based on multi-spectral technology,and developed a portable detection apparatus to achieve rapid and non-destructive detection of anthocyanins in eggplant peel.The main research contents and results are as follows:Firstly,the quantitative relationship between anthocyanins and spectral information of eggplant pericarp was explored,and the characteristic wavelengths of anthocyanins were extracted,and partial least squares regression（PLSR）prediction models were developed.Eighty-two purple-skinned dwarf eggplants from Hanzhong were used as experimental samples to collect spectral data in the 390～1050 nm band.The anthocyanin content of these eggplants in the peel was determined by the p H difference method.After eliminating four abnormal samples by Monte Carlo cross validation（MCCV）method,the anthocyanin characteristic wavelengths were extracted by Successive projections algorithm（SPA）,competitive adaptive reweighted sampling（CARS）,elimination of uninformative variables（UVE）,CARS-SPA and UVE-SPA.Respectively,the PLSR prediction models were constructed and validated for accuracy.Then,combining with regression the weights of each characteristic wavelength in its wavelength combination were analyzed by combining the model variable coefficients,and the three characteristic wavelengths with the highest contribution were selected:450 nm,490 nm and 530 nm.Secondly,The hardware system design of the detection device was completed based on multi-featured LEDs combined with photoelectric sensors.With STM32F103RCT6microcontroller as the core unit,the device mainly consists of six functional modules,including spectral detection module,processor module,power management module,data storage module,data transmission module and human-computer interaction module.Respectively,the diffuse reflection optical path of the detection probe is designed according to the mechanism of the interaction between the eggplant peel and the spectrum,and the experimental structure of the spectral detection probe with adjustable light-sensing distance is designed according to the demand of the probe parameter optimization.Then with the STM32 microcontroller as the core,combined with the modular function and peripheral interface location,the integrated PCB control boards were design and the detection device package structure design are completed.Thirdly,the software control system of the eggplant peel anthocyanin content detection device was developed.The initialization system,spectral data acquisition system,Fat Fs file management system,data transmission system and interrupt control system were designed based on the application programming interface provided by STM32F103 library functions.And a friendly human-computer interaction was realized based on the USART HMI graphical interface design software.In addition,the portability of the program was improved through the encapsulation of functional functions.Finally,the parameters of the detection device were optimized,the anthocyanin content prediction model was preferred and the accuracy verification was completed.The optimal driving current of LED light source was determined through the light source stability test.The best light-sensing distance was determined by reflectance testing based on the rack and pinion mechanism to realize the linear motion of the photoelectric sensor.The spectral data of 83 purple-skinned dwarf eggplants from Hanzhong and 71 purple-skinned round eggplants from Shandong were collected by the detection device.Then the multiple linear regression（MLR）,PLSR and support vector regression（SVR）calibration models for anthocyanin content were established.The determination coefficients of the calibration model with the purple-skinned dwarf eggplants samples from Hanzhong were 0.922 and0.872 for the calibration set and prediction set,respectively,the root mean square errors were0.096 g·kg^-1and 0.113 g·kg^-1,the ratio of standard deviation to root mean square error of prediction was 2.66.The determination coefficient of the calibration model with purple-skinned round eggplants samples from Shandong were 0.885 and 0.782 for the calibration set and prediction set,respectively,the root mean square error was 0.135 g·kg^-1and 0.146 g·kg^-1,the ratio of standard deviation to root mean square error of prediction was2.23.The results of this paper show that the MLR calibration model is the best model for portable detection devices with high prediction accuracy,and the detection device designed in this paper has met the actual detection needs was proved by the high accuracy and reliability of models,and this device has a certain degree of universality for different varieties of eggplant samples.