Estimating maturity by measuring pH, sugar, dry matter, water and vitamin C content of cashew apple (Anacardium occidentale) from remote spectral reflectance data using neural network

Abstract

In agricultural sector, maturity is the main decision criterion for starting the harvest. This criterion is usually revealed by a number of parameters such as pH, sugar, dry matter, water and vitamin C, which are informative but technically tedious to measure. The cashew apple is the hypertrophied peduncle which is attached to the cashew nut. It is a nutritious (very juicy fruit (85 to 90% water), sweet (7 to 13% carbohydrates), acidic and vitamin C content) fruit with high therapeutic and medicinal properties. The cashew apple is used as a raw material for many industrial applications (juice and alcohol). This research was conducted as a preliminary step towards the development of a real-time remote sensing technique for assessing the quality of tropical fruits. Spectral acquisitions were carried out from intact cashew apple using optical system composed reflector coupled with spectrometer USB 4000 FL from Ocean Optics (350-1100 nm). Immediately after spectral acquisition, the samples were analyzed by using chemical methods (sugar content, dry matter content, water content, vitamin C and pH). Preprocessing treatment method, bootstrap method was required to create statistical new samples and to increase the number of samples required. This method was used to improve the predictive performance of calibration model. Statistical models of prediction were developed using an artificial neural network (ANN) method. The results obtained from the models built by ANN showed strong relationships between predicted and experimental values: (Rsquare = 0.9870, RMSE= 0.0262) for pH, (Rsquare=0.9869, RMSE=0.1392) for Sugar, (Rsquare=0.9726, RMSE=0.3333) for water content, (Rsquare=0.9703, RMSE=0.3464) for vitamin C and (Rsquare=0.9922, RMSE= 5.0304, RMSE=5.0304) for dry matter. These results confirm the potential of visible spectroscopy to predict quality parameters of cashew apples remotely and make decisions about best harvest time