Application of Successive Projection Algorithm in Nondestructive Detection of Soluble Solid Content in Apple

Soluble solid content of apple is one of the important indexes to evaluate taste of apple. In order to eliminate collinearity between original spectral variables, reduce calculation of modeling variables, and improve correction speed and modeling accuracy, this paper applies successive projection algorithm (SPA) to the establishment of near-infrared correction model of soluble solid content of apple. The sample set partitioning based on be x-y distances method (SPXY) selected representative of the calibrating samples, then it is processed by window smoothing method and standard normal variate transformation method (SNV), and variable selection is conducted by SPA. 19 optimal characteristic variables were selected for modeling, and the root mean square error of prediction and predictive correlation coefficient were 0.4071 and 0.9351 respectively. The results show that the continuous projection algorithm can effectively improve the correction speed, and this algorithm is effective and feasible.