The work aims to propose a selection method of Chinese wolfberry in the multi-feature combination of machine vision technology, in order to improve the separation efficiency of Chinese wolfberry and enable the sorting result to be convenient for uniform packaging. The shape and plumpness of Chinese wolfberry were two important features and basis of selection. Firstly, through digital image processing technology, the transverse diameter and longitudinal diameter of the wolfberry image were obtained and regarded as the shape characteristic parameters; then, the ratio of the quality of Chinese wolfberry to its image area was obtained and regarded as its plumpness characteristic parameter. Based on the clustering analysis of these two kinds of characteristic parameters, the multi-feature selection classifier of Chinese wolfberry was established by BP neural network. When the shape type of Chinese wolfberry was different, its shape characteristic and plumpness characteristic value were different. The BP neural network classifier displayed the Chinese wolfberry cluster of different shape types. Such method could effectively select the shape type of Chinese wolfberry and the selection accuracy was up to 97.6%. The selection effect is the best when the shape and plumpness characteristics are combined. Such method can realize the selection of the shape type of Chinese wolfberry.