Abstract: Machine vision technology was used to evaluate inner quality of tomato fruits qualitatively and quantitatively. Sixty-eight tomato samples were collected with different inner quality. A multifunctional camera system was developed to take the tomato images. Four halogen lamps were used as lighting resource and the illuminance of the camera system was about 600 lx. The camera was set in three heights, 0.5 m, 1 m, and 1.5 m, and six directions, top, bottom, left, right, front, and back. The features of the images from RGB color model, L*a*b* color model, and gray level co-occurrence matrix were calculated. In quantitative analysis, four important indexes of tomato inner quality, sugar content, acid content, amino acid content, and water content, were selected for prediction by machine vision technology. The correlations between each feature of the images and each index of inner quality were investigated and the estimation models of all four indexes were established by the features of the images with BP neural network. The correlation coefficient observed between acid content and image features was 0.536. The results showed a possibility of using image features to predict the acid content of tomato fruit. However, no significant correlations were observed between other indexes and the image features. In qualitative analysis, all tomato samples were divided into five groups based on inner quality, and then the classification and identification were conducted by the features of the images with BP neural network, too. The effect of two important model parameters, hidden node and training function, on the precision of the network was analyzed and finally optimal model parameters were determined. Twenty-eight samples were used as validation group to check the model of classification. Twenty-two samples were identified correctly. The results show the prospect to use machine vision to identify inner quality of tomato fruits.