Abstract: By analysis of different hues of fruit, and according to fractal dimension of chroma distribution on superficies of fruit, the classification of fruits was carried out. The eigenvalues including both the accumulative characteristics and space distribution characteristics of various hue points make the hue classification corresponding to the practical situation. The identifying model of artificial neuron network was built when the fractal dimension in every hue area was considered as pattern and then was processed. After studying, the accuracy of classification reached 95%, which can meet the requirement of real time production line for the classification of fruits by computer vision.