Abstract: The feasibility was investigated for online detection of soluble solids content（SSC） of Gannan navel orange by visible-near infrared（visible-NIR） diffuse transmission spectroscopy coupled with least square support vector machine（LS-SVM） algorithm. 139 samples were divided into the calibration and prediction sets（103∶ 36）for developing calibration models and assessing their performance. The partial least square（PLS） regression and LS-SVM model were developed with the pretreatment by the combination of first derivative（1D），Smoothing and multiplicative scattering correction（MSC）. The new samples of prediction set were applied to evaluate the performance of the model. Compared with PLS model，the performance of LS-SVM model was better with the root mean square error of prediction（RMSEP） of 0.6423% and the correlation coefficient of prediction of 0.9059. And the spectral dimension reduction method of principal component analysis（PCA） and the kernel function of radial basis function（RBF） were suitable to improve the predictive ability of LS-SVM model. The results suggested that it was feasible for online detection of SSC of Gannan navel orange by visible-NIR diffuse transmission spectroscopy combined with LS-SVM algorithm.