Sugar beet is the second biggest world contributor to sugar production and the only one grown in Europe. Great efforts are currently being made to preserve and enhance its competitiveness. Within this context, one of the main bottlenecks identified is the lack of effective tools for assessing sugar content in unprocessed sugar beet roots. In the sugar industry, NIRS has been long used to monitor product quality on sugar processing lines in both sugar beet and cane industrial processes. The main work have been performed on beet brei although some authors studied the feasability of predicting sucrose content of intact and sliced beets. The results are of very good quality, however the beet peel is still a real problem to acquire signals of sufficient quality. Optical fibers are an essential component for remote optical measurements and have been extensively used to deliver and collect light in optical setups. By invasive but non destructive measurement, the fiber-optic probe could be an effective tools for rapid assessment of sugar content of intact beets and deal with the problem caused by the presence of beet peel. The objective of this work is to evaluate the use of a self-designed fiber-optic probe for the estimation of Soluble Solid Content (SSC) in intact sugar beets. Two geometries were tested : a single-fiber probe and a multiple-fiber probe. The impact of different components of the probe were analyzed, such as fiber diameter, distance between illumination and collection fiber in order to optimize optical setup of the probe. Total reflectance over the [400-2500 nm] spectral range, of 180 sugar beet samples, were measured with the two probes. Immediately after spectrum acquisition, Soluble Solids Content measurements were achieved for each sample. Partial Least Square (PLS) algorithm was used to calibrate the SSC estimation of the sugar beets With the single-fiber probes and for the two tested diameters, the results were not satisfactory. The better correlation coefficient and the ratio of sample standard deviation to standard error of prediction are respectively equal to 0.757 and 2.02. For this configuration, the measured signal is mainly composed by the specularly reflected light, which overlaps and therefore masks the information related to chemical absorbance of SSC. On the contrary, Partial Least Squares regression (PLSr) models showed good performance in the estimation of SSC with the multiple-fiber probes. For two different positions of the collecting fiber, results showed good correlation between SSC and reflectance spectra for intact sugar beets, with a correlation coefficient higher than 0.90 and a ratio of sample standard deviation to standard error of prediction higher than 3.30. These first results are very promising. However, further improvement are needed in order to achieve a satisfactory accuracy: •optimization of the distance between emitting and collecting fiber, •reducing the multiple scattering using several preatraitements, •increasing the number of samples in order to increase the robustness of the models.