This paper deals with the design of a model-based soft-sensor to improve the process monitoring and control in industrial sugar crystallization. This soft-sensor is based on an original model dedicated to the last stage of crystallization, avoiding the solving of the population balance. Additional information like the mass of crystals in the solution, the concentration of dissolved sucrose and the purity are relevant to improve the manufacturing process. As these physicochemical properties are not measurable on-line, a model based soft-sensor is developed. The effectiveness of the soft sensor is demonstrated using real plant data from an industrial crystallization process.