In this study, entrainment monitoring algorithms were developed, trained and implemented on the batch flotation of three synthetic mixtures of galena and quartz with different particle size ranges for the quartz mineral. An online image-based soft sensor framework was developed to estimate product grade and recovery using support vector regression. A dynamic fundamental model was developed with emphasis on the entrainment and drainage sub-processes. The model was reconciled with online soft sensor measurements and was updated in real-time by estimating the states and parameters using an extended Kalman filter. Along with the online measurements of quartz entrainment recovery, measurements of entrainment and the true flotation contribution for galena particles were obtained in real-time. The proposed monitoring framework was shown to be effective in monitoring entrainment and the grade and recovery of the desired minerals.