Fine mode aerosol optical depth in Xianghe was calculated, using data from the onboard multi-angle polarization sensor POLDER-3 in the PARASOL satellite. The retrieval results were compared with the operational products of POLDER, MODIS, and AERONET data. The results show that the accuracy of POLDER using polarization remote sensing is significantly better than that of unpolarized MODIS. The correlation coefficient is increased from 0.67 to 0.93, and the average error is reduced from 0.32 to 0.15. As combined with the neural network (NN) method, the method gave correlation coefficient of 0.94 and standard deviation of only 0.11. Then, NN was applied to Hangzhou and Hong Kong, respectively. The verification results show that it has similar accuracy in Hangzhou but poor applicability in Hong Kong. The research shows it is feasible to use the NN to extract fine mode aerosol information from polarized signals.