Long-term exposure to PM_2.5 can lead to great adverse health effect on human health. To better guide public policies that aim to reduce PM_2.5 population exposure, this work combined multi-source data to realize high-resolution PM_2.5 exposure risk assessment in Nanchang urban region. The land use regression (LUR) model was used to simulate the seasonal-spatial variations of PM_2.5 concentrations at 100-m resolution, and building information extracted from IKONOS image was applied to spatialize population at 100-m resolution. An improved piece-wise population exposure approach was introduced to evaluate the exposure risk, and results were compared with two classical approaches. In all seasons, results by the absolute concentration approach are very different from the other two, showing obvious spatial smoothing effect. Results by population-weighted and piece-wise exposure approaches are similar in spring and autumn, and different in summer and winter. In winter, the area and population percentages divided to severity level 7 by population-weighted exposure approach are 5.21% and 2.35% lower than that by piece-wise exposure approach. When in summer, the area and population percentages divided to severity level 7 by population-weighted exposure approach are 6.77% and 24.79% higher than that by piece-wise exposure approach. The absolute concentration approach is disadvantageous for the identification of high-risk areas, the population-weighted exposure approach would underestimate or overestimate the population exposure when air is seriously polluted or remarkably clean, and the proposed piece-wise exposure approach would be more reasonable. The integrated methodology is effective in exposure risk assessment and can be applied to other regions and pollutants.