Abstract: In order to solve the problem of poor objectivity of index weight assignment for land resources carrying capacity evaluation, the projection pursuit evaluation model optimized by particle swarm optimization algorithm was proposed and applied to evaluate the land resources carrying capacity of Sanjiang plain in 2008. The results showed that the model was feasible for effectively avoiding human disturbance in index weight assignment and obtaining the evaluation results in agreement with facts. The evaluation results indicated that the comprehensive evaluation value of the land resources carrying capacity grade of the whole Sanjiang plain was 2.621 in 2008, and the land resources utilization had strong carrying capacity for social development and environment protection but had low guarantee degree for economic and technical development. The land resources carrying capacity of the cities and counties showed significant differences in Sanjiang plain. Thus scientific programs for land use and socioeconomic development ought to be made according to the regional resources characteristics.