Land carrying capacity is an important indicator of the regional population, resources and environment evaluation of sustainable development. In this study, based on fully considering the influencing factors of land resources carrying capacity, land resources carrying capacity evaluation indicator system is constructed, composed of four sub target, i.e. the utilization of land resources, social development, economic and technology and the ecological environment. Real-coded accelerating genetic algorithm is used to optimize projection pursuit model to carry on comprehensive evaluation of land resources carrying capacity under different scales. Results show that national resources carrying capacity level of large-scale land falls under grade III under national scale; Heilongjiang Province and Sanjiang Plain belong to grade II under medium scale; The city of Jixi, Hegang, Shuangyashan and Jiamusi fall on grade I and the city of Qitai River, Muling and Yilan belong to grade II under small scale. The measures of effective utilization of land resources can be further proposed combined with the regional land resource evaluation results.