There are few researches on architectural population planning in urban ecologically sensitive areas, and the planning effect is poor. Therefore, a high-dimensional spatial planning method of architectural colony in urban ecologically sensitive areas is put forward. Based on the multi-scale feature, the optimized building group image is segmented by using rectangular fitting optimization method to process the architectural group image in the ecologically sensitive area. Based on the segmented image, the building group data is converted into high dimensional spatial data, the building planning target model is established, and the particle swarm optimization algorithm is used to solve the target model, and the global optimal value of the target model is obtained. To complete the high-dimensional space of the urban ecological sensitive area. Planning. The experimental results show that the building image fitting error of the proposed method is less than 2.9%, the fitting accuracy is high, and the segmentation effect is better than the traditional method, and the searching time of the optimal value of the target model is less than 40% of the traditional method. The result of planning effect evaluation is good.