With the rapid development of remote sensing technology and computer technology, the quality of geographic image(GI) obtained by various remote sensing image acquisition methods is getting higher and higher, and the content of image expression is more and more complex. Degree requirements are getting higher and higher. In the GI obtained by remote sensing equipment, it can be seen that the global features of urban areas have a high degree of similarity. Most of the current image retrieval methods are based on global features, and a large number of repeated streets, residential areas, and commercial areas in urban GI make it difficult to It is difficult to retrieve regional information for geographic location positioning through traditional methods. Therefore, in this paper, a GI feature retrieval method based on spiking neural network(SNN) model, after comparing the effectiveness of SNN algorithm and LM-FNN, FNN, RNN and other neural network algorithms, it is verified that spiking neurons can optimize network performance. The experiments of GI texture feature extraction and local feature retrieval show that the proposed algorithm is more feasible in GI feature analysis.