Abstract: In order to understand the composition of fish species diversity and assemblage structure in the Dianpu river, Chuanyang river, Dazhi river, Jinhui river and Longquan river these 5 Shanghai backbone rivers, this paper carried out water collection, eDNA filtration and extraction, 12SrRNA genetic marker amplification, high-throughput sequencing, bioinformatic analysis and other analysis processes in June (summer) and November (autumn) of 2021.The results showed that a total of 80 fish species were detected in 20 sections of 5 backbone rivers during the two surveys. Excluding 10 marine species that were detected due to kitchen waste flowing into rivers, 70 species might actually inhabit, belonging to 10 orders, 22 families and 55 genera. Including 1 species of second-class state protected fish-Myxocyprinus asiaticus and 5 invasive species. 48 and 62 fish species were detected in summer and autumn respectively, with an average of 35.2 and 44.0 species per river. Howerver, Chao1 index and ACE index of 20 sections in summer were higher than the average in autumn, while the mean values of Shannon index and Simpson index in summer were lower than those in autumn. The common dominant species in summer and autumn were Coilia nasus, Carassius auratus and Cyprinus carpio. Jaccard similarity analysis showed that Longquan river with the best similarity between sections was very similar, but the Dazhi river with the worst similarity was also medium similar. Cluster analysis and NMDS sequencing analysis showed that the 5 backbone rivers could be divided into 3 groups with extremely significant discrepancies. The Longquan river, Jinhui river and Dazhi river sections excluding Z1 were group Ⅰ, the sections of  the Chuanyang river and Dianpu river were Ⅱ, and Z1 section was Group Ⅲ. Coilia nasus, Carassius auratus and Acheilognathus macropterus were the main discriminating species causing the differences among the three groups. Although the number of fish species detected in this study was higher than that recorded in the main stream of the Huangpu River, food waste flowing into the river could also be detected, which might have some “false positive” phenomenon. Nevertheless, the eDNA matabarcoding technology still shows good application prospects in the assessment of fish diversity in urban rivers.