Abstract: Reservoir leakage evaluation is one of the core contents of geological survey for hydraulic engineering projects, directly determining the reservoir formation conditions, engineering construction plans, and subsequent operational safety. Addressing the deficiencies of traditional reservoir leakage evaluations, such as a single evaluation carrier, fragmented multi-source survey data, deviations in interpretation results from different survey methods, and a lack of cross-checking of survey results, this paper takes a reservoir project in Southwest China as the research object and constructs a set of three-dimensional visual leakage evaluation technology processes based on multi-source data fusion. The system collects data results such as geological mapping of the reservoir area, borehole core logging, rock mass permeability coefficient, and stable groundwater level in boreholes, while simultaneously integrating geophysical exploration survey data such as high-density electrical method, natural source surface wave method, and induced polarization method. Relying on MATLAB programs, it completes standardized correction, spatial registration, outlier removal, and normalization processing of discrete data for multi-source survey data, achieving collaborative mutual checking and joint interpretation of multi-source survey information. A refined three-dimensional geological model is built, which can characterize the stratum lithology, groundwater occurrence patterns, and spatial characteristics of rock mass permeability in the reservoir area. Based on this model, visual spatial analysis and comprehensive research and judgment of leakage characteristics are carried out. The results indicate that a dual-layer groundwater system is developed in the reservoir area, with a continuous leakage channel at the reservoir bottom discharging to the lower adjacent valley and downstream. The proposed dam site valley is a suspended valley, lacking the geological conditions for reservoir formation. Engineering case verification shows that the modeling and evaluation approach constructed in this paper can accurately depict the geological structure of the reservoir area, groundwater distribution, and spatial distribution characteristics of unfavorable geological bodies, achieving intuitive identification and quantitative analysis of leakage channels and hydrogeological key parameters. This study effectively supplements the analytical methods of traditional leakage evaluations, further enhancing the scientificity and reliability of evaluation results, and improving the modeling implementation path for multi-source geological data fusion in hydraulic engineering projects. The research results can provide technical references for leakage analysis and reservoir formation condition research and judgment in similar complex reservoir areas, and have certain theoretical and technical reference significance for promoting the digitalization and refinement transformation of geological survey in hydraulic engineering projects.