Abstract: Three-dimensional spatial characterization of fracture systems at high resolution constitutes a critical technical challenge in oil and gas exploration and development, with direct implications for drilling operational safety. Taking Block X in the Pearl River Mouth Basin as the research object, this paper integrates imaging logging fracture interpretation and seismic multi-attribute analysis technologies, proposes a seismic attribute optimization and fusion method based on Partial Least Squares Regression (PLSR), and establishes a high-precision three-dimensional fracture density model. First, FMI imaging logging data are used to identify fracture occurrence characteristics, and five seismic attributes are extracted, including variance cube, ant body, chaos cube, curvature cube and fault distance cube. The PLSR algorithm is then adopted to solve the multicollinearity problem among attributes, establish a well-seismic mapping relationship, and quantitatively evaluate the contribution weight of each attribute to fracture development using the Variable Importance in Projection (VIP) index. The results show that the curvature attribute plays a dominant role in fracture prediction (VIP>1), with the coefficient of determination (R²) of the model reaching 0.925. A 3D fracture density volume is constructed by fusing multiple attributes based on weight coefficients, which clearly characterizes the spatial distribution pattern and heterogeneity of the fracture system in the study area. The research results provide an effective approach for the 3D spatial characterization of fractured reservoirs and engineering risk prediction.