In response to the significant color difference between pothole repair coatings and original asphalt pavement surfaces, this study employed waterborne acrylate-modified polyurethane (WPUA) as the matrix, with carbon black, titanium dioxide, and yellow iron oxide as pigments. Through single-factor, two-factor, and ternary system experiments, the regulation mechanisms of multiple pigments on the CIELab color space were systematically investigated. Based on the ternary system data, a Lab dataset covering typical pavement color intervals was constructed, and the orthogonal regulation characteristics of the pigments and the selective absorption mechanism of yellow iron oxide were analyzed. A response surface methodology (RSM) model was established as an interpretable quadratic polynomial model to reveal the lightness regulation and interaction patterns of carbon black and titanium dioxide. Subsequently, a BP neural network was constructed for high-precision inverse prediction. The average chromatic aberration (ΔE) of the test set decreased from 3.42 (RSM) to 1.86 (BP), and the acceptability rate (ΔE < 2.0) improved from 42% to 78%. The established RSM-BP complementary dual-model system enables rapid inverse design from target Lab values to pigment ratios, effectively reducing visual differences between repaired and original pavement surfaces and providing a data-driven intelligent color matching method for refined asphalt pavement maintenance.