We present a cadence optimization strategy to unveil a large population of kilonovae using optical imaging alone. These transients are generated during binary neutron star and potentially neutron star-black hole mergers and are electromagnetic counterparts to gravitational-wave signals detectable in nearby events with Advanced LIGO, Advanced Virgo, and other interferometers that will come online in the near future. Discovering a large population of kilonovae will allow us to determine how heavy element production varies with the intrinsic parameters of the merger and across cosmic time. The rate of binary neutron star mergers is still uncertain, but only few (less than 15) events with associated kilonovae may be detectable per year within the horizon of next-generation ground-based interferometers. The rapid evolution (hours to days) at optical/infrared wavelengths, relatively low luminosity, and the low volumetric rate of kilonovae makes their discovery difficult, especially during blind surveys of the sky. We propose future large surveys to adopt a rolling cadence in which g-i observations are taken nightly for blocks of 10 consecutive nights. With the current baseline2018a cadence designed for the Large Synoptic Survey Telescope (LSST), less than 7.5 poorly-sampled kilonovae are expected to be detected in both the Wide Fast Deep (WFD) and Deep Drilling Fields (DDF) surveys per year, under optimistic assumptions on their rate, duration, and luminosity. We estimate the proposed strategy to return up to about 272 GW170817-like kilonovae throughout the LSST WFD survey, discovered independently from gravitational-wave triggers.