Population monitoring requires techniques that produce estimates with low bias and adequate precision. Distance sampling using ground-based thermal infrared imaging (ground imaging) and spotlight surveys is commonly used to estimate population densities of white-tailed deer (Odocoileus virginianus). These surveys are often conducted along roads, which may violate assumptions of distance sampling and result in density estimates that are biased high. Aerial vertical-looking infrared imaging (aerial imaging) is not restricted to roads and therefore enables random sampling and detection. We compared estimates of population density and precision, and evaluated potential sources of bias for these 3 techniques for deer on Arnold Air Force Base in Tennessee, USA, during January–February 2010. Using data from aerial imaging conducted along systematic strip transects, we found that deer were distributed close to roads and deer responded to the landscape along the road edge or to observers driving along roads. As a result of these distributional patterns, estimated deer density based on ground imaging and spotlighting from road-based surveys was 3.0–7.6 times greater than density estimated from strip transects using aerial imaging. Ground imaging did not produce better estimates than spotlighting. Observers on the ground counting all deer seen at test plots with hand-held thermal imagers saw fewer deer than were seen on aerial images, suggesting high detection of deer by aerial imaging. Despite its higher cost (US$10,000) over spotlight surveys, we recommend aerial imaging instead of road-based ground surveys for monitoring populations of deer and discourage the continued use of non-random road-based surveys as a method for estimating white-tailed deer populations. © 2014 The Wildlife Society.