Ground coverage (GC) is a simple and important trait to monitor crop growth and development, which can be easily captured by visual camera attached on the ground and aerial based platform. However, the accuracy of GC is determined by pixel size related to the size of green crops. Here, we explored the impacts of image resolution on the GC. High-resolution reference images were manually taken at 13 time points during the growing season of a wheat trial. A visual camera was vertically pointed downward at about 0.7 m above the top of canopy (Canon 550D). The wheat trial included 2 nitrogen (high and low nitrogen) and 2 irrigation (rain-fed and irrigation) treatments for 8 cultivars, and each treatment included three replicates. Two photos were taken in each plot to cover about 80% regions. In total, 2496 photos were captured during the whole season. For each reference photo, the original resolution was gradually degraded into several coarse resolutions (up to 5 cm/pixel, 27 levels in total) using cubic interpolation algorithm. The serial image resolutions were used to mimic the image acquisition at different flight heights of unmanned aerial vehicle (UAV) based platform. A machine learning algorithm was used to calculate GC based on the same training dataset. Compared with reference images, GCs were underestimated and overestimated when GCs were less and more than 0.5, respectively, but depending on the growth stage of wheat (mainly leaf width). The optimum image resolution (i.e. flight height) was explored for UAV-based platform to estimate GC.