Gamma-rays are very important for probing high energy particle accelerators and to study the nature of dark matter. The DArk Matter Particle Explorer (DAMPE), launched on 17th December, 2015, is designed for high energy cosmic-ray and gamma-ray observatory. Gamma-rays constitute only a tiny fraction of all the particles that impinge on our detector, it is thus critical to reject a large amount of ‘background’ to get a clean gamma-ray event sample. Here we present an algorithm to identify gamma-rays from the remaining background, using a silicon-tungsten tracker converter which promotes pair conversion and measures the directions of incident particles, and a segmented anti- coincidence detector made from plastic scintillator which measures the charge of particles. We utilize a GEANT4-based Monte Carlo simulation to validate our selection algorithm and estimate its efficiency. This method has been applied to the flight data, and a preliminary clean GeV gamma-ray sky-map has been obtained.