Using camera traps can provide valuable information for estimating wildlife density, and further contributes to conservation activities. For species that can be individually recognized, such as tigers and leopards, camera trapping combined with mark-recapture techniques can provide reliable estimates of population density. However, most species cannot be individually recognized, and no sophisticated models are available for accurate estimation of their densities. At present, the only available model is Rowcliffe et al.'s gas molecule movement model, which assumes animals behave like ideal gas particles, moving randomly and independently of one another. Such a model is not appropriate for either territorial or social species, or elusive species that usually move along trails. We developed a novel method to estimate the population density of animals that cannot be individually recognized. This method is based on the simulation of animal movement and pseudo camera trapping processes at a series of population densities. We matched results of real camera trapping with those of simulated camera trapping to estimate population density. The method was coded using the R language. We deployed 25 cameras (LTL5120) in the 25 hm² forest dynamics plot in the Changbaishan National Nature Reserve, China for 41 days in the winter of 2011 and 40 days in the winter of 2012. The Siberian chipmunk (Eutamias sibiricus)and Korean field mouse (Apodemus peninsulae) are two dominant species in the plot. Animal movement was simulated by setting a starting location and a series of moving directions and step lengths. The starting location was a randomly selected point in the survey area. The direction of the first movement, θ, was also randomly selected from a range of 0-2π. The length and angle of deflection of subsequent movements followed normal distributions N (1 m, 0.1 m) and N (0, 30 degrees), respectively. We also defined a home range for each species by forcing the simulated animal to return to the starting location (assumed to be a mouse hole) at a rate of D/50, where D is the distance in meters between the current and original locations. The simulations of animal movement were run under a series of population densities. We matched the simulated results and the observed photo records using the random forest algorithm to estimate the population density and its confidence intervals. This analysis determined that the density of the Siberian chipmunk is 1.96hm², and 2.71hm² for the Korean field mouse. Our method has a number of limitations. First, the movement pattern of the target species must be known. In this study, we selected movement parameters (step length, angle of deflection, home range size, etc.) by visually checking the simulated footprint chains, which should be replaced by field tracking. Second, the cameras must be deployed in the field systematically, at regular intervals, so that virtual camera trapping can be simulated accordingly. In spite of the limitations, this method can provide reliable estimates for population density for animals that cannot be individually recognized. Our new method can be used for other camera trapping practices, as long as the movement pattern of the species is known.