红外相机技术已在野生动物调查中得到了广泛应用。然而，红外相机所监测的许多物种无法进行个体识别，因而不能采用标记重捕模型计算这些物种的密度；如何准确估计这些物种的种群密度成为限制红外相机应用的关键问题。本项目研究影响密度估计的若干关键问题，并建立基于模拟个体运动和红外相机监测相匹配的方法来估计种群密度。我们以长白山自然保护区为研究基地，通过对重要动物进行GPS项圈遥测和足迹链跟踪，确定这些动物的运动参数和领域大小；然后根据每个物种的运动模式，模拟动物个体的运动过程，记录在不同种群密度时虚拟相机拍摄的照片数；最后利用机器学习（随机森林）的算法建立种群密度与照片数量的回归关系，进而匹配实际的红外相机监测数据计算动物种群密度和置信区间。我们将设计R语言软件包完成上述分析并给出种群密度。该方法是一个全新的基于红外相机数据的动物种群密度估计方法，将显著提高对不可个体识别物种种群密度估计的准确性。

Camera trapping has been extensively used for wildlife survey. However, many species that are surveyed can’t be individually recognized, so that the capture-recapture model can't be applied for estimating population density of these species. How to estimate population density of species that can't be individually recognized has become the key factor limiting the advantage of camera trapping. This project aims to study on the key questions about how to estimate population density of these species, and plan to develop a new method for population density estimation by simulating animal movement and virtual camera trapping. We will conduct GPS collar tracking and footprint tracking in Changbaishan Nature Reserve, inder to obtain the movement parameters for the wild boar, roe deer, and saber, et al. The parameters include step length, step bearing, and their variation, as well as the size of the territory. We will simulate the animal movement in a series of densities in a virtual plot with a number of virtual cameras. We will use a machine learning algorithm, random forest, to quantify the relationship between population density and the number of pictures that are taken during the simulation. Therefore we can match the simulated and real camera trapping data to estimate the animal population density and its confident interval. We will develop a R package so as to publish our model to the public. The model can significantly improve the accuracy of population density estimation for the species that can't be individually recognized, and further contribute to wildlife monitoring and management.