Camera trap is an image sensor that is widely used in monitoring biodiversity, identifying and tracking species in natural life. In this study, we investigate human-animal distinction in image dataset obtained from camera traps for the purpose of smuggling detection and prevention. The dataset includes human and animal images capturing during both night and day light hours. In the preprocessing stage, the objects are firstly cropped from the background. Then Scale Invariant Feature Transform (SIFT), Color Histogram, Local Binary Pattern (LBP) and Histogram of Oriented Gradient (HOG) descriptors are extracted from these cropped images. Support Vector Machine (SVM), k-NN and random forest algorithms are used to classify the data in two class as human and animal. The experiments are conducted on different type of dataset such that original dataset are separated by images captured in night and day light. The other one is obtainted by dividing dataset randomly as equal number of human and animal images. The experimental results show that color histogram features on random forest algorithm give always best accuracy results for all dataset. Moreover, the images captured in night give more accuracy than the images captured in day light for all classification algorithms.