Foreground segmentation (FS) plays a fundamental and important role in computer vision. In recent years, scholars have put forward many effective methods. However, these methods are limited in low illumination scenes. In order to improve the performance of FS in low illumination scene, a simple but effective method is proposed via feature extraction and RGB-D camera. Firstly, the coined Iterated Robust CUR (IRCUR) is used to get candidate foreground for depth image sequence. At the same time, the RGB image sequence is segmented using the simple linear iterative clustering (SLIC). Then, feature extraction is performed on the candidate matrix region corresponding to the super-pixel block. Then, the neural network is trained by using the acquired super-pixel features. Experiments show that the average F-measure value of this method is 35% higher than that of other methods only based on RGB images.