Moving object detection technology, which is always used in many aspects such as video classification, animal motion capture and human-computer interaction, is now the most common technology in computer vision and video information processing. Although the current moving object detection technology has been quite mature, due to environmental factors and a series of reasons, it still faces many challenges such as the impact of noise, camera shake and automatic adjustment, natural light changes, moving objects stagnation, background object movement, dynamic background and so on. In this paper, we use the convolution neural network to solve the problem of moving object detection, the proposed method will extract the feature of moving object through a lot of training. The model proposed in this paper is called 3D-CNN, it combines information of both time and space to extract feature which makes the model more rational. According to the experiment results, our 3D-CNN model achieves a higher accuracy than the most traditional methods.