Natural gas gathering station plays an important role in providing stable gas supply for downstream users. With the trend of the unattended station, the safety of the station becomes a major challenge. Intelligent video surveillance system, as an important part of unattended system, is mainly responsible for monitoring suspicious intruders. Target tracking technology is the key technology to achieve this goal. Off-line tracker is one of the main ways to achieve target tracking tasks. However, it requires a large number of samples for pre-training. Considering that human is the main monitoring object in intelligent video surveillance system, facial features can provide an effective basis for target tracking. Therefore, we introduce information maximizing generative adversarial nets as a generative model, which takes CelebA dataset as a benchmark to generate a large number of face samples with continuous change of feature attributes. Qualitative evaluation shows that the quality of synthetic face samples is high, which provides reliable data support for further training off-line tracker.