Ultrasound is progressing toward becoming an affordable and versatile solution tomedical imaging. In recent years, the need for a fully autonomous ultrasound system has been accelerated due to its labor-intensive nature and the advent of COVID-19 global pandemic. In this work, we tackle the important yet seldom-studied problem of scan target localization, under the setting of lung ultrasound imaging. We propose a purely vision-based, data driven method that incorporates learning-based computer vision techniques. We combined a human pose estimation model with a specially designed interpolation model to predict the lung US scan targets, while multi-view stereo vision is deployed to enhance the accuracy of 3D target localization. We collected data from 30 human subjects for testing, and obtained satisfactory result from test experiments, achieving a success rate above 80% for all scan targets under an error threshold of 25mm. Finally, our approach can serve as a general solution to other types of US scan, with many potential improvements in terms of model complexity and runtime. The code is available at this url.