This paper presents a new system, running on smartphone, to provide augmented reality in unconstrained context like an urban environment. Our approach is mainly based on interest points extraction and description. We use two new improvements of existing methods to increase the general performance. The first one allows optimization of the matching step with a BRIEF descriptor and a corner detector. The second one describes the functioning of a new tracker. We add other improvements to get more stable points especially in our urban environment and speed up the matching process between interest points. Thanks to these improvements, we were able to implement real time augmented reality on smartphones in urban environment. Demonstration videos are provided with the paper.