This paper presents a real time monocular EKF SLAM process that uses only Cartesian defined landmarks. This representation is easy to handle , light and consequently fast. However , it is prone to linearization errors which can cause the filter to diverge. Here , we will first clearly identify and explain when those problems take place. Then , a solution , able to reduce or avoid the errors involved by the linearization process , will be proposed. Combined with an EKF , our method uses resources parsimoniously by conserving landmarks for a long period of time without requiring many points to be efficient. Our solution is based on a method to properly compute the projection of a 3D uncertainty into the image frame in order to track landmarks efficiently. The second part of this solution relies on a correction of the Kalman gain that reduces the impact of the update when it is incoherent. This approach was applied to a real data set presenting difficult conditions such as severe distortions , reflections , blur or sunshine to illustrate its robustness. I. INTRODUCTION In order to be autonomous , a vehicle must be able to localize itself in an unknown environment. This problem , known as the Simultaneous Localization And Mapping (SLAM) problem , has been extensively studied throughout the last two decades [ 11 ]. Though mathematical solutions have been proposed [ 10 ] , they are usually not sufficient in real world applications. Moreover , they mostly consider range and bearing sensors [ 2 ]. Laser Range Finders (LRFs) , for instance , are expensive and rarely furnish enough information to efficiently track a feature. These two constraints are essential and must be avoided in order to develop and spread the use of Intelligent Vehicles. Conversely to LRFs , cameras are cheap and rich in terms of information. However , they add other constraints. First , cameras are bearing-only sensors which means that the depth of a landmark cannot be accurately estimated without having several observations with a sufficient parallax. Then , environmental factors can degrade the image quality : blur while turning , distortions , reflections or sunshine. These effects appear a lot especially in urban environments which are information rich. This is why a system must be tested under difficult conditions in order to prove its capabilities in real life applications. To do so , it is necessary to take into account the way features are initialized and handled within the state vector. Indeed , with difficult conditions , it is obvious that landmarks will be discarded quickly. With a bearing-only sensor , it is an major issue as several observations of the same landmark are needed to have an accurate estimate of its position. Two main solutions exist in the literature : delayed and undelayed initializations. The first way is straightforward : it only uses well-defined landmarks [ 1 ] [ 9 ]. This means that no wrong