The development and employment of UAV have grown in last years. With this, the applications have also been increase. Among the applications, an important process is to automatic estimation of the Digital Elevation Model for a specific area. In many cases, it is important that the flight be made autonomously. The successful autonomous aerial navigation is dependent on several factors, among them the location of the unmanned aerial vehicle into the geographical space. In order to achieve this goal, there exist techniques that combine the Satellite Position System (SPS) with the Inertial Navigation System (INS). However, these techniques have been dependent on external data (signal from the SPS), that could not be available during the vehicle operation. When the SPS and the INS are not available, Computational Vision is an alternative for the navigation. Some studies have been developed about navigation with the use of images and Digital Elevation Models. In this paper it is presented an approach to automatically estimate a Digital Elevation Model using aerial images. Different techniques have been used to obtain characteristic points, from two images obtained at different instants. The Zernike Moment and Particle Collision Algorithm are used to find correspondence between the aerial images. The results show that the proposed techniques have been suitable for this problem.