Due to the nature of Unmanned Aerial Vehicles (UAVs) that operate automatically or semi-automatically, there is a possibility of emergency situations in which the landing zone cannot be reached during the mission. Since it is important to quickly find Safe Landing Zones (SLZ) by analyzing the terrain of the ground, in this paper, a system to detect the SLZs using RANSAC plane fitting algorithm and several constraints is presented. Also, the relative state of SLZs and UAV is modeled as Random Finite Set, in order to estimate the uncertainty associated with the states of UAVs as well as the number and state parameters of SLZs using a Gaussian Mixture Probability Hypothesis Density (GM-PHD) filter. In addition, this paper classifies surveillance regions that change in response to sensor movement and presents a method for predicting birth target intensity using measurements and sensor movement. Furthermore, a state estimator based on the Interacting Multiple Model (IMM) filter and a stochastic phase update method are proposed to account for the camera model’s uncertainties. The proposed systems are validated via a comparative evaluation using benchmark models in the ROS/GAZEBO simulation environment.