Moving objects detection is one of the important problems in recent years in computer vision. Accuracy and reliability of moving object detection algorithms have a direct impact on problems like object tracking and locating. One of the main issues of the moving object detection is the movement of camerawhich is called Ego-motion. Ego-motion concerns 2D motion observed in an image sequence that is caused by 3D camera motion. For moving object detection in this situation we must distinguish Ego-motion from motion that is caused by movement of objects. In most of methods in the first step, Ego-motion is estimated and compensated from images sequence, thenthe usual methods that are developed for moving object detection in fixed camera images is performed. In this research at the first we introduce a mathematical model for Ego-motion. Then the model is used to group current methods. Among the discussed methods, features matching approach for its low time and computation complexity and independency form hardware in comparison with other methods was considered. In this research we introduce a measure that shows the necessity of making sub-images for extracting 1D features form two consequent images. Then we show that for extracting 1D features form aerial image making sub-imagesare not necessary. Hence we developed an efficient method for moving object detection in aerial image of moving camera which have lower time and computation cost in comparison with others method. Beside this advantage, the proposed method is able to detect moving objects in scenes that have a uniform background and moving objects have small movement. The proposed method implemented in MATLAB and run this in system with processor CORE i3 with frequency 2.53 GHz and 4 GB RAM in windows 7, on the images with resolution 480×640, moving objects are detectedin 0/6 second.
