A simple and computationally efficient Fourier transform based algorithm for moving target detection in noisy background with no a priori target information

The topic of object detection and target tracking is of great interest in image sequence and dynamic scene analysis. The complexity and unpredictability of the content of an image scene contribute to the difficulty in locating the target object and then estimating its motion in the image sequence.; Designing an object detection algorithm that is applicable to a variety of image sequences, with no a priori knowledge about the content of the image scene, increases the level of complexity of the analysis. Usually, certain assumptions such as the target's shape, size, initial location or velocity have to be made about the object of interest, to help in its detection. This study presents a simple computationally efficient Fourier transform based algorithm for target detection in a noisy background without a priori target information. A simple and efficient adaptive filtering technique is presented and implemented in the three-dimensional Fourier domain. A complete algorithm, requiring no human interaction or a priori knowledge about the image scene content, is developed, implemented and tested in detecting multiple moving targets in “real-world” image sequences. For comparison purposes, a complete moving target detection algorithm based on image frame differencing is developed and presented. That algorithm embodies an improvement over the classical frame-differencing technique.