Research Of Anomaly Detection Algorithms Of Infrared Imagery

The infrared anomaly detection has received considerable interest in the remote sensing community. With strong adaptability to various circumstances, good hiding capacity and precise detecting ability, it plays an important role in civilian and military domain. So a close study of this technology has both theoretical significance and utilitarian value. Aiming at solving the problems such as the weakness of the detected image signal, the lowness of the image signal to noise ratio, and the distraction by the background and the noise, the research results can be summarized as follows:1. Considering the characteristics that the image’s background varies smoothly, the non-parameter regression based on kernel is adapted to predict the image background, so that the background and the anomaly could be separated. And eventually the targets could be detected by removing the noises from the redundant image.2. Aiming at the infrared target detection in the sea background with non-stationary and non-linear features, a novel infrared target detection algorithm is proposed. The image entropy is used to locate the target, and Kmeans cluster algorithm is used to find the cluster indices. The targets are extracted effectively as well as the outline with higher speed and better performance.3. In order to promote the accuracy and the speed of the target detection algorithm, the FKT based on saliency infrared anomaly detection is proposed. The spectral residual is used to process the original image at first, and then the FKT is used afterwards. The result is much better than the original methods by enhancing the signal-to-clutter ratio and increasing the accuracy of target detection.