Anomaly detection in multi-temporal infrared thermography

Anomaly detection is an important tool in various types of image processing and was widely investigated in the area of hyperspectral imaging. This research focuses on anomaly detection within multi temporal thermal images. We used three types of datasets; I) anomaly-free images, II) synthetically anomaly images, III) images with small metal objects, both buried and exposed. In this article, we introduce a new algorithm called RXmin in which we examine the metric distance between the suspected pixel to other pixels in the image. In contrast to visible light imagery, this method, when operating in the infrared is indifferent to the presence of sunlight and therefore can be used during the night. The proposed algorithms are general in nature and can be used for other types of information or functions such as video analysis, array processing, seismic signal processing etc.