TARGET DETECTION IN OPTICAL, MICROWAVE AND LIDAR DATA

Abstract

Historically, in the security-defence environment, information is derived through a subjective analytical approach principally based on the experience and the skills of the analyst who visually interprets the image(s). The spatial and contextual way to proceed varies and depends on the objective of the study. Spatial, pattern, texture, and, in general, spectral information is most of the time improved by standard image processing technics (i.e., image enhancement) for increasing the visual distinction between features. Different collateral/ancillary data, spatially and temporally correlated with the imagery, made available through different sources, may complement the analytical process providing worthwhile information, essential in helping, confirming, etc. the interpretation course and its inferences. For target detection in remotely sensed images, targets can be referred to as man-made or natural object or an event or activity of interest. Detection of an object or activity, such as military vehicles or vehicle tracks, is common in both military and civilian applications. Hyperspectral target identification algorithms look at the spectrum of each pixel to find targets based on the spectral characteristics of the target's surface material. Targets of interest may not be clearly resolved depending on the sensor's spatial resolution, hence the first fundamental characteristic of the hyperspectral target detection problem is that a "target present" versus "target absent" decision must be made individually for each pixel of a hyperspectral image. In this project, the various target detection algorithms along with the dimensionality reduction methods are explored and presented with valid results and discussions. The work involves the use of multi-platform, multi-dimensional and multi-sensor datasets making this work an important example of the various target detection approaches in remote sensing.