Automatic Battle Damage Assessment Of Airports Based On Image Analysis

Accurate and real-time assessment of battle damage plays an important role in warfare. It is essential in planning attacks. In the past, battle damage assessment was carried out manually by the photo-analysts. They analyzed images available from various sensors and then drew a conclusion, which was both time-consuming and subjective. There is a natural need to develop a reliable and timely way to automatically evaluate damages inflicted by a strike to maximize the probability of kill with the minimum expenditure of offensive resources. In this thesis, we describe an automatic damage assessment process we developed to access the extent of battle damage of airports based on images taken before and after a strike. The process is composed of three modules, the extraction of pre-strike airport, the registration of the two images and the damage assessment module. At first, we introduce several auxiliary parts of an airport. Then, the function of each part is analyzed and the corresponding damage assessment rules are given. In order to extract airport from the imagery, several image-processing techniques including median filtering, HOUGH transform, mathematical morphology and region growing are used. After segmenting the key parts, such as runways, taxiways, shelters, dragging plane roads and linking roads between runways and taxiways, the images taken before and after the strike are registered. In the registration module, the particular registration method we chosen is one that uses both feature and area information of the images. In the damage assessment module, an "AND" operation is first implemented to locate the damages to each auxiliary part of the stricken airport. Then, based on assessment rules, the damage extent of each part is determined. This process enables fast and accurate evaluation of a strike with as little human supervision as possible. Experiments prove that the method described in this thesis needs much less time and the results are accurate. So, the method can offer a timely and exact reference to force commanders.