Created with Sketch.
Sun-glint false alarm mitigation in a maritime scenario
Hyperspectral sensors are gaining ever-increasing importance in the field of avionics systems due to the high information content that they are able to extract from the observed objects on the Earth's surface. This information resides in the high spectral resolution with which these sensors analyze the radiation reflected from objects illuminated by sunlight. In the maritime environment, hyperspectral sensors can be exploited for detecting rare and small objects, even sub-pixel, given the high contrast between the radiation reflected by the sea water and from the floating objects on the sea surface. Measurement campaigns were planned in a maritime environment using the Selex-Es SIM-GA hyperspectral sensor installed on avionics platforms. In both the Visible-Near InfraRed (VNIR) and Short Wave InfraRed (SWIR) spectral bands, the contrast between the objects, anomalous with respect to the sea water, and the water itself, is very high. Thus, the detection of the anomalous deployed objects is easily accomplished by means of a local statistical anomaly detection approach. To accomplish this task, Selex-Es recently focused on the development of a local anomaly detection algorithm based on the well-known Reed-Xiaoli (RX) approach, which can run on a processor unit on real-time, furnishing the object location during the reconnaissance flight. Unfortunately, it has been noticed that the performance of the anomaly detection algorithm is affected by sun-glint. The reflection on the sea surface of the solar radiation produces a high density of alarms, that makes challenging the task of detecting the objects of interest. In the literature, algorithms proposed to reduce alarms due to sun-glint, are mainly based on two steps. Firstly, potential sun-glint pixels are detected exploiting the NIR spectral channels. Then, sun-glint pixels are recovered relying to a linear stretching of the spatial regions where the phenomenon of the sun-glint has been detected. In this paper, it is introduced a strategy aimed at discriminating the sun-glint false alarms from the effective alarms related to targets of potential interest. The designed algorithm exploits the entire VNIR and SWIR spectral bands. False alarms due to glint are mitigated performing a local spatio-spectral analysis on each alarm furnished by the anomaly detector. Each potential anomalous pixel is decided to belong to sun-glint or object classes on the basis of a metric of similarity between the pixel itself and the surrounding background. The proposed strategy has been implemented in the real-time processing architecture developed by Selex-ES. The technique has been tested on hyperspectral images collected during the measurement campaign conducted by CISAM in conjunction with University of Pisa, CSSN and Selex-ES in Viareggio, Italy on 8-9 of May, 2013. Several experiments were conducted, setting up scenarios with small man-made objects deployed on the sea surface, so as to simulate search and rescue operations (e.g., life jackets, buoys, suits).The results have highlighted the effectiveness of the solution in terms of mitigation of false alarms due to glints on maritime scenario.
Rossi Alessandro, Riccobono Aldo, Landini Stefano
Paper for Seminar/Symposium/Conference
SPIE 2014 - International Conference on Remote Sensing (22-25 September 2014, Amsterdam, The Netherlands)
Inserisci il testo di ricerca