Object level HSI-LIDAR data fusion for automated detection of difficult targets

Abstract

Data fusion from disparate sensors significantly improves automated man-made target detection performance compared to that of just an individual sensor. In particular, it can solve hyperspectral imagery (HSI) detection problems pertaining to low-radiance man-made objects and objects in shadows. We present an algorithm that fuses HSI and LIDAR data for automated detection of man-made objects. LIDAR is used to define a set of potential targets based on physical dimensions, and HSI is then used to discriminate between man-made and natural objects. The discrimination technique is a novel HSI detection concept that uses an HSI detection score localization metric capable of distinguishing between wide-area score distributions inherent to natural objects and highly localized score distributions indicative of man-made targets. A typical man-made localization score was found to be around 0.5 compared to natural background typical localization scores being less than 0.1.

1541nm GmAPD LADAR system

“…12,14 None the less, working at 1541 nm requires a smaller band gap in the semiconductor comprising the GM-APDs which results in an increase in dark noise by slightly more than a factor of 2 compared with similar GM-APDs optimized for 1064 nm. 12 During daytime use this disadvantage will likely be offset somewhat as there is more than twice as much solar radiation at 1064 nm as at 1541 nm at sea level.…”

Registration of Urban Aerial Image and LiDAR Based on Line Vectors

“…Various reference frames in airborne LiDAR data and imagery, along with systematic errors inside of the LiDAR system and aerial photography system, cause spatial position deviations between the LiDAR data and the image, which may greatly affect the accuracy of the data results. Both of them, before being integrated for application, should be included in a unified coordinate system to achieve their spatially-accurate registration, namely the registration of LiDAR data and aerial images [8].…”

Fusion of waveform LiDAR data and hyperspectral imagery for land cover classification

“…Since HI records surface spectral reflectance characteristics of objects, while FWL reveals scattering properties and geometrical structure (roughness, slope, spatial distribution) of targets, their complementary nature suggests that the fusion of HI and FWL can be beneficial for applications such as land cover classification, forest inventory estimation, and obscured target detection Kanaev et al, 2011). Although previous work has addressed the issue of fusing LiDAR data and HI, for example (Dalponte et al, 2008;Erdody and Moskal, 2010;Jones et al, 2008;Dinuls et al, 2012), the work has predominantly been limited to using simple gridded discrete return LiDAR data, such as Digital Elevation Models (DEM), Digital Terrain Models (DTM), or using data from a FWL system that has been converted to discrete return point clouds (Anderson et al, 2008;Asner et al, 2007;Ranjani et al, 2014;Paris and Bruzzone, 2015 feature level fusion with HI (Sarrazin et al, 2011;Jung, 2011).…”