Spectral DAISY: a combined target spatial-spectral dense feature descriptor for improved tracking performance

Abstract

In EO tracking, target spatial and spectral features can be used to improve performance since they help distinguish the targets from each other when confusion occurs during normal kinematic tracking. In this paper we introduce a method to encode a target's descriptive spatial information into a multi-dimensional signature vector, allowing us to convert the problem of spatial template matching into a form similar to spectral signature matching. This allows us to leverage multivariate algorithms commonly used with hyperspectral data to the problem of exploiting panchromatic imagery. We show how this spatial signature formulation naturally leads to a hybrid spatial-spectral descriptor vector that supports exploitation using commonly-used spectral algorithms.We introduce a new descriptor called Spectral DAISY for encoding spatial information into a signature vector, based on the concept of the DAISY 1 dense descriptor. We demonstrate the process on real data and show how the combined spatial/spectral feature can be used to improve target/track association over spectral or spatial features alone.

Assessment of Spatial–Spectral Feature-Level Fusion for Hyperspectral Target Detection

“…2 and 3. S-DAISY is a local image descriptor that does not explicitly have RTS invariance, but has been empirically shown to be robust to minor levels of rotation and scale distortion in [12]. About each spatial pixel under consideration, Q rings of a predetermined radius R are defined.…”

“…The average number of confusers at full association (ANCFA) metric presented in [12] is an effective way of evaluating target in the absence of a statistically significant number of targets needed to produce a receiver-operating characteristic (ROC) curve. Previous applications of this metric [18] were limited to comparing individual pairs of target signatures extracted from the spatial location nearest to each target's centroid in F 0 and F 1 .…”

“…In [11], spectral anomalies found in HSI imagery are filtered with a spatial anomaly detector that operates on coincident regions in HRI imagery to eliminate false spectral cues. Local image descriptors and image moments provide natural building blocks for creating spatial-spectral target detection algorithms through the application of these techniques at each spatial pixel within each spectral band [12], [13]. Some descriptors are robust to spatial transformations [14]- [16], and may provide similar robustness when they are configured to capture an object's shape as well as texture.…”

“…The spatial feature vectors produced from each band can be concatenated, forming a spatial-spectral feature vector or pseudospectrum that simultaneously captures spectral, shape, and textural aspects about a pixel. The basis of this concept comes from S-DAISY [12], an extension to the local image pattern descriptor DAISY [17]. This structure allows us to frame the spatial-spectral target detection problem in one similar to that of hyperspectral material detection.…”

“…A signature migration step was added as a means of compensating for illumination differences affecting target signatures over time. Extending the evaluation approach from [12], we determined that pansharpened HSI in combination with image moments that are robust to rotation, translation, and scale (RTS) distortion can decrease false alarms among Bobcat 2013 targets versus using either imaging modality alone, particularly when a target changes position or orientation after its signature is measured.…”