An Exploration of Feature Detector Performance in the Thermal-Infrared Modality

Abstract

Thermal-infrared images have superior statistical properties compared with visible-spectrum images in many low-light or no-light scenarios. However, a detailed understanding of feature detector performance in the thermal modality lags behind that of the visible modality. To address this, the first comprehensive study on feature detector performance on thermal-infrared images is conducted. A dataset is presented which explores a total of ten different environments with a range of statistical properties. An investigation is conducted into the effects of several digital and physical image transformations on detector repeatability in these environments. The effect of non-uniformity noise, unique to the thermal modality, is analyzed. The accumulation of sensor non-uniformities beyond the minimum possible level was found to have only a small negative effect. A limiting of feature counts was found to improve the repeatability performance of several detectors. Most other image transformations had predictable effects on feature stability. The best-performing detector varied considerably depending on the nature of the scene and the test.

Performance Evaluation of Feature Detectors and Descriptors Beyond the Visible

“…However, several modifications had to be made to the protocol. Indeed, as reported by [33], the original protocol presented limitations that made it particularly unsuitable for performance analysis in the thermal modality. The first limitation relates to the utilisation of feature detectors with default parameters which may lead to relatively low numbers of features on thermal images.…”

“…This disparity could be explained by the relatively recent introduction of thermal cameras (compared to visible-band sensors) in the field of computer vision. Vidas et al [33] investigated the performance of feature detectors on thermal images using a similar protocol to the one introduced by [18]. However, the authors did not consider feature descriptors.…”

“…Vidas et al [33] studied the effect of non-uniformities on the repeatability property of feature extractors. A very slight decay in performance over several minutes was reported in the results for most detectors.…”

“…These may render some computer vision algorithms unusable. For instance, optimal settings of feature detection pipelines were designed for visible-band images and their performance may degrade severely if they are not properly tuned and adapted for infrared modality [33]. Therefore, choosing the right image processing tools to operate in an exclusively infrared-based imagery environment can prove very challenging.…”

Evaluation of two-view geometry methods with automatic ground-truth generation

“…The same data set and a closely related method for determining correspondence is used in [10] to evaluate local descriptors. A similar data set and evaluation of feature performance in the thermal infra-red domain was reported in [11].…”