Kalie Knecht, PhD

Kalie Knecht, PhD

Kalie is an Applied Data Scientist working at the intersection of radiation detection, multimodal sensor fusion, robotics, and machine learning.

Talks and presentations

AI for Source Search

June 10, 2026

Invited talk, DNN Tech Modern and Emerging AI for Nuclear Nonproliferation Summer School, Oak Ridge, TN

Enhanced Use of Contextual Data for Quantitative Compton Imaging

November 08, 2023

Conference talk, IEEE Nuclear Science Symposium, Vancouver, BC, Canada

Contextual sensors can be equipped to radiation detection systems to allow free-moving 3D gamma-ray imaging through scene data fusion. Scene data fusion captures relevant scene information to construct a 3D map of an environment, which can be used to constrain image reconstruction techniques. While developed to enable free-moving imaging measurements, scene data fusion also has applications in cases that require multiple static measurements to generate 3D images, where the scene information can improve results. One way we seek to further leverage contextual scene information is to use the 3D map to automatically identify the radiological object of interest in a scene, and then determine optimal measurement poses around that object for quantitative Compton imaging. We also seek to improve the use of contextual information by identifying materials present in an environment to inform attenuation estimates for intervening materials. We present the approach for further leveraging the contextual information, and the preliminary results of using an SDF-enabled gamma-ray imager to find optimal measurement positions for a distributed source scenario to produce quantitative Compton images with minimization of associated uncertainties.

Scene-informed Optimal Measurement Positions for Quantitative Safeguards Measurements

May 25, 2023

Conference talk, INMM Annual Meeting, Vienna, Austria

Portable radiation detection systems can be equipped with contextual sensors to allow free-moving 3D gamma-ray imaging through scene data fusion (SDF). While developed for free-moving measurements, SDF also has applications in cases that require multiple static measurements to generate quantitative 3D images, where the scene information can improve results. The scene information captured by these devices can be leveraged to determine optimal measurement positions for quantification measurements, such as in safeguards applications, where limited time is provided to quantify nuclear material present.

Scene-informed Optimization of Measurement Locations for Radiological Assessments

November 08, 2022

Conference talk, IEEE Nuclear Science Symposium, Milan, Italy

Portable radiation detection systems can be equipped with contextual sensors to allow free-moving 3D gamma-ray imaging in a method called scene data fusion (SDF). The scene information provided by the contextual sensors can be used to enable 3D imaging and constrain image reconstruction to improve imaging accuracy and computational efficiency. While developed for free-moving measurements, SDF also has applications in cases that require multiple static measurements to generate 3D images, where the scene information can improve results. The scene information captured by these devices can be leveraged to determine optimal measurement positions for quantification measurements, such as in safeguards applications, where limited time is provided to quantify nuclear material present. In this work we are utilizing the 3D scene information obtained from an initial survey to create a measurement space that allows the determination of optimal positions and orientations (pose) of an instrument to assess radiological materials of interest. We will present our approach and results for performing an optimization to quantify properties of interest in low count rate environments and the minimization of associated uncertainties. %For quantification in low count rates, it is desired to have maximum and uniform sensitivity to the object of interest. Optimization techniques are developed to determine the measurement positions which best fulfill this criteria.

3D Compton Imaging of Distributed Sources around the Chernobyl Nuclear Power Plant

October 19, 2021

Conference talk, IEEE Nuclear Science Symposium, Piscataway, NJ

Portable radiation detection systems can be equipped with contextual sensors to allow free-moving 3D gamma-ray imaging in a method called scene data fusion (SDF). The scene information provided by the contextual sensors can be used to enable 3D imaging and constrain image reconstruction to improve imaging accuracy and computational efficiency. SDF could be a useful tool in many applications, including radiation mapping for accident remediation. To demonstrate this concept, Polaris-LAMP, a commercially available detector that has been integrated with contextual sensors, was operated in and around the Chernobyl Nuclear Power Plant. Compton imaging techniques are applied to the data collected over a series of 20-50 minute dynamic measurements around Pripyat, a nearby town which has been abandoned since the accident. These reconstruction techniques created 3D maps of gamma-ray sources and successfully identified hot spot candidates on key features of the scenes investigated, demonstrating the usefulness of SDF in radiation mapping of unknown distributed source environments. The contextual data captured by Polaris-LAMP also generated interesting visual products of scenes around Pripyat.

Evaluating 3D Gamma-ray Imaging Techniques for Distributed Sources at the Fukushima Daiichi Nuclear Power Station

November 03, 2020

Conference talk, IEEE Nuclear Science Symposium, Boston, MA

Portable radiation detection systems can be equipped with contextual sensors to allow free-moving 3D gamma-ray imaging in a method called scene data fusion (SDF). The scene information provided by the contextual sensors can be used to enable 3D imaging and constrain image reconstruction to improve imaging accuracy and computational efficiency. SDF could be a useful tool in many applications, including radiation mapping for contamination remediation. To demonstrate this concept, Polaris-LAMP, a commercially available detector that has been integrated with contextual sensors, was operated in a parking lot of some of the vehicles that were used in the cleanup of the March 2011 Fukushima Daiichi Nuclear Power Station accident. Proximity mapping and Compton imaging techniques are applied to the data collected over a series of short measurements in this parking lot. These reconstruction techniques created accurate 3D maps of gamma-ray sources and successfully identified hot spot candidates on key features of vehicles in the parking lot, demonstrating the usefulness of SDF in radiation mapping of unknown distributed source environments.