multi-sensor fusion and 3d mapping

UK-US Academic Network in Nuclear Security and Nonproliferation Skills Virtual Workshop

Multi-Sensor Fusion and 3D Mapping

Kai VetterUniversity of California, Berkeley

1Kai Vetter, [email protected]

Consortium for Monitoring, Technology, and Verification


Introduction and Motivation• Detection of (dynamic) weak nuclear signatures embedded in (dynamic) naturally

occurring and man-made radiological backgrounds• E.g. Proliferation detection

• Effective assessment and monitoring of facilities, materials, and operations• E.g. Safeguards

• Effective and safe assessment of objects and mapping of contamination and verification of decontamination• E.g. Emergency response

• Drive and utilize (~linear) developments in nuclear radiation detection and imaging and combine it with (~exponential) advances in non-radiological sensing, data processing, and computer vision


Performance

Time1900 20001950

Fluorescence & Film

Geiger Müller

CounterNaI(Tl) (1948)

Si & Ge (& CdZnTe) detectors

MWPCTPC

TomographyPET & SPECT

Dual-ModalityECT + CT

BGO (1982)

LSO (1992)

CLLBC(2016)

Segmented CdZnTe

(CPG & pixels)Scintillator

arrays

Radiation detection & imaging(Selected examples)

Evolution of relevant concepts and technologies


Selected radiation detection & imaging developments at UC Berkeley and LBNL

20 mm

10 mm

10 mm

5 mm

NewOriginalCompact, room-temperature,

<2% 𝛥Erel

Coplanar Grid CdZnTe(CPG CZT)

HEMIHigh-Efficiency Multimode Imager

96 (1cm3) CPG detectors arranged in 2 planes serving as active coded

aperture (FOV<60 deg) and Compton imager (360 deg)

>100 CPG (1cm3) detectors arranged on sphere serving as active coded aperture (360 deg) and Compton

imager (360 deg)

PRISMPortable Radiation Imaging Spectroscopy and Mapping

15cm 8cm

~60 (1cm3) CPG detectors arranged in cube for 360 deg FOV gamma-ray imaging (coded aperture and Compton imaging)

Mini-PRISMNG-PRISM or

NG-LAMP

~60 (1.27cm)3 CLLBC detectors (3-4% 𝛥Erel) as

Mini-PRISM for gamma-ray detection and imaging and neutron mapping at ultra-

high-count rates (>10Mcps)

HEMI on unmanned RMAX in Fukushima

Gamma-ray detection and imaging + Neutron detection mapping

1995

2014

2008 2018 2019 2020


+ Beta-(only) detection (e.g. Sr-90)

Selected radiation detection & imaging developments at UC Berkeley and LBNL

CZT detectors2 x 20 x 20 x 5 mm3& DOI Readout

5 cm

Activ

ity [n

Ci]

-0.5

0.0

0.5

1.0

1.5

2.0

2.5

Sr-8

9Sr

-90

Cs-1

34

Cs-1

37

K-4

0Y-

91Zr

-95

Nb-

95M

o-99

Tc-9

9mRu

-103

Te-1

29m

I-131

I-132

Te-1

32Cs

-13

6Ba

-140

La-1

40Ce

-141

Pr-1

43Ce

-144

Pr-1

44Pm

-147

Nd-

147

Np-

237

Activity Reconstruction of Soil Sample [nCi] in sample holder (13g)

Eckert&ZieglerLBNL

Total Activity: EZ: 8.4 nCiLBNL: 12.0 nCi

Sensitivity to x10 below Derived Response Level (DRL) within 60 min for Sr-90

Gamma-ray detection and imaging + Neutron detection mapping

20 mm

10 mm

10 mm

5 mm

NewOriginalCompact, room-temperature,

<2% 𝛥Erel

Coplanar Grid CdZnTe(CPG CZT)

1995


Performance

Time1900 20001950

Personal Computer

Z1, ENIAC, “Manchester Baby”

E.g. 1,024 bits & 1,000 ops

Laptop

Smart phones

+ NetworksOptical and wireless

e.g. 5GEdge computing

Quantum computingNeuromorphic computing

E.g. A14 Bionic with 6-core CPU and 4-core

GPU; 12x109 transistors or 11x1012 ops by neural

engine+ 12MP CMOS visual

& LiDAR sensors

Cloud computing

Computing & sensing(Selected examples)

CCD sensor1880’s Digital cameras

(Active pixel/ CMOS)

CMOS (active pixel)

sensorAnalog cameras

RangingRadar Lidar

Evolution of relevant concepts and technologies+ Mixed Reality

AR & VR


Performance

Time1900 20001950

Radiation detection & imaging(Selected examples)

Computing & sensing(Selected examples)⊗

Evolution of relevant concepts and technologies

⇒ Advances in all areas of nuclear detection, assessment, and mapping


Technical Approach - 3D Mapping • Combine radiation detectors and imagers with contextual sensorso Localization And Mapping Platform LAMPo E.g. NG-LAMP w/ LiDAR, visual camera, GPS/IMU

UF6 in storage cylinders

Material Accountancy

Consequence management

Bamboo forest – Cs-137 map ~100 m

Contaminated home

Fukushima Prefecture (Namie Town)

Reactor 2 floorDebris removal “claw”

Chernobyl

Hold-up

Np-237 in Pu-238 facility

Pre-event Search/Survey

Map of 49er Levi’s stadium

Cs-137 contamination


Technical Approach - Detection, Localization, Visualization and Quantification• Combine radiation detectors and imagers with contextual sensors

Detector Module

Contextual Data Acquisition

DAQ & Power

PMT & Digibase

NaI(Tl) Crystals

Contextual Sensors

Lidar

IMU

CameraSensor Interfaces

Antenna

INS

OBDii

Ruggedized Computer

Hexagonal Array

Cs-137 source

Ba-133 source

3D map with two sources detected, identified, and localized in 4 min flight

UAS-based search

3D Gamma-ray imaging

3D Neutron mapping

3D map of heavily shielded Pu-surrogate source detected and localized within seconds

Remote control and visualization

Object racking and attribution from static and mobile systems

Utilize advanced detection and identification algorithms: Non-negative Matrix Factorization, Dense Autoencoder, Convolutional and Recurrent Neural Networks

Vehicle-based Urban Search

3D Semantically labelled scenes

Object-count attribution


Technical Approach - Detection, Localization, Visualization and Quantification

Activity and attenuation estimation

Top-down view of 3D Model

Air dose-rate in µSv/hr at 1m AGL

Co-60 source (0.44mCi)Estimate: 0.33±0.08mCi

Accurate and fast detection and (3D) localization ANDEstimation of source activity and dose rates with uncertainties

Cs-137 source (1.71mCi)Estimate: 1.55mCi±0.12mCi

Scanning, detection, 3D localization, and activity estimation of containers within minutes

• Combine radiation detectors and imagers with contextual sensors

Accurate mapping of container stack, detection, identification, 3D localization, and activity estimation of

Cs-137 source within several minutes

Localization of Cs-137 w/ attenuation (within 30 cm)

Reconstructed location w/o attenuation


Conclusion• The integration of recent developments in radiation detection and imaging and

advancements in computer vision and data processing provide new means to detect, map, and visualize nuclear materials.• Multi-sensor fusion provides• Enormous potential for enhancing capabilities across several missions in nuclear

security and nonproliferation;• Outstanding opportunities for students and the next generation to engage in

multi-disciplinary research and teams;• Excellent opportunities exist for collaborations to integrate nuclear and non-nuclear

instruments and fuse their data and to optimize the extraction of relevant features and quantities for specific needs.


Acknowledgements

• All Members of the Berkeley Applied Nuclear Physics Program• Our Sponsors from DOE/NNSA, DOD/DTRA, DHS/CWMD, and JAEA

12

multi-sensor fusion and 3d mapping

Documents