Deep Learning for LiDAR Waveforms with Multiple Returns

Andreas Aßmann; Brian Stewart; Andrew Michael Wallace

Deep Learning for LiDAR Waveforms with Multiple Returns

Andreas Aßmann, Brian Stewart, Andrew Michael Wallace

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

We present LiDARNet, a novel data driven approach
to LiDAR waveform processing utilising convolutional neural
networks to extract depth information. To effectively leverage
deep learning, an efficient LiDAR toolchain was developed, which
can generate realistic waveform datasets based on either specific
experimental parameters or synthetic scenes at scale. This enables
us to generate a large volume of waveforms in varying conditions
with meaningful underlying data. To validate our simulation
approach, we model a super resolution benchmark and crossvalidate
the network with real unseen data. We demonstrate
the ability to resolve peaks in close proximity, as well as to
extract multiple returns from waveforms with low signal-to-noise
ratio simultaneously with over 99% accuracy. This approach is
fast, flexible and highly parallelizable for arrayed imagers. We
provide explainability in the deep learning process by matching
intermediate outputs to a robust underlying signal model.

Original language	English
Title of host publication	Proceedings of the 28th European Signal Processing Conference
Publication status	Accepted/In press - 21 Sep 2020
Event	28th European Signal Processing Conference - Amsterdam, Netherlands Duration: 18 Jan 2021 → 22 Jan 2021 https://eusipco2020.org/

Conference

Conference	28th European Signal Processing Conference
Abbreviated title	EUSIPCO 2020
Country	Netherlands
City	Amsterdam
Period	18/01/21 → 22/01/21
Internet address	https://eusipco2020.org/

Fingerprint Dive into the research topics of 'Deep Learning for LiDAR Waveforms with Multiple Returns'. Together they form a unique fingerprint.

Cite this

@inproceedings{29360c61b37b429e8a4205b910699682,

title = "Deep Learning for LiDAR Waveforms with Multiple Returns",

abstract = "We present LiDARNet, a novel data driven approachto LiDAR waveform processing utilising convolutional neuralnetworks to extract depth information. To effectively leveragedeep learning, an efficient LiDAR toolchain was developed, whichcan generate realistic waveform datasets based on either specificexperimental parameters or synthetic scenes at scale. This enablesus to generate a large volume of waveforms in varying conditionswith meaningful underlying data. To validate our simulationapproach, we model a super resolution benchmark and crossvalidatethe network with real unseen data. We demonstratethe ability to resolve peaks in close proximity, as well as toextract multiple returns from waveforms with low signal-to-noiseratio simultaneously with over 99% accuracy. This approach isfast, flexible and highly parallelizable for arrayed imagers. Weprovide explainability in the deep learning process by matchingintermediate outputs to a robust underlying signal model.",

author = "Andreas A{\ss}mann and Brian Stewart and Wallace, {Andrew Michael}",

year = "2020",

month = sep,

day = "21",

language = "English",

booktitle = "Proceedings of the 28th European Signal Processing Conference",

note = "28th European Signal Processing Conference, EUSIPCO 2020 ; Conference date: 18-01-2021 Through 22-01-2021",

url = "https://eusipco2020.org/",

}

TY - GEN

T1 - Deep Learning for LiDAR Waveforms with Multiple Returns

AU - Aßmann, Andreas

AU - Stewart, Brian

AU - Wallace, Andrew Michael

PY - 2020/9/21

Y1 - 2020/9/21

N2 - We present LiDARNet, a novel data driven approachto LiDAR waveform processing utilising convolutional neuralnetworks to extract depth information. To effectively leveragedeep learning, an efficient LiDAR toolchain was developed, whichcan generate realistic waveform datasets based on either specificexperimental parameters or synthetic scenes at scale. This enablesus to generate a large volume of waveforms in varying conditionswith meaningful underlying data. To validate our simulationapproach, we model a super resolution benchmark and crossvalidatethe network with real unseen data. We demonstratethe ability to resolve peaks in close proximity, as well as toextract multiple returns from waveforms with low signal-to-noiseratio simultaneously with over 99% accuracy. This approach isfast, flexible and highly parallelizable for arrayed imagers. Weprovide explainability in the deep learning process by matchingintermediate outputs to a robust underlying signal model.

AB - We present LiDARNet, a novel data driven approachto LiDAR waveform processing utilising convolutional neuralnetworks to extract depth information. To effectively leveragedeep learning, an efficient LiDAR toolchain was developed, whichcan generate realistic waveform datasets based on either specificexperimental parameters or synthetic scenes at scale. This enablesus to generate a large volume of waveforms in varying conditionswith meaningful underlying data. To validate our simulationapproach, we model a super resolution benchmark and crossvalidatethe network with real unseen data. We demonstratethe ability to resolve peaks in close proximity, as well as toextract multiple returns from waveforms with low signal-to-noiseratio simultaneously with over 99% accuracy. This approach isfast, flexible and highly parallelizable for arrayed imagers. Weprovide explainability in the deep learning process by matchingintermediate outputs to a robust underlying signal model.

M3 - Conference contribution

BT - Proceedings of the 28th European Signal Processing Conference

T2 - 28th European Signal Processing Conference

Y2 - 18 January 2021 through 22 January 2021

ER -