In progress – view active session
Conference 13039
Automatic Target Recognition XXXIV
22 - 24 April 2024 | National Harbor 5
Conference Co-Sponsor
All sponsors 
Show conference sponsors +
View Session
Opening Remarks
22 April 2024 • 8:00 AM - 8:10 AM EDT | National Harbor 5
Session Chair:
Timothy L. Overman, Prime Solutions Group, Inc. (United States)
Opening remarks for Automatic Target Recognition XXXIV.
Session 1:
Machine Learning for Automatic Target Recognition I: Joint Session with Conferences 13036 and 13039
22 April 2024 • 8:10 AM - 10:10 AM EDT | National Harbor 5
Session Chair:
Timothy L. Overman, Prime Solutions Group, Inc. (United States)
13039-1
Image data for training ML models: how would synthetic methods help
(Keynote Presentation)
22 April 2024 • 8:10 AM - 8:50 AM EDT | National Harbor 5
13039-2
22 April 2024 • 8:50 AM - 9:10 AM EDT | National Harbor 5
Show Abstract +
The intricacies of visual scenes in Automatic Target Recognition (ATR) necessitate sophisticated models for nuanced interpretation. Vision-language models, notably CLIP (Contrastive Language-Image Pre-training), bridge visual perception and linguistic description. However, their effectiveness in ATR relies on targeted fine-tuning, challenged by the unimodal nature of datasets like the Defense Systems Information Analysis Center (DSIAC) ATR data. We propose a novel fine-tuning approach for CLIP, enriching DSIAC data with algorithmically generated captions for a multimodal training environment. Central to our innovation is a homotopy-based multi-objective optimization strategy, adept at balancing model accuracy, generalization, and interpretability—key factors for ATR success. Implemented in PyTorch Lightning, our approach propels the frontier of ATR model optimization while also effectively addressing the intricacies of real-world ATR requirements.
13039-3
22 April 2024 • 9:10 AM - 9:30 AM EDT | National Harbor 5
Show Abstract +
This study introduces a depth-aware approach for detecting small-scale camouflaged objects, leveraging the Swin Transformer and Ghost Convolution Layer. We employ multimodal depth maps to enhance spatial understanding, which is crucial for identifying camouflaged items. The Swin Transformer captures extensive contextual data, while the Ghost Convolution Layer boosts computational efficiency. We validate our method on unique quasi-synthetic and comparative synthetic datasets created for this study. An ablation study and GRAD-CAM visualization further substantiate the model's effectiveness. This research offers a novel framework for improving object detection in challenging camouflaged environments.
13039-4
22 April 2024 • 9:30 AM - 9:50 AM EDT | National Harbor 5
Show Abstract +
This work involves performing black box adversarial attacks using light as a medium against image classifier neural networks. The method of generating these adversarial examples involves querying the target network to inform decisions on designing a pattern upon the Fourier plane. The shapes are designed to target regions of the Fourier domain effectively without being able to back-propagate loss toward said plane.
13039-5
22 April 2024 • 9:50 AM - 10:10 AM EDT | National Harbor 5
Show Abstract +
In recent decades, there have been notable advancements in Automatic Target Recognition (ATR) systems. One technique that has played a crucial role in improving the accuracy and efficiency of these systems is dictionary-learning. This paper provides a thorough examination, documenting the evolutionary progression of dictionary-learning methodologies in the field of Automatic Target Recognition (ATR). Commencing with initial approaches such as K-SVD and MOD, we examine their fundamental influence and subsequent evolution towards more adaptable methodologies, such as online and convolutional dictionary learning. The focus is on comprehending the enhancements in target recognition achieved by dictionary-learning methods, particularly in demanding scenarios characterized by factors such as noise, occlusions, and diverse target orientations. In addition, we investigate the recent incorporation of deep learning principles into conventional dictionary-based frameworks, revealing a hybrid paradigm that holds the potential to significantly transform automatic target recognition (ATR) capabilities.
Break
Coffee Break 10:10 AM - 11:00 AM
Session 2:
Electro-Optical and Infrared Detection and Tracking I
22 April 2024 • 11:00 AM - 12:00 PM EDT | National Harbor 5
Session Chair:
Kenny Chen, Lockheed Martin Missiles and Fire Control (United States)
13039-7
22 April 2024 • 11:00 AM - 11:20 AM EDT | National Harbor 5
Show Abstract +
Advances in camera design have resulted in the development of next generation “event-based” imaging sensors. These imaging sensors provide super-high-temporal resolution in individual pixels, but only pick-up changes in the scene. This enables interesting new capabilities like bullet-tracking and hostile fire detection, and their low power-consumption is important for edge AiTR systems.
However, AiTR algorithms require massive amounts of data for system training and development and these data collections are expensive and time-consuming. Therefore, it is of interest to explore whether and how current data can be modified to simulate event-based images for training and evaluation.
In this work, we present results from training and testing CNN and non-CNN architectures on both simulated and real event-based imaging sensor systems.
13039-8
22 April 2024 • 11:20 AM - 11:40 AM EDT | National Harbor 5
Show Abstract +
In this study, we present a real-time vehicle detection program that combines the YOLO-X object detection algorithm with a multi-object Kalman filter tracker, specifically designed for analyzing 3-Dimensional (3-D) LiDAR data. Our approach involves capturing videos of 8 vehicles using an ASC 3-D Flash LiDAR camera, which provides intensity and range data sequences. These sequences are then converted into representative RGB images, used to train the YOLO-X object detector neural network. To further enhance the detection accuracy for obscured vehicles and minimize the missed detection rate, we integrate Kalman filter trackers into the detection algorithm. The resulting algorithm is lightweight and capable of producing highly accurate inference results in near real-time on a live stream of LiDAR data. To demonstrate the applicability of our approach on small, unmanned vehicles/drones, we deploy the application on NVIDIA's Jetson Orin Nano embedded processor for AI.
13039-9
22 April 2024 • 11:40 AM - 12:00 PM EDT | National Harbor 5
Show Abstract +
Visually detecting camouflaged objects is a hard problem for humans as well as computer vision algorithms. Strong similarities between object and background appearance make the task significantly more challenging than traditional object detection or segmentation tasks. Current state-of-the-art models use either convolutional neural networks or vision transformers as feature extractors. They are trained in a fully supervised manner and thus need a large amount of labeled training data. In this paper, self-supervised frugal learning methods are introduced to camouflaged object detection. The overall goal is to fine-tune two methods, namely SINet-V2 and HitNet, pre-trained for camouflaged animal detection to the task of camouflaged human detection. Therefore, we use the public dataset CPD1K that contains camouflaged humans in a forest environment. We create a strong baseline using supervised frugal transfer learning for the fine-tuning task. Then, we analyze three pseudo-labeling approaches to perform the fine-tuning task in a self-supervised manner. Our experiments show that we achieve similar performance by pure self-supervision compared to supervised frugal learning.
Break
Lunch Break 12:00 PM - 1:30 PM
Session 3:
Radar Frequency and Synthetic Aperture Radar Automatic Target Recognition I
22 April 2024 • 1:30 PM - 2:30 PM EDT | National Harbor 5
Session Chair:
Kristen Jaskie, Prime Solutions Group, Inc. (United States)
13039-11
22 April 2024 • 1:30 PM - 1:50 PM EDT | National Harbor 5
Show Abstract +
Radar target recognition with Random Forests (RF) and using stepped-frequency radar features is the focus of this paper. Recent comparative studies between RF and convolutional neural networks (CNN) showed that RF yields reliable robust target recognition results with relatively fast training and testing time. The appeal of RF is that they can be implemented in parallel and have far fewer tunable parameters than CNN. In addition to providing measures of variable significance, and permitting differential class weighting, RF can help with imputation of missing data [1]. These RF properties make them a good target recognition alternative tool especially in scenarios where the data is occluded, or corrupted with extraneous scatterer, or when the target signature at certain azimuth position changes drastically compared to other likely positions (or aspect angles). This paper uses real radar data of commercial aircraft models recorded in a compact range. The results show that RF offers a fast and reliable alternative for target recognition systems especially under realistic radar operating conditions.
13039-12
22 April 2024 • 1:50 PM - 2:10 PM EDT | National Harbor 5
Show Abstract +
Reliable computer vision object classification in imagery is import in security applications that may have high stakes decisions made from automated algorithms. In real world scenarios, it is often impractical to meet the implicit assumption that all relevant, labelled data may be attained prior to training. To avoid performance degradation, the recently developed Open-set Automatic Target Recognition (ATR) frame is applied to the classification of ships from clutter in satellite, Electro-Optical (EO) imagery, and is shown to reliably identify data that is out of distribution from training data. This enables and operator to know whether to believe classification results from the deep learning-based algorithm.
13039-31
22 April 2024 • 2:10 PM - 2:30 PM EDT | National Harbor 5
Show Abstract +
This paper describes a comprehensive computational imaging field trial conducted in Meppen, Germany, aimed at assessing the performance of cutting-edge computational imaging systems (compressive hyperspectral, visible/shortwave infrared single-pixel, wide-area infrared, neuromorphic, high-speed, photon counting cameras, and many more) by the members of NATO SET-RTG-310. The trial encompassed a diverse set of targets, including dismounts equipped with various two-handheld objects and adorned with a range of camouflage patterns, as well as fixed and rotary-wing Unmanned Aerial System (UAS) targets. These targets covered the entire spectrum of spatial, temporal, and spectral signatures, forming a comprehensive trade space for performance evaluation of each system.
The trial, which serves as the foundation for subsequent data analysis, encompassed a multitude of scenarios designed to challenge the limits of computational imaging technologies. The diverse set of targets, each with its unique set of challenges, allows for the examination of system performance across various environmental and operational conditions.
Break
Coffee Break 2:30 PM - 3:05 PM
Session 4:
Electro-Optical and Infrared Detection and Tracking II
22 April 2024 • 3:05 PM - 3:45 PM EDT | National Harbor 5
Session Chair:
Timothy L. Overman, Prime Solutions Group, Inc. (United States)
13039-17
22 April 2024 • 3:05 PM - 3:25 PM EDT | National Harbor 5
Show Abstract +
Hybrid dynamical systems are a natural model for missions in which several behaviors are required to achieve the goal of the mission. Missions are tasks featuring interacting subtasks, such as the decision of where and how to search and when to transition between behaviors. While the discrete nature of mission actions (which subtask to accomplish) and the continuous nature of real-world physical state spaces make hybrid systems a good model, control in such systems is poorly understood. Despite this, we find the formalism to have significant value and develop hierarchical state estimation tools to control agents in a hybrid framework and execute missions. In past work, we developed hierarchical dynamic target modeling to estimate the progress of search and track scenarios. In this work, we consider the related problem of searching for stationary targets that appear in formation. Executing such a search efficiently and gaining situational awareness presents unique challenges. We develop a generative hierarchical model for target locations that relies on stochastic clustering techniques and ideas from object Simultaneous Location and Mapping (SLAM) to address these challenges.
13039-18
22 April 2024 • 3:25 PM - 3:45 PM EDT | National Harbor 5
Show Abstract +
Low-resolution image object recognition and tracking is often required for battlefield
reconnaissance. We propose a fast detector-agnostic multi-hypothesis tracker for improving situational
awareness using electro-optical video data. Our approach uses standard techniques such as
YOLO, match filters, or shape transforms to segment objects of interest in an image. From two
or more successive detections, we initialize a linear quadratic estimator that extrapolates
probable trajectories of the target of interest in the image space.
Symposium Plenary
22 April 2024 • 5:00 PM - 6:30 PM EDT | Potomac A
Session Chairs:
Tien Pham, The MITRE Corp. (United States), Douglas R. Droege, L3Harris Technologies, Inc. (United States)
View Full Details: spie.org/dcs/symposium-plenary
Chair welcome and introduction
22 April 2024 • 5:00 PM - 5:05 PM EDT
DoD's microelectronics for the defense and commercial sensing ecosystem (Plenary Presentation)
Presenter(s): Dev Shenoy, Principal Director for Microelectronics, Office of the Under Secretary of Defense for Research and Engineering (United States)
22 April 2024 • 5:05 PM - 5:45 PM EDT
NATO DIANA: a case study for reimagining defence innovation (Plenary Presentation)
Presenter(s): Deeph Chana, Managing Director, NATO Defence Innovation Accelerator for the North Atlantic (DIANA) (United Kingdom)
22 April 2024 • 5:50 PM - 6:30 PM EDT
Symposium Panel on Microelectronics Commercial Crossover
23 April 2024 • 8:30 AM - 10:00 AM EDT | Potomac A
View Full Details: spie.org/dcs/symposium-panel
The CHIPS Act Microelectronics Commons network is accelerating the pace of microelectronics technology development in the U.S. This panel discussion will explore opportunities for crossover from commercial technology into DoD systems and applications, discussing what emerging commercial microelectronics technologies could be most impactful on photonics and sensors and how the DoD might best leverage commercial innovations in microelectronics.
Moderator:
John Pellegrino, Electro-Optical Systems Lab., Georgia Tech Research Institute (retired) (United States)
Panelists:
Shamik Das, The MITRE Corporation (United States)
Erin Gawron-Hyla, OUSD (R&E) (United States)
Carl McCants, Defense Advanced Research Projects Agency (United States)
Kyle Squires, Ira A. Fulton Schools of Engineering, Arizona State Univ. (United States)
Anil Rao, Intel Corporation (United States)
Break
Coffee Break 10:00 AM - 10:20 AM
Opening Remarks
23 April 2024 • 10:20 AM - 10:30 AM EDT | National Harbor 5
Session Chair:
Riad I. Hammoud, PlusAI, Inc. (United States)
Opening remarks for Automatic Target Recognition XXXIV.
Session 5:
Machine Learning for Automatic Target Recognition II
23 April 2024 • 10:30 AM - 12:10 PM EDT | National Harbor 5
Session Chair:
Kristen Jaskie, Prime Solutions Group, Inc. (United States)
13039-19
End-to-end machine learning for co-optimized sensing and automated target recognition
(Keynote Presentation)
23 April 2024 • 10:30 AM - 11:10 AM EDT | National Harbor 5
Show Abstract +
Many sensors produce data that rarely, if ever, is viewed by a human, and yet sensors are often designed to maximize subjective image quality. For sensors whose data is intended for embedded exploitation, maximizing the subjective image quality to a human will generally decrease the performance of downstream exploitation. In recent years, computational imaging researchers have developed end-to-end learning methods that co-optimize the sensing hardware with downstream exploitation via end-to-end machine learning. This talk will describe two such approaches at Kitware. In the first, we use an end-to-end ML approach to design a multispectral sensor that’s optimized for scene segmentation and, in the second, we optimize post-capture super-resolution in order to improve the performance of airplane detection in overhead imagery.
13039-32
23 April 2024 • 11:10 AM - 11:50 AM EDT | National Harbor 5
Show Abstract +
Remotes sensing and vision problems like object detection and recognition from various active and passive sensors are of great value to many DoD use cases. Usually due to sensor or communication link limitations the images received are of low resolution, quality and have compression artifacts. To combat this we developed a new direct vision task feature pyramid recovery with a joint frequency and pixel domain neural learning approach. It had many successes in problems like ATR from low resolution SAR and EO images, joint delburring and target detection, as well as the very low bit rate complex SAR image compression for phase recovery.
13039-21
23 April 2024 • 11:50 AM - 12:10 PM EDT | National Harbor 5
Show Abstract +
The field of quantum computing, especially quantum machine learning (QML), has been the subject of much research in recent years. Leveraging the quantum properties of superposition and entanglement promises exponential decrease in computation costs. With the promises of increased speed and accuracy in the quantum paradigm, many classical machine learning algorithms have been adapted to run on quantum computers, typically using a quantum-classical hybrid model. While some work has been done to compare classical and quantum classification algorithms in the Electro-Optical (EO) image domain, this paper will compare the performance of classical and quantum-hybrid classification algorithms in their applications on Synthetic Aperture Radar (SAR) data using the MSTAR dataset. We find that there is no significant difference in classification performance when training with quantum algorithms in ideal simulators as compared to their classical counterparts. However, the true performance benefits will become more apparent as the hardware matures.
Break
Lunch/Exhibition Break 12:10 PM - 1:40 PM
Session 6:
Panel Discussion: Machine Learning for Automatic Target Recognition
23 April 2024 • 1:40 PM - 3:10 PM EDT | National Harbor 5
Session Chair:
Peter A. Torrione, Covar, LLC (United States)
Automatic Target Recognition (ATR), traditionally rooted in predefined algorithms and rule-based systems, has long been a cornerstone in defense operations. In its conventional form, ATR relied on handcrafted features and rigid frameworks, meeting the challenges of its time. However, the landscape of modern defense scenarios demands a paradigm shift.
In response to evolving complexities, ATR is seamlessly transitioning into the realm of artificial intelligence (AI), embracing a future marked by innovation and adaptability. The traditional rule-based approaches are giving way to dynamic, data-driven methodologies empowered by AI. This shift is not merely a technological upgrade; it represents a strategic move to tackle the intricate challenges of contemporary defense.
The integration of Transformer-based architectures in ATR reflects a fundamental departure from the limitations of predefined algorithms. This evolution is particularly notable in tasks requiring nuanced understanding, such as anomaly detection and trajectory prediction. Moreover, explainable AI (XAI) techniques are becoming integral, ensuring accuracy, transparency, and user trust in ATR systems.
Looking forward, the trajectory of ATR is shaped by the promise of Generative Adversarial Networks (GANs) addressing data scarcity and Quantum Machine Learning optimizing high-dimensional analyses. Academic pursuits like federated learning and meta-learning provide the intellectual backbone for a future-ready ATR landscape.
This narrative unfolds at the intersection of tradition and innovation, where the definition of ATR expands beyond its conventional boundaries, ushering in an era where AI becomes the linchpin in meeting the challenges of modern defense.
Moderators:
Asif Mehmood, Chief Digital and Artificial Intelligence Office (United States)
Panelists:
Zhu Li University of Missouri (United States)
Shuvra Bhattacharyya University of Maryland (United States)
Edmund Zelnio Air Force Research Lab. (United States)
Peter A. Torrione Covar, LLC (United States)
In response to evolving complexities, ATR is seamlessly transitioning into the realm of artificial intelligence (AI), embracing a future marked by innovation and adaptability. The traditional rule-based approaches are giving way to dynamic, data-driven methodologies empowered by AI. This shift is not merely a technological upgrade; it represents a strategic move to tackle the intricate challenges of contemporary defense.
The integration of Transformer-based architectures in ATR reflects a fundamental departure from the limitations of predefined algorithms. This evolution is particularly notable in tasks requiring nuanced understanding, such as anomaly detection and trajectory prediction. Moreover, explainable AI (XAI) techniques are becoming integral, ensuring accuracy, transparency, and user trust in ATR systems.
Looking forward, the trajectory of ATR is shaped by the promise of Generative Adversarial Networks (GANs) addressing data scarcity and Quantum Machine Learning optimizing high-dimensional analyses. Academic pursuits like federated learning and meta-learning provide the intellectual backbone for a future-ready ATR landscape.
This narrative unfolds at the intersection of tradition and innovation, where the definition of ATR expands beyond its conventional boundaries, ushering in an era where AI becomes the linchpin in meeting the challenges of modern defense.
Moderators:
Asif Mehmood, Chief Digital and Artificial Intelligence Office (United States)
Panelists:
Zhu Li University of Missouri (United States)
Shuvra Bhattacharyya University of Maryland (United States)
Edmund Zelnio Air Force Research Lab. (United States)
Peter A. Torrione Covar, LLC (United States)
Break
Coffee Break 3:10 PM - 3:40 PM
Session 7:
Radar Frequency and Synthetic Aperture Radar Automatic Target Recognition II
23 April 2024 • 3:40 PM - 5:00 PM EDT | National Harbor 5
Session Chair:
Kristen Jaskie, Prime Solutions Group, Inc. (United States)
13039-22
23 April 2024 • 3:40 PM - 4:00 PM EDT | National Harbor 5
Show Abstract +
This paper summarizes our work in alleviating the vulnerability of neural networks for Synthetic Aperture Radar (SAR) Automatic Target Recognition (ATR) to adversarial perturbations. We propose an approach of robust SAR image classification that integrates Bayesian Neural Networks (BNNs) to harness epistemic uncertainty for distinguishing between clean and adversarially manipulated SAR images. Additionally, we introduce a visual explanation method that employs a probabilistic variant of Guided Backpropagation (GBP) specifically adapted for BNNs. This method generates saliency maps highlighting critical pixels, thereby aiding human decision-makers in identifying adversarial scatterers within SAR imagery. Our experiments demonstrate the effectiveness of our approach in maintaining high True Positive Rates (TPR) while limiting False Positive Rates (FPR), and in accurately identifying adversarial scatterers, showcasing our method's potential to enhance the reliability and interpretability of SAR ATR systems in the face of adversarial threats.
13039-23
23 April 2024 • 4:00 PM - 4:20 PM EDT | National Harbor 5
Show Abstract +
We compare the effectiveness of using a trained-from-scratch, unsupervised deep generative Variational Autoencoder (VAE) model as a solution to generic representation learning problems for Synthetic Aperture Radar (SAR) data as compared to the more common approach of using an Electric Optical (EO) transfer learning method. We find that a simple, unsupervised VAE training framework outperforms an EO transfer learning model at classification.
13039-24
23 April 2024 • 4:20 PM - 4:40 PM EDT | National Harbor 5
Show Abstract +
The lack of large, relevant labeled datasets for SAR ATR poses a challenge for deep neural network approaches. Transfer learning offers promise to train on a data rich source domain and fine tune a model on a data poor target domain. Here, we apply a set of model and dataset transferability analysis techniques to investigate the efficacy of transfer learning for SAR ATR. We use multiple neural network models trained on different source SAR datasets to test the insights of these transferability analysis techniques.
13039-25
23 April 2024 • 4:40 PM - 5:00 PM EDT | National Harbor 5
Show Abstract +
In this paper, we first provide an overview of explainability and interpretability techniques introducing their concepts and the insights they produce. Next we summarize several methods for computing specific approaches to explainability and interpretability as well as analyzing their outputs. Finally, we demonstrate the application of several attribution map methods and apply both attribution analysis metrics as well as localization interpretability analysis to six neural network models trained on the SAMPLE dataset to illustrate the insights these methods offer for analyzing SAR ATR performance.
Poster Session
23 April 2024 • 6:00 PM - 7:30 PM EDT | Potomac C
Conference attendees are invited to attend the symposium-wide poster session on Tuesday evening. Come view the SPIE DCS posters, enjoy light refreshments, ask questions, and network with colleagues in your field. Authors of poster papers will be present to answer questions concerning their papers. Attendees are required to wear their conference registration badges to the poster session.
Poster Setup: Tuesday 12:00 PM - 5:30 PM
Poster authors, view poster presentation guidelines and set-up instructions at http://spie.org/DCSPosterGuidelines.
Poster Setup: Tuesday 12:00 PM - 5:30 PM
Poster authors, view poster presentation guidelines and set-up instructions at http://spie.org/DCSPosterGuidelines.
13039-26
On demand | Presented live 23 April 2024
Show Abstract +
Common automatic target recognition (ATR) algorithms based on convolutional neural network (CNN) and detection transformer (DETR) architectures have shown diminished performance when inferencing on data containing anomalies outside of the original training dataset. This papers seeks to characterize the comparative performance impact of common optical sensor calibration artifacts as they affect industry-standard CNN and DETR-based object detectors identifying targets in satellite imagery.
Symposium Plenary on AI/ML + Sustainability
24 April 2024 • 8:30 AM - 10:00 AM EDT | Potomac A
Session Chairs:
Latasha Solomon, DEVCOM Army Research Lab. (United States), Ann Marie Raynal, Sandia National Labs. (United States)
View Full Details: spie.org/dcs/plenary-ai-ml-sustainability
Welcome and opening remarks
24 April 2024 • 8:30 AM - 8:40 AM EDT
Army intelligence data and AI in modern warfare (Plenary Presentation)
Presenter(s): David Pierce, U.S. Army Intelligence (United States)
24 April 2024 • 8:40 AM - 9:20 AM EDT
FUTUR-IC: A three-dimensional optimization path towards building a sustainable microchip industry (Plenary Presentation)
Presenter(s): Anu Agarwal, Massachusetts Institute of Technology, Microphotonics Ctr. and Materials Research Lab. (United States)
24 April 2024 • 9:20 AM - 10:00 AM EDT
Artificial Intelligence and Deep Learning: Joint Session with Conferences 13039 and 13046
24 April 2024 • 1:20 PM - 3:20 PM EDT | National Harbor 2
Session Chairs:
Michael T. Eismann, Air Force Research Lab. (United States), Kenny Chen, Lockheed Martin Missiles and Fire Control (United States)
13039-27
24 April 2024 • 1:20 PM - 1:40 PM EDT | National Harbor 2
Show Abstract +
We present a streamlined pipeline that generates a YOLO object detection application using MATLAB and NVIDIA hardware. The application utilizes MATLAB’s GPU Coder toolbox and NVIDIA TensorRT to accelerate inferencing on NVIDIA processors, specifically the latest Jetson Orin embedded processor. We evaluated the object detector on the open U.S. Army Automated Target Recognition (ATR) Development Image Dataset (ADID) for multi-class vehicle detection and classification. Overall, this workflow decreases development time over traditional approaches and provides a quick route to low-code deployment on the latest NVIDIA Jetson Orin. This work offers value to researchers and practitioners aiming to harness the power of NVIDIA processors for rapid, efficient object detection solutions.
13046-42
24 April 2024 • 1:40 PM - 2:00 PM EDT | National Harbor 2
Show Abstract +
Traditionally, computer vision systems, like object detection, primarily relied on supervised learning and predetermined object categories. However, this approach's limitations in terms of generality and the need for additional labeled data are more pronounced for infrared images due to the difficulty of obtaining training datasets. In contrast, the rise of contrastive vision-language models, such as CLIP, has transformed the field. These models, pre-trained on vast image-text pairs, offer more versatile visual representations aligned with rich language semantics. CLIP's feature transferability has become a foundation for various visible image tasks. This paper introduces zero-shot object detection for infrared images using pre-trained vision and language models, extending CLIP's benefits to this domain. Experimental results show the promise of this approach, and the paper initiates a preliminary exploration of domain shift issues between infrared and visible images.
13046-43
24 April 2024 • 2:00 PM - 2:20 PM EDT | National Harbor 2
Show Abstract +
The availability of high power mobile processors designed for embedded products featuring multiple compute cores including CPUs, GPUs, DSPs and ISPs enables system developers the ability to integrated software capabilities never possible on embedded processors. Mobile processors from Qualcomm and NVIDIA now feature 50 TOPS compute power while operating on as low as 5 watts. Teledyne FLIR will describe the creation of libraries compiled to run Qualcomm Open CL and NVIDIA CUDA based hardware.
13039-28
24 April 2024 • 2:20 PM - 2:40 PM EDT | National Harbor 2
Show Abstract +
Automatic target recognition (ATR) algorithms that rely on machine learning approaches are limited by the quality of the training dataset and the out-of-domain performance. The performance of a two-step ATR algorithm that on fusing thermal imagery with environmental data is investigated using thermal imagery containing buried and surface object collected in New Hampshire, Mississippi, Arizona, and Panama. A autoencoder neural network is used to encode the salient environmental conditions for a given climatic condition into an environmental feature vector. The environmental feature vector allows for the inclusion of environmental data with varying dimensions to robustly treat missing data. Using this architecture, we evaluate the performance of the two-step ATR on a test dataset collected in an unseen climatic condition, e.g., tropical wet climate when the training dataset contains imagery collected in a similar condition, e.g., subtropical, and dissimilar climates. Lastly, it is shown that performance for out-of-domain climates can be further improved by incorporating physics-based synthetic data into the training dataset.
13039-29
24 April 2024 • 2:40 PM - 3:00 PM EDT | National Harbor 2
Show Abstract +
Traditional data collects of high priority targets require immense planning and resources. When novel operating
conditions (OCs) or imaging parameters need to be explored, typically synthetic simulations are leveraged.
While synthetic data can be used to assess automatic target recognitions (ATR) algorithms; some simulation
environments may inaccurately represent sensor phenomenology. To levitate this issue, a scale model approach is
utilized to provide accurate data in a laboratory setting. This work demonstrates the effectiveness of a resource
cognizant approach for collecting IR imagery suitable to assessing ATR algorithms. A target of is interest is
3D printed at 1/60th scale with a commercial printer and readily available materials. The printed models are
imaged with a commercially available IR camera in a simple laboratory setup. The collected imagery is used to
test ATR algorithms when trained on a standard IR ATR dataset; the publicly available ARL Comanche FLIR
dataset. The performance of the selected ATR algorithms when given sampled of scale model data is compared
to the performance of the same algorithms when using the provided measured data
13046-45
24 April 2024 • 3:00 PM - 3:20 PM EDT | National Harbor 2
Show Abstract +
The ability to accurately ascertain an observer’s position directly from imaged scenery is an important technological capacity, especially in light of the susceptibility of global positioning system (GPS) signals to interference. Horizon matching is a technique that shows promise as a method of self-localization in regions with sufficiently rich topography. By utilizing preexisting digital elevation data, simulated imagery of horizon contours and other landscape features is first generated. Real imagery is then processed to extract these features, and they are compared and analytically fitted to simulation in order to back-calculate the imager position. Here, we demonstrate this functionality with a vehicle-integrated, gimbal mounted multi-band imaging platform that facilitates platform geopositioning with visible, NIR, SWIR, MWIR and LWIR imagers. We compare the benefits and limitations of these bands, evaluating their localization accuracies given competing range performance, image resolution, and target-to-background contrast.
Program Committee
Megan King
U.S. Army Combat Capabilities Development Command (United States)
Program Committee
Jason P. Luck
Lockheed Martin Missiles and Fire Control (United States)
Program Committee
Anurag Paul
PlusAI, Inc. (United States)
Program Committee
Inderjot Singh Saggu
PlusAI, Inc. (United States)
Program Committee
ESPOL Polytechnic Univ. (Ecuador), Vintra Inc. (United States), Univ. Autònoma de Barcelona (Spain)
Additional Information
View call for papers
What you will need to submit:
- Presentation title
- Author(s) information
- Speaker biography (1000-character max including spaces)
- Abstract for technical review (200-300 words; text only)
- Summary of abstract for display in the program (50-150 words; text only)
- Keywords used in search for your paper (optional)
- Check the individual conference call for papers for additional requirements (i.e. extended abstract PDF upload for review or instructions for award competitions)