[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-122447-en":3,"doc-seo-122447-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},122447,1374391974564,"Clementine","https://ap-avatar.wpscdn.com/avatar/14000253aa45c000a9e?x-image-process=image/resize,m_fixed,w_180,h_180&k=1779874745381141002",8,"Research & Report","Machine Learning Models for Improved Tracking from Range-Doppler Map Images","Accurate statistical tracking relies on trustworthy target measurements and well-calibrated uncertainty estimates. This work introduces machine learning models for target detection and uncertainty estimation directly from range-Doppler map (RDM) images for GMTI (Ground Moving Target Indicator) radars. The detector (a UNet) improves performance over a CFAR baseline, while a conditional variational autoencoder (CVAE) learns the output distribution conditioned on UNet features. Using these model outputs within a multiple hypothesis tracker significantly improves multi-target air-to-ground tracking in complex scenarios.","Machine Learning Models for Improved Tracking from Range-Doppler Map Images  \n1st Elizabeth Hou Systems & Technology Research (STR) Arlington, VA 22203, USA [elizabeth.hou@str.us](elizabeth.hou@str.us)  \n2nd Ross Greenwood Systems & Technology Research (STR) Woburn, MA 01801, USA [ross.greenwood@str.us](ross.greenwood@str.us)  \n3rd Piyush Kumar Systems & Technology Research (STR) Woburn, MA 01801, USA  \npiyush.kumar@str.us  \narXiv :2407 .03 140v 1 [ cs .CV] 3 Jul 2024  \nAbstract—Statistical tracking filters depend on accurate target measurements and uncertainty estimates for good tracking performance. In this work, we propose novel machine learning models for target detection and uncertainty estimation in range-Doppler map (RDM) images for Ground Moving Target Indicator (GMTI) radars. We show that by using the outputs of these models, we can significantly improve the performance of a multiple hypothesis tracker for complex multi-target air-toground tracking scenarios.  \nIndex Terms—neural networks, target detection, uncertainty estimation, GMTI radars, multi-target tracking  \nI. INTRODUCTION  \nMachine learning (ML) systems provide improved accuracy in many data-rich domains like object detection. However, these systems often fail to accurately propagate uncertainty to their outputs given a faithful characterization of the typical input noise – a feat which traditional statistics-based models in these domains may accomplish straightforwardly. In applications where accurate representation of uncertainties in outputs is crucial e.g., a statistics-based estimator such as a Kalman filter, the superior accuracy of such ML systems is not useful due to the lack of reliable estimates of their outputs’ variability. In this paper, we develop a novel detection system consisting of coupled machine learning models: i) a UNet based neural network architecture that gives improved target detection performance (w.r.t. CFAR baseline) in simulated air-toground scenarios, and ii) a conditional variational autoencoder (CVAE) based neural network that estimates the uncertainties of the UNet model’s predictions. Rather than propagate input uncertainty through the UNet model layers or evaluate UNet multiple times for each input to build sample statistics, the CVAE learns the output distribution conditioned on the UNet’s penultimate layer features. This allows us to predict the output distribution using a single new sample (from an intrinsic input distribution), whereas uncertainty propagation and direct sampling do not 1. We show that by replacing traditional methods with our more powerful machine learning models ina detection and tracking pipeline, we are able to significantly  \nSupported by the Defense Advanced Research Projects Agency (DARPA) under contract number HR00112290111 . Any opinions, findings and conclusions expressed in this material are those of the authors and do not necessarily reflect the views of DARPA.  \n1In the latter cases, many samples are needed to estimate input distribution for propagation or feed through the UNet to generate output samples.  \nimprove the performance of the tracker in complex multi-target tracking scenarios.  \nA. Related Work  \nDetecting targets in range-Doppler map (RDM) images from airborne radars is a relatively “niche” domain limited to mostly defense applications. Unlike more mainstream domains such as electro-optical (EO) or even infrared (IR) where there are prolific amounts of literature (from the machine learning community) on object detection in RGB images, there is very little published literature on machine learning models for detection in RDM images. The authors in [14] and [15] both use convolutional neural networks (CNNs) to classify humans from other types of objects ([e.g. cars](e.g. cars), dogs, drones) in RDM images. And, the authors in [16] and [17] also use CNNsto classify human activities from micro-Doppler and rangeDoppler signatures. However, even in these other works, they are focused","cbCaiqUtMhv9tKvg","https://ap.wps.com/l/cbCaiqUtMhv9tKvg","pdf",1877182,1,9,"English","en",105,"# Introduction\n## Related Work\n## Outline\n# Problem Definition\n## State and Measurement Modeling","[{\"question\":\"Why is uncertainty propagation important for tracking performance?\",\"answer\":\"Tracking accuracy depends on faithful uncertainty estimates tied to target measurements. If output variability is not reliably represented, even a highly accurate detection model may not improve tracking when uncertainty is required.\"},{\"question\":\"What machine learning models are proposed for RDM-based GMTI tracking?\",\"answer\":\"The approach couples a UNet-based detector for improved target detection with a CVAE-based network that estimates the uncertainty of the detector’s predictions.\"},{\"question\":\"How does the CVAE reduce the need for multiple samples during uncertainty estimation?\",\"answer\":\"Instead of propagating uncertainty through all UNet layers or repeatedly sampling outputs, the CVAE learns the conditional output distribution based on UNet’s penultimate-layer features, enabling uncertainty prediction using a single new sample.\"}]","Machine Learning Models for Improved Tracking from Range-Doppler Map Images | PDF",1785810677,23,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"machine-learning-models-for-improved-tracking-from-range-doppler-map-images","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/machine-learning-models-for-improved-tracking-from-range-doppler-map-images/122447/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-04",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why is uncertainty propagation important for tracking performance?","Question",{"text":75,"@type":76},"Tracking accuracy depends on faithful uncertainty estimates tied to target measurements. If output variability is not reliably represented, even a highly accurate detection model may not improve tracking when uncertainty is required.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What machine learning models are proposed for RDM-based GMTI tracking?",{"text":80,"@type":76},"The approach couples a UNet-based detector for improved target detection with a CVAE-based network that estimates the uncertainty of the detector’s predictions.",{"name":82,"@type":73,"acceptedAnswer":83},"How does the CVAE reduce the need for multiple samples during uncertainty estimation?",{"text":84,"@type":76},"Instead of propagating uncertainty through all UNet layers or repeatedly sampling outputs, the CVAE learns the conditional output distribution based on UNet’s penultimate-layer features, enabling uncertainty prediction using a single new sample.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,127,130,134],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":21,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]