Background-aware Siamese Network Tracking based on Salient Feature Fusion

Background-aware Siamese Network Tracking based on Salient Feature Fusion | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Background-aware Siamese Network Tracking based on Salient Feature Fusion Lifang Zhou, Jing Dai, Bangjun Lei, Ruixin Xue This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-3862057/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract In the realm of object tracking, complex backgrounds pose significant challenges, often leading to the underperformance of existing trackers that predominantly rely on deep features from the final layer of feature extraction networks. These features, while semantically rich, are not always sufficient to distinguish targets from cluttered environments. To address this limitation, we introduce the Background-aware Siamese Network (BASNet), an innovative approach that enhances the salience of features in intricate scenes. Central to our method is the Dual-Feature Fusion (DFF) model, meticulously crafted to stabilize target representation amidst distracting backgrounds. BASNet leverages the complementary strengths of deep and shallow features, harnessing the former’s semantic depth and the latter’s precision in spatial localization. This fusion not only elevates feature utilization but also mitigates the shortcomings associated with the superficial output of shallow networks. Further refining the target’s spatial definition, our attention-focusing module (AF) plays a pivotal role. It accentuates pertinent features 1 and attenuates noise emanating from the multi-layer feature amalgamation. The module’s ingenuity lies in its cyclic attention operations, which enable each pixel to establish extensive dependencies with all others , thereby spotlighting salient features. For precise target localization, BASNet employs the SIoU loss within its regression branch, a strategic choice that contributes to the model’s accuracy. Extensive experiments demonstrate the competitive performance of our method on six datasets: OTB50, OTB100, UAV123, VOT2016, VOT2018, and VOT2019. Object tracking features fusion deep learning Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-3862057","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":267209178,"identity":"51357f0b-be81-4851-8812-9e61a9f1eaf6","order_by":0,"name":"Lifang Zhou","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzElEQVRIiWNgGAWjYDACCSB+YGMD4zITqSUhLQ3EZGwgRcthErTIz24+JpGQcF6e70bu8QcMFdaJDexnD+DVwjjnWBpQy23DmTfyEhsYzqQnNvDkJeDVwiyRYyaR+ON2gsGNHMMGxrbDiQ0SPAZ4tbCBtCQknINq+UeEFh6IlgNQLQ1EaJGQSEu2SEhINpx55l3ijIRj6cZtPDn4tcjPSD5440OCnTzf8dwDHz7UWMv2s5/BrwUBDvAAIwjkOyLVQ7WMglEwCkbBKMAGAGYQRVdIOXDzAAAAAElFTkSuQmCC","orcid":"","institution":"Chongqing University of Posts and Telecommunications","correspondingAuthor":true,"prefix":"","firstName":"Lifang","middleName":"","lastName":"Zhou","suffix":""},{"id":267209179,"identity":"f814a32f-8c37-4bf8-bf2c-fd0c66334865","order_by":1,"name":"Jing Dai","email":"","orcid":"","institution":"Chongqing University of Posts and Telecommunications","correspondingAuthor":false,"prefix":"","firstName":"Jing","middleName":"","lastName":"Dai","suffix":""},{"id":267209180,"identity":"ba07b95c-aff4-4ab9-b661-e5385d4ff128","order_by":2,"name":"Bangjun Lei","email":"","orcid":"","institution":"Chongqing University of Posts and Telecommunications","correspondingAuthor":false,"prefix":"","firstName":"Bangjun","middleName":"","lastName":"Lei","suffix":""},{"id":267209181,"identity":"40c64dcf-7566-4818-9515-0a96eb43eecf","order_by":3,"name":"Ruixin Xue","email":"","orcid":"","institution":"Chongqing University of Posts and Telecommunications","correspondingAuthor":false,"prefix":"","firstName":"Ruixin","middleName":"","lastName":"Xue","suffix":""}],"badges":[],"createdAt":"2024-01-14 04:14:07","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-3862057/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-3862057/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":62323878,"identity":"7a48c16b-094d-4594-8e1a-849b73a06eb7","added_by":"auto","created_at":"2024-08-13 02:22:28","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2531768,"visible":true,"origin":"","legend":"","description":"","filename":"BASNet.pdf","url":"https://assets-eu.researchsquare.com/files/rs-3862057/v1_covered_48829657-a594-418c-a696-07f06e552bc8.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Background-aware Siamese Network Tracking based on Salient Feature Fusion","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Object tracking, features fusion, deep learning","lastPublishedDoi":"10.21203/rs.3.rs-3862057/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-3862057/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"In the realm of object tracking, complex backgrounds pose significant challenges, often leading to the underperformance of existing trackers that predominantly rely on deep features from the final layer of feature extraction networks. These features, while semantically rich, are not always sufficient to distinguish targets from cluttered environments. To address this limitation, we introduce the Background-aware Siamese Network (BASNet), an innovative approach that enhances the salience of features in intricate scenes. Central to our method is the Dual-Feature Fusion (DFF) model, meticulously crafted to stabilize target representation amidst distracting backgrounds. BASNet leverages the complementary strengths of deep and shallow features, harnessing the former’s semantic depth and the latter’s precision in spatial localization. This fusion not only elevates feature utilization but also mitigates the shortcomings associated with the superficial output of shallow networks. Further refining the target’s spatial definition, our attention-focusing module (AF) plays a pivotal role. It accentuates pertinent features 1 and attenuates noise emanating from the multi-layer feature amalgamation. The module’s ingenuity lies in its cyclic attention operations, which enable each pixel to establish extensive dependencies with all others , thereby spotlighting salient features. For precise target localization, BASNet employs the SIoU loss within its regression branch, a strategic choice that contributes to the model’s accuracy. Extensive experiments demonstrate the competitive performance of our method on six datasets: OTB50, OTB100, UAV123, VOT2016, VOT2018, and VOT2019.","manuscriptTitle":"Background-aware Siamese Network Tracking based on Salient Feature Fusion","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-01-17 16:00:27","doi":"10.21203/rs.3.rs-3862057/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"ac590c7c-8f95-40b3-a6c3-89d99fd95863","owner":[],"postedDate":"January 17th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-08-13T02:14:19+00:00","versionOfRecord":[],"versionCreatedAt":"2024-01-17 16:00:27","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-3862057","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"identity":"rs-3862057","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}