A Lightweight Monocular Visual Odometry with Depth Scale Alignment

A Lightweight Monocular Visual Odometry with Depth Scale Alignment | 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 A Lightweight Monocular Visual Odometry with Depth Scale Alignment Bingming Tong, Wei Chen, Luyao Du, Xingzhuo Yan, Fangquan Si, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7314978/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 5 You are reading this latest preprint version Abstract Monocular visual odometry (VO) suffers from inherent scale ambiguity, which limits its accuracy and robustness in navigation tasks. Although recent self-supervised end-to-end methods have shown promising results, their reliability remains insufficient for real-world deployment. To address this challenge, we propose a self-supervised monocular VO framework that estimates relative camera motion using dense optical flow and incorporates a dedicated depth estimation network for explicit scale alignment. The depth network features a hybrid convolutional neural network (CNN)–Transformer encoder and a semi-dense decoder, enabling efficient and accurate monocular depth prediction. During inference, the predicted depth maps are aligned with sparse point clouds reconstructed via two-view triangulation, allowing the system to recover absolute scale and enhance global consistency. The proposed method is evaluated on the KITTI odometry benchmark, achieving an average translational error of 3.10% and a rotational error of 0.64°/100 m. Compared with a ResNet-18-based baseline, our approach reduces the translational pose error by 41.95% on average and decreases the number of model parameters by 56.8%. Monocular Visual Odometry Self-Supervised Learning Transformer-Based Architecture Depth Alignment Optical Flow Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 11 Aug, 2025 Reviewers invited by journal 11 Aug, 2025 Editor assigned by journal 07 Aug, 2025 Submission checks completed at journal 07 Aug, 2025 First submitted to journal 07 Aug, 2025 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-7314978","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":499023673,"identity":"641be751-2af7-4a5d-b2c7-cc1d8c3121bc","order_by":0,"name":"Bingming Tong","email":"","orcid":"","institution":"Wuhan University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Bingming","middleName":"","lastName":"Tong","suffix":""},{"id":499023674,"identity":"e324b7f8-cfc4-49d7-ac7c-e4f562bf12ed","order_by":1,"name":"Wei Chen","email":"","orcid":"","institution":"Wuhan University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Wei","middleName":"","lastName":"Chen","suffix":""},{"id":499023676,"identity":"75e98093-141f-471c-9af6-be2430f76585","order_by":2,"name":"Luyao Du","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzElEQVRIiWNgGAWjYDCCA8wNIEqOQQLMZSZGCyNYizFQC4hFgpbEBqK18N1IbJPmqbFJ33C7x/wBQ4V1YgP72QN4tUiCtRxLy91w54xhA8OZ9MQGnrwEvFoMwFrYDuduu5Fj2MDYdhjoQh4DIrT8O5xuBtbyj1gtvG2HEyBaGojQInnmYbPl3L40w/030gpnJBxLN27jycGvhe948sEbb77ZyEvOSN7w4UONtWw/+xn8WkCAiQfGSgBiNoLqgYDxBzGqRsEoGAWjYOQCAJvjSjr3K0psAAAAAElFTkSuQmCC","orcid":"","institution":"Wuhan University of Technology","correspondingAuthor":true,"prefix":"","firstName":"Luyao","middleName":"","lastName":"Du","suffix":""},{"id":499023678,"identity":"8777768d-281f-4770-9d8c-59faeef76724","order_by":3,"name":"Xingzhuo Yan","email":"","orcid":"","institution":"Wuhan University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Xingzhuo","middleName":"","lastName":"Yan","suffix":""},{"id":499023679,"identity":"57eb103c-aa8a-4127-96eb-46fff2720a9b","order_by":4,"name":"Fangquan Si","email":"","orcid":"","institution":"Wuhan University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Fangquan","middleName":"","lastName":"Si","suffix":""},{"id":499023681,"identity":"b0b308df-d72c-4f4e-ab46-1f3da95bd1c6","order_by":5,"name":"Zhengpeng Xu3","email":"","orcid":"","institution":"Jiangxi Winsky Intelligent Technology Co.,Ltd","correspondingAuthor":false,"prefix":"","firstName":"Zhengpeng","middleName":"","lastName":"Xu3","suffix":""},{"id":499023685,"identity":"24d8b1e2-35a8-4e39-bc8d-78934222b865","order_by":6,"name":"Ping Gong","email":"","orcid":"","institution":"Jiangxi Winsky Intelligent Technology Co.,Ltd","correspondingAuthor":false,"prefix":"","firstName":"Ping","middleName":"","lastName":"Gong","suffix":""}],"badges":[],"createdAt":"2025-08-07 05:38:29","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7314978/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7314978/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":89036147,"identity":"a0685f95-8ade-4451-8deb-4bc1810340b5","added_by":"auto","created_at":"2025-08-14 04:05:43","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":764485,"visible":true,"origin":"","legend":"","description":"","filename":"manuscrip.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7314978/v1_covered_49d59cc1-d383-4719-a0b5-e55b846ad963.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"A Lightweight Monocular Visual Odometry with Depth Scale Alignment","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"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":"signal-image-and-video-processing","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"sivp","sideBox":"Learn more about [Signal, Image and Video Processing](http://link.springer.com/journal/11760)","snPcode":"11760","submissionUrl":"https://submission.nature.com/new-submission/11760/3","title":"Signal, Image and Video Processing","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"Monocular Visual Odometry, Self-Supervised Learning, Transformer-Based Architecture, Depth Alignment, Optical Flow","lastPublishedDoi":"10.21203/rs.3.rs-7314978/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7314978/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eMonocular visual odometry (VO) suffers from inherent scale ambiguity, which limits its accuracy and robustness in navigation tasks. Although recent self-supervised end-to-end methods have shown promising results, their reliability remains insufficient for real-world deployment. To address this challenge, we propose a self-supervised monocular VO framework that estimates relative camera motion using dense optical flow and incorporates a dedicated depth estimation network for explicit scale alignment. The depth network features a hybrid convolutional neural network (CNN)\u0026ndash;Transformer encoder and a semi-dense decoder, enabling efficient and accurate monocular depth prediction. During inference, the predicted depth maps are aligned with sparse point clouds reconstructed via two-view triangulation, allowing the system to recover absolute scale and enhance global consistency. The proposed method is evaluated on the KITTI odometry benchmark, achieving an average translational error of 3.10% and a rotational error of 0.64\u0026deg;/100 m. Compared with a ResNet-18-based baseline, our approach reduces the translational pose error by 41.95% on average and decreases the number of model parameters by 56.8%.\u003c/p\u003e","manuscriptTitle":"A Lightweight Monocular Visual Odometry with Depth Scale Alignment","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-08-14 03:41:26","doi":"10.21203/rs.3.rs-7314978/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2025-08-11T21:39:54+00:00","index":"","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-08-11T21:39:12+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-08-07T17:21:23+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-08-07T17:16:57+00:00","index":"","fulltext":""},{"type":"submitted","content":"Signal, Image and Video Processing","date":"2025-08-07T05:36:54+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"signal-image-and-video-processing","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"sivp","sideBox":"Learn more about [Signal, Image and Video Processing](http://link.springer.com/journal/11760)","snPcode":"11760","submissionUrl":"https://submission.nature.com/new-submission/11760/3","title":"Signal, Image and Video Processing","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"Springer Hybrid","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"a7fdfb62-cd91-41dc-a5f6-b128ecd61451","owner":[],"postedDate":"August 14th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-08-21T21:38:15+00:00","versionOfRecord":[],"versionCreatedAt":"2025-08-14 03:41:26","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7314978","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7314978","identity":"rs-7314978","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}