MambaAlign fusion framework for detecting defects missed by inspection systems

Researchers develop an efficient system that detects subtle defects missed by existing industrial visual inspection systems

Shibaura Institute of Technology

FacebookXLinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

Researchers propose a new alignment-aware state-space fusion framework called MambaAlign that produces tighter, less fragmented anomaly maps, and is substantially more robust to modest misalignment than prior fusion or attention-heavy approaches

view more 

Credit: Dr. Phan Xuan Tan from Shibaura Institute of Technology, Japan, and Dr. Dinh-Cuong Hoang from FPT University, Vietnam Source link: https://academic.oup.com/jcde/article/13/1/514/8405688

Industrial quality inspection plays a critical role in manufacturing, from ensuring the reliability of electronics and vehicles to preventing costly failures in aerospace and energy systems. Traditional vision-based inspection systems typically rely on Red, Green, Blue (RGB) cameras, which are fast and inexpensive but often miss defects related to geometry (scratches or dents), material structure, or heat dissipation. While additional sensors, such as thermal cameras or depth scanners, can reveal these hidden anomalies, effectively combining information from multiple sensors remains a major technical challenge. Many existing fusion approaches either lose fine spatial detail, require heavy computation, or fail when sensors are not perfectly aligned—common issues in factory settings.

To address these, a research team led by Associate Professor Phan Xuan Tan from the Innovative Global Program, College of Engineering, Shibaura Institute of Technology, Japan, along with Dr. Dinh-Cuong Hoang from the FPT University, Vietnam, has proposed a new framework, termed MambaAlign, which enables computationally efficient fusion of multimodal sensor data while remaining robust to modest sensor misregistration. The study was made available online on December 27, 2025, and was published in Volume 13, Issue 1 of the Journal of Computational Design and Engineering on January 01, 2026.

“Existing systems miss geometric and material/thermal defects, amplify sensor artifacts, lose localization, or are brittle to modest misregistration. In addition, efficiently capturing long-range, orientation-sensitive context (important for thin/oblique defects) without the quadratic cost of dense attention remained unresolved. These challenges of existing systems motivated us to develop a fusion approach that is alignment aware, uses state-space recurrences to collect long-range directional context, and exchanges semantic guidance at deep stages via lightweight cross-recurrence (Cross Mamba Interaction), and then reconstitutes low-level channels top-down to preserve precise localization,” says Dr. Tan.

MambaAlign introduces an alignment-aware state-space fusion framework for multimodal industrial anomaly detection. The method captures long-range and orientation-aware context using state-space refinement, which is particularly effective for detecting thin or oblique defects such as scratches and cracks. Instead of relying on computationally expensive global attention, MambaAlign exchanges semantic guidance between sensors only at high-level feature stages, keeping the computational cost close to linear. A top-down reconstruction mechanism then reconstitutes low-level feature channels, allowing the system to tolerate modest sensor misalignment while preserving precise pixel-level localization.

Extensive experiments demonstrate the effectiveness of the approach. Averaged across three RGB-plus-auxiliary-modality (RGB-X) datasets, MambaAlign improves image-level area under the receiver operating characteristic curve (AUROC) by approximately 4.8%, pixel-level AUROC by about 5.0%, and area under the per-region overlap curve by roughly 6.5% compared with prior methods. Importantly, these gains come without excessive computational overhead. The model sustains close to 30 frames per second at moderate resolutions, with controlled memory usage, making it practical for deployment in real production lines.

“MambaAlign achieves state-of-the-art localization with parameters and runtime suitable for real-time inspection. It not only provides higher detection accuracy but also tighter and less fragmented anomaly maps. This translates directly into fewer false alarms, fewer missed defects, and more actionable outputs for engineers on the factory floor,” says Dr. Tan.

Overall, the study highlights wide-ranging industrial relevance. In electronics and printed circuit board inspection, MambaAlign can detect micro-cracks or missing components that subtly alter thermal or geometric patterns. In aerospace and composite manufacturing, fusing RGB and thermal data helps reveal subsurface delamination invisible to standard cameras. Automotive body inspection benefits from improved detection of dents, scratches, and seam defects, while the system’s real-time performance enables inline inspection on conveyor belts or robotic vision stations. By reducing manual inspection effort, minimizing scrap, and improving reliability under realistic sensor conditions, MambaAlign addresses a long-standing bottleneck in industrial quality assurance.

 

Reference

DOI: https://doi.org/10.1093/jcde/qwaf143

 

About Shibaura Institute of Technology (SIT), Japan

Shibaura Institute of Technology (SIT) is a private university with campuses in Tokyo and Saitama. Since the establishment of its predecessor, Tokyo Higher School of Industry and Commerce, in 1927, it has maintained “learning through practice” as its philosophy in the education of engineers. SIT was the only private science and engineering university selected for the Top Global University Project sponsored by the Ministry of Education, Culture, Sports, Science and Technology and had received support from the ministry for 10 years starting from the 2014 academic year. Its motto, “Nurturing engineers who learn from society and contribute to society,” reflects its mission of fostering scientists and engineers who can contribute to the sustainable growth of the world by exposing their over 9,500 students to culturally diverse environments, where they learn to cope, collaborate, and relate with fellow students from around the world.

Website: https://www.shibaura-it.ac.jp/en/

 

About Associate Professor Phan Xuan Tan from SIT, Japan

Dr. Phan Xuan Tan is an Associate Professor in the Innovative Global Program, College of Engineering, Shibaura Institute of Technology (SIT), Japan. He earned a B.E. in Electrical-Electronic Engineering from Le Quy Don Technical University, and an M.S. in Computer and Communication Engineering from Hanoi University of Science & Technology, Vietnam. He received his Ph.D. in Functional Control Systems from SIT in 2018. His academic work bridges engineering and artificial intelligence (AI), with research centered on computer vision, image processing, generative AI, and AI safety.

---

Journal

Journal of Computational Design and Engineering

DOI

10.1093/jcde/qwaf143 

Method of Research

Computational simulation/modeling

Subject of Research

Not applicable

Article Title

MambaAlign: Alignment-aware state-space fusion for RGB-X industrial anomaly detection

 

How horses whinny: Whistling while singing

Cell Press

FacebookXLinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

A horse mid-whinny.

view more 

Credit: Elodie Briefer

A horse’s whinny is an unusually distinctive mix of sounds including both high and low frequencies. Reporting in the Cell Press journal Current Biology on February 23, researchers demonstrate how horses produce high-frequency sounds that defy their large size while simultaneously producing lower tones: they whistle through their larynx while vibrating their vocal folds as a human does while singing. Horses likely evolved these vocalizations to be able to convey multiple messages to one another at the same time, says the team.  

“We now finally know how the two fundamental frequencies that make up a whinny are produced by horses,” says author Elodie Briefer of the University of Copenhagen. “In the past, we found that these two frequencies are important for horses, as they convey different messages about the horses’ own emotions. We now have compelling evidence that they are also produced through distinct mechanisms." 

While domesticated horses have lived closely with humans for over 4,000 years, horse vocal communication is poorly understood. Larger mammals generally make lower-pitched sounds because the size of the larynx typically increases in step with body size. But researchers have noted exceptions to this rule—and horses' whinnies are a prominent example. 

To learn more, the team sought to explore the mechanics behind horses’ whinnies. They found that the whinny represents an unusual vocal phenomenon, known as “biphonation,” in which a vocalization has two independent frequency components: low and high.  

While the low frequency is made by vocal-fold vibration, just like a human singing or a cat meowing, the origin of the high-frequency component has remained mysterious until now. To learn more about this high-frequency sound, the team gathered a wide range of data through careful study of the animals’ vocal anatomy, clinical data, and acoustic analysis.  

"Solving this biomechanical puzzle required combining approaches from veterinary medicine to acoustic physics,” says author Romain Lefèvre of the University of Copenhagen. 

They found that the high-frequency component of the whinny is generated by a laryngeal whistle. The researchers describe it as similar in principle to a normal human whistle, except that the turbulent airstream that creates the whistle sound is created within the horse's larynx. Although some small rodents like rats and mice produce laryngeal whistles, horses are the first large mammal species found to whistle in this way and the only animals known to do so simultaneously with vocal-fold vibration. 

To demonstrate this whistle, the researchers performed excised larynx experiments, blowing a stream of air through the larynx removed from deceased horses. While doing so, they switched the airstream into the larynx from air to helium and back again. They explained that because the speed of sound is higher in helium, it causes whistles to shift to higher frequencies, while the frequency of normal vocal-fold vibration remains unchanged. As predicted, the high-frequency component of the whinnies shifted upward when using helium, while the low frequencies did not change. 

"When we blew helium through the larynges for the first time, the frequency shift was immediately obvious, and we knew we'd solved the mystery,” says author William Tecumseh Fitch of the University of Vienna. “We were thrilled!" 

The new findings help to explain how the two overlapping pitches—or biphonation—occur, says the team. They suggest that horses’ biphonation likely evolved to convey multiple independent messages concurrently.  

The researchers also note that Przewalski's horses—a species closely related to the domesticated horse—also produce whinnies with biphonation, but more distant horse relatives like donkeys and zebras appear to lack the high component of the whinny, suggesting that horses have unique vocal adaptations which allow them to produce a richer and more complex spectrum of calls compared to other mammals. 

"Understanding how and why biphonation has evolved is an important step towards elucidating the origins of the amazing vocal diversity of mammalian vocal behavior,” says author David Reby of the University of Lyon/Saint-Etienne. 

### 

This work was supported by the Swiss National Science Foundation, the Austrian Science Fund, and Institut Universitaire de France. 

Current Biology, Lefèvre et al., “The high fundamental frequency in horse whinnies is generated by an aerodynamic whistle” https://www.cell.com/current-biology/fulltext/S0960-9822(26)00004-7

Current Biology (@CurrentBiology), published by Cell Press, is a bimonthly journal that features papers across all areas of biology. Current Biology strives to foster communication across fields of biology, both by publishing important findings of general interest and through highly accessible front matter for non-specialists. Visit: http://www.cell.com/current-biology. To receive Cell Press media alerts, contact press@cell.com. 

Horse whinny [AUDIO] |

Journal

Current Biology

DOI

10.1016/j.cub.2026.01.004 

Method of Research

Experimental study

Subject of Research

Animals

Article Title

The high fundamental frequency in horse whinnies is generated by an aerodynamic whistle

Article Publication Date

23-Feb-2026

 

Roadmap for Europe’s biodiversity monitoring system

New study charts how digital technologies, DNA, and coordinated governance can transform biodiversity observation across Europe and support global conservation goals.

German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig

FacebookXLinkedInWeChatBlueskyMessageWhatsAppEmail

 

image: 

Europe-wide monitoring is designed to track biodiversity change and support nature conservation. The photo shows a beech forest in the Hainich National Park, Thuringia, Germany.

view more 

Credit: Stefan Bernhardt, iDiv

Biodiversity is changing across the planet, yet governments still lack the robust, consistent data needed to track these changes and guide effective conservation. Now, a new study led by the University of Amsterdam (UvA), the German Centre for Integrative Biodiversity Research (iDiv), and the Martin Luther University Halle-Wittenberg (MLU), proposes a comprehensive roadmap to build a modern, integrated Biodiversity Observation Network (BON) for Europe – one that could become a global model for biodiversity monitoring in the 21st century. The study has been published in the journal Nature Reviews Biodiversity.

“Our proposal provides a plan for Europe to fix its messy and disconnected monitoring systems,” says lead author Dr Daniel Kissling, Associate Professor at the UvA. “We want to create one coordinated, continent-wide network that can track changes in species and ecosystems – from the DNA of plants and animals to entire forests, rivers, and oceans.”

A unified monitoring system for Europe’s biodiverse heritage

The roadmap identifies 84 Essential Biodiversity Variables (EBVs) that form the backbone of a harmonised monitoring system. These EBVs – from bird abundance and insect phenology to seagrass extent, genetic diversity, and ecosystem productivity – provide Europe with a consistent, standardised checklist for measuring the state and change of its biodiversity.

“Europe has hundreds of monitoring programmes, but the data are often siloed, incompatible, or incomplete,” says senior author Prof Henrique Pereira, research group head at iDiv and the MLU. “Our roadmap provides the architecture for a truly integrated, transnational system – one that brings all observations together into a coherent whole.”

To enable this transition, the authors propose establishing a European Biodiversity Observation Coordination Centre (EBOCC). This new EU-level body would coordinate workflows, harmonise methods, ensure transparent data governance, align monitoring with EU policy needs, and act as the central hub for national and European data infrastructures.

High-tech biodiversity monitoring with people

A key message of the roadmap is that Europe must harness the combined strengths of technological innovation and human expertise, including the potential of new digital technologies, including:

• Automated digital sensors such as acoustic bird recorders, wildlife and insect cameras, and biological and weather radars
• AI for species recognition and automated data processing
• Environmental DNA (eDNA) and metabarcoding for detecting species and communities from water, soil, or air
• State-of-the-art remote sensing from satellites (including Copernicus), aircraft, and drones to observe habitats, vegetation structure, and ecosystem change

The roadmap also highlights that people remain central to biodiversity monitoring. Citizen scientists, taxonomic experts and professional monitoring networks provide essential observations, expertise, and continuity. New technologies complement and strengthen their contributions, making biodiversity monitoring more efficient, scalable and inclusive, while ensuring that human knowledge and engagement remain fundamental to Europe’s monitoring system.

Bridging data gaps through unified workflows

Europe’s current biodiversity data are extensive but scattered. The roadmap proposes to build data pipelines that can integrate information from many different sources – like professional field notes, reports from the public, electronic sensors, DNA samples, and satellite images, and merge them into scalable EBV datasets.

These new pipelines should enable Europe to create clear reports for decision-makers, spot trends quickly, and give early warnings of ecological change.

Step forward for international biodiversity policy

The roadmap was designed by EuropaBON, a Horizon 2020 project involving 15 research organisations across Europe and has led to a strong policy response, as the European Parliament has already approved a preparatory action for the EBOCC to start implementing parts of the roadmap. The roadmap for the EBOCC is directly aligned with the EU Biodiversity Strategy for 2030, the Nature Restoration Regulation (NRR), and other major EU environmental legislation, including the Birds Directive, Habitats Directive, Water Framework Directive, and Marine Strategy Framework Directive. By delivering harmonised biodiversity data, an EBOCC would significantly improve reporting and support implementation across Member States.

Globally, the system would help track progress toward the Kunming–Montreal Global Biodiversity Framework (GBF), support assessments of the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES), and contribute to GEO BON (Group on Earth Observations Biodiversity Observation Network).

---

Journal

Nature Reviews Biodiversity

DOI

10.1038/s44358-026-00140-6 

Method of Research

Commentary/editorial

Subject of Research

Not applicable

Article Title

Building the backbone for Europe’s biodiversity monitoring

Article Publication Date

23-Feb-2026