Living by water characterises the development of civilisations since ancient times. Water, enveloping 71% of the Earth’s surface, sustains a diverse range of aquatic life, harbours the enigmas of life’s beginnings, and chronicles environmental transformations and ecological rhythms across time. Today, access to clean drinking water stands as pivotal for human wellbeing, enhancing livelihoods and advancing progress. Aquatic organisms serve as a lens into the aquatic world, offering glimpses into animal behaviour and ecology. This helps reveal the complex interplays between organisms and their surroundings, unveiling how they adapt to and shape their environments. With technological advancements, scientists now employ advanced behavioural tracking tools to scrutinise these dynamics.
Led by Professor Ke Wei from the Faculty of Applied Sciences, PhD students Wu Zewei, Wang Cui and Zhang Wei in Computer Applied Technology collaborated with researchers from Beihang University and Beijing Forestry University. This collaborative effort culminated in the creation of a cutting-edge dual-camera system designed to monitor the actions of aquatic organisms in a real-time, three-dimensional setting. This system facilitates comprehensive collection of data on aquatic animal behaviour, laying a robust groundwork for advancing research in ecological preservation and environmental stewardship.
Three-dimensional Imaging for Comprehensive Aquatic Monitoring
Observing the behaviours of aquatic life, such as their movements, feeding patterns, and reproductive behaviours, is essential for evaluating their wellbeing and habitat quality. Timely detection of anomalies is crucial for enabling prompt interventions and remedial actions. Conventional underwater monitoring systems, typically reliant on two-dimensional imagery captured from a single angle, encounter limitations due to constrained perspectives and potential visual obstructions, hindering thorough behaviour tracking and accurate analyses.
The research team introduces an innovative dual-camera tracking system with demonstrated precision and dependability for studying zebrafish behaviour. Working in harmony, akin to synchronised dancers, the cameras delve into the underwater realm, capturing the nuanced motions of zebrafish from multiple angles simultaneously. This approach significantly mitigates issues related to visual obstructions and provides a robust basis for precise tracking.
The effectiveness of the dual-camera system is primarily due to its advanced object detection algorithms and early reconstruction strategies. The research team created a fish head detector that functions with the precision of a keen-eyed detective, adeptly pinpointing target zebrafish in video sequences. The early reconstruction strategy operates like a skilled sculptor, swiftly reconstructing the three-dimensional coordinates of the zebrafish during tracking. This enhances the real-time accuracy of behaviour tracking, showcasing the impressive capabilities of modern technology and the meticulous nature of scientific inquiry.
To further enhance the accuracy and reliability of the tracking process, the research team incorporated a multi-view contrastive learning framework alongside a Kalman filter. The contrastive learning framework serves as a bridge that connects observations from different camera angles, enhancing the precision of image matching. The Kalman filter, acting like an astute prophet, streamlines the fish movement model, effectively estimating the state of the three-dimensional targets. This integration fortifies the predictive and corrective functions of the tracking system, adeptly tackling technical hurdles and elevating the system’s overall performance.
Aquatic Ecological Health and Pollution Monitoring
Ecological health encompasses the wellbeing of human habitats, biological environments, and metabolic systems, all of which have profound impacts on human life. The extensive use of pesticides, production and incineration of plastics, discharges of industrial and urban sewage, and frequent chemical spills pose significant threats to aquatic ecosystems and their inhabitants. Human-induced climate change results in elevated water temperatures, diminishing dissolved oxygen levels, and amplifying the metabolic rates of aquatic organisms. These factors heighten the toxicity of micropollutants such as heavy metals, adversely affecting the feeding, growth, and detoxification capabilities of aquatic animals.
Fish, as higher-order members of aquatic ecosystems, serve as robust indicators of changes in water conditions, which are often reflected in their behaviour. Monitoring fish behaviour has become a vital global technique for assessing the health of aquatic ecosystems. Changes in fish behaviour, including variations in swimming speed, proximity to pollution sources, and surface activities, serve as valuable indicators for assessing water quality and the outcome of aquatic ecology restoration initiatives. For example, zebrafish exhibit heightened sensitivity to environmental pollutants, displaying noticeable behavioural changes in reaction to toxic substances like arsenic and manganese. The dual-camera tracking system enhances the accuracy and efficiency of fish behaviour monitoring. By capturing and analysing real-time behavioural data, this system not only indicates the current health status of the aquatic environment but also provides a scientific foundation and early alert system for endeavours aimed at restoring aquatic ecology.
Intelligentisation of Aquaculture Industry
In the aquaculture industry, detecting abnormal behaviour is a key method for safeguarding the wellbeing of aquatic animals. Fluctuations in temperature, dissolved oxygen levels, and parasitic infestations can significantly alter the behaviour of aquaculture species. Researchers utilise extensive video and image datasets to track the whereabouts and movement patterns of these creatures, analysing their behaviour to detect irregularities. This information is processed using methods such as similarity measurements and machine learning techniques to pinpoint unusual behaviours. For instance, meticulous analysis of fish behaviour before, during, and after feeding times is vital for the early detection of hunger states or stress reactions.
China stands as a global frontrunner in aquaculture, contributing 65% of the global output. Over the past two decades, the country’s aquaculture output has doubled, catering to approximately one-third of the high-quality animal protein consumed by its populace. Upholding safety and health standards in this growing sector is of paramount importance. Traditional aquaculture approaches face challenges such as low production efficiency, increasing environmental pressures, and high cultivation risks. These challenges are further exacerbated by rising labour costs and an aging workforce. Transitioning to more efficient, eco-friendly, precise, and intelligent aquaculture practices is crucial for the sustainable development of aquaculture industry. The incorporation of dual-camera systems in aquaculture monitoring can substantially ease the labour burden on aquaculture practitioners, increase production efficiency, and diminish environmental footprints, which are essential for a sustainable future in aquaculture.
Enhancing Animal Welfare through Scientific Advancement
There is a growing international emphasis on animal welfare, prompting the scientific community to address ethical dilemmas related to the suffering animals may endure in experiments, as well as the significant time and resources involved. This has led to a gradual shift towards alternatives to using animals in research. Scientists now widely advocate for the 3Rs principle – Replacement, Reduction, and Refinement – aiming to minimising animal use while maximising the ethics and efficiency of experiments, thereby steering scientific research towards more sustainable practices.
Zebrafish have emerged as a prominent substitute to traditional laboratory animals due to their distinctive physiological traits and suitability for a range of experiments, earning them the moniker “aquatic laboratory rats.” While their genetic makeup is about 70% similar to humans, their physiological systems and sensory capacities are simpler, rendering them excellent models for investigating human diseases. As such, there is an increasing demand for precise monitoring of zebrafish in scientific research, particularly in the field of biomedicine. The utilisation of dual-camera systems to observe zebrafish and collect behavioural data enhances the development of genetic disease models and facilitates research on gene function and developmental biology. This approach enriches the breadth of experimental data and insights across disciplines such as cardiology, oncology, neuroscience, behavioural science, and ecotoxicology. It also enables researchers to quickly assess the impact of various compounds on zebrafish larvae behaviour, neural development, and overall function, thereby advancing a more ethical exploration of neurological disorders, cardiovascular diseases, and cancer. Zebrafish are prolific breeders, yielding over 200 fertilised eggs per cycle, supporting large-scale experiments and minimising potential errors.
The advanced underwater behavioural tracking system endorses more ethical research practices, aligning with the feasibility and benefits of using zebrafish as an alternative in research. Minimising harm to animals while enhancing the precision and efficiency of experiments, this approach creates a mutually beneficial scenario that promotes both the progression of scientific research and the coexistence of contemporary scientific pursuits with academic ethics.
Expertise in AI and Computer Vision
Professor Ke Wei, Coordinator of the PhD programme in Computer Applied Technology and a core researcher of the Ministry of Education Engineering Research Centre of Applied Technology on Machine Translation and Artificial Intelligence, boasts nearly two decades of experience in teaching and research. His significant contributions span the realms of artificial intelligence, computer vision, and natural language processing, with over 140 papers published in international academic journals. Notably, his work on enhancing the precision and stability of image recognition has been featured in esteemed publications such as Advanced Engineering Informatics.
Professor Ke stands steadfast in advancing technological innovation and fostering academic exchange. He actively cultivates interdisciplinary partnerships and secures academic resources to support research endeavours. His recent project, “Open Learning Theories and Methods for Humanoid Vision,” conducted in collaboration with three universities, has achieved breakthroughs in computer vision and deep learning theories. This project secured joint funding from the National Natural Science Foundation of China and the Macao Science and Technology Development Fund. As a professor in Computer Science, he passionately nurtures emerging talent, a commitment that has earned him a “National Teaching Achievement Award.” Additionally, Professor Ke is deeply involved in marine conservation research, leveraging his expertise to drive advancements in this field. His work is pivotal in the behavioural analysis of aquatic organisms and the evaluation of ecosystem health, providing critical scientific insights and technical support for the conservation of aquatic life and the preservation of ecosystem diversity.
Twisting the Magnetic Dance
Advancements in Nano-magnetism Propel Semiconductor Industry