Assoc. Prof. Pavel Loskot

ZJU-UIUC Institute, China

Speech Title: Estimation Problems in Array Signal Processing

Abstract: Spatial signals are generated by spatially distributed sources, which are then simultaneously observed in multiple spatial locations. The spatial dimension allows one to assume the spatial signals to be a special case of images, however, the former are also time-varying, and they are often modeled as Gaussian random processes. The example scenarios include radiowave propagation between the transmitting and receiving antennas in wireless communication systems, multi-channel EEG measurements, remote sensing in complex environments, and also multi-projections in general tomographic reconstructions. The signal processing problems appearing in all these scenarios always involve some form of inverse problems requiring to infer the parameters of the spatial signals in order to spatially resolve and combine the signals, spatially extrapolate the signals, and extract or reconstruct information about the sources. In this talk, we will focus on reviewing the key ideas and results in array signal processing where the spatial locations of the observers are known, and the task is to infer the key parameters of the incoming time-varying 3D random field such as the direction of arrival, amplitude and frequency shift, which are then used to design beamforming schemes for antenna arrays in wireless communications. We will show that the problem formulation and the solution can differ significantly when the signals can be considered to be narrowband and when they must be considered to be wideband.

Biography : Pavel Loskot joined the ZJU-UIUC Institute, Haining, China, in January 2021 as Associate Professor after 14 years being the Senior Lecturer at Swansea University in the UK. He obtained his PhD degree in Wireless Communications from the University of Alberta in Canada, and the MSc and BSc degrees in Radioelectronics and Biomedical Electronics, respectively, from the Czech Technical University of Prague in the Czech Republic. In the past 25 years, he was involved in numerous collaborative research and development projects, and also held a number of consultancy contracts with industry. Pavel Loskot is a Senior Member of the IEEE, a Fellow of the Higher Education Academy in the UK, and the Recognized Research Supervisor of the UK Council for Graduate Education. His current research interests focus on mathematical and probabilistic modeling, statistical signal processing and classical machine learning for multi-sensor data in biomedicine, computational molecular biology, and wireless communications.

Asst. Prof. Yanglong Lu

Hong Kong University of Science and Technology, China

Speech Title: Image Compression and Denoising Using Physics-Constrained Dictionary Learning

Abstract:Image compression and denoising are crucial tasks in image processing, each presenting unique challenges and employing different techniques. In recent years, compressed sensing (CS) has emerged as a method to improve data acquisition efficiency by leveraging the sparse representation of signals. CS has found extensive applications in image compression and denoising. Dictionary learning has also been developed to enhance the compression ratio in CS by training the basis matrix with specific signal types. However, existing approaches do not optimize the measurement matrix, which determines the pixel locations to be stored, limiting the customization potential for maximizing image compression ratios. To address this limitation, this study introduces a novel approach that combines image compression and denoising using physics-constrained dictionary learning (PCDL). PCDL is a recently developed method that aims to enhance compression ratios and reconstruction accuracy by simultaneously optimizing both the measurement matrix and the basis matrix. The measurement matrix, optimized using a constrained FrameSense algorithm, plays a crucial role in indicating the pixel locations to be stored within the images. On the other hand, the basis matrix is trained using the K-SVD algorithm. By inversely estimating a sparse coefficient vector through PCDL, the original image can be reconstructed while incorporating denoising effects through a linear combination of the basis matrix and the coefficient vector. The effectiveness of PCDL in image compression and denoising tasks is demonstrated in this work. Moreover, the compression ratio is further improved by incorporating constraints that facilitate the selection of the most important regions while eliminating redundant information. The PCDL framework has been successfully applied to medical images and optical images within the context of manufacturing process monitoring.

Biography: Dr. Lu holds a Ph.D. and B.S. degrees from the Department of Mechanical Engineering at the Georgia Institute of Technology. In 2022, he joined the Hong Kong University of Science and Technology as an assistant professor, following his work as a postdoctoral research fellow at the University of Michigan. Dr. Lu was a finalist in the 2023 NSF Manufacturing Blue Sky Competition and received the ASME Computers and Information in Engineering Division (CIE) Best Ph.D. Dissertation Award in 2021. Dr. Lu's research focuses on several areas, including process modeling and monitoring in additive manufacturing, design optimization, and human health monitoring and diagnosis. In the field of additive manufacturing, one of the major challenges is the variability of build qualities. Dr. Lu and his group have developed innovative sensing techniques that integrate physical models and machine learning methods to enhance efficiency and accuracy in process monitoring. Additionally, Dr. Lu explores the application of physics-informed machine learning in human health monitoring, a domain with limited available data. This emerging direction shows promise in improving monitoring and diagnosis methods for human health.

Dr. Sergii Khlamov

Kharkiv National University of Radio Electornics, Ukraine

Speech Title: Astronomical Image Processing by the Lemur Software

Abstract: The astronomical images are made by cameras with the charge-coupled device (CCD). They can be received from the different sources, like servers, clusters, predefined series, archives or “live” (online) data streams. The astronomical images processing is focused on but not limited to the following tasks: data mining, knowledge discovery, big astronomical data processing, filtering, background alignment, brightness equalization, segmentation, classification, image recognition, object’s image detection, object’s astrometry and photometry, moving object detection, parameters determination of the object’s image and apparent motion, reference objects selection and others. The modern Lemur software of the Collection Light Technology (CoLiTec) project (https://colitec.space) was developed using the described above technologies and approaches. The Lemur software is designed to perform a sequence of the following main steps: pre-processing (astronomical information collection -> worst data rejection -> useful data extraction -> data mining -> classification -> background alignment -> brightness equalization), image processing (segmentation -> typical form analysis -> recognition patterns applying -> detection of the object’s image -> astrometry -> photometry -> objects identification -> tracks detection), knowledge discovery (Solar System objects or artificial satellites to be discovered, tracks parameters for the investigation, light curves of the variable stars, scientific reports in the international formats). The paper describes the modern features for the astronomical image processing implemented in the Lemur software. It has assisted in making over 1700 discoveries of asteroids, including 5 NEOs, 21 Trojan asteroids of Jupiter, 1 Centaur. In total it has been used in about 800 000 observations, during which five comets were discovered.

Biography: Dr. Sergii Khlamov holds a Ph.D, MSc and BSc degrees with honors at the Kharkiv National University of Radio Electronics, Ukraine, where continues working. The Ph.D dissertation title was "Computational data processing methods for detecting objects with near-zero apparent motion" of the specialty 01.05.02 "Mathematical modeling and computational methods". Dr. Khlamov's research focuses on several areas, including computational methods, mathematical modeling (statistical and in situ), image recognition, image filtering, image processing, machine/computer vision, observational astronomy, computer science, big data and data science, data mining, knowledge discovery in databases, machine learning, internet of things, artificial intelligence, etc. Since 2014 Dr. Sergii Khlamov is a senior researcher of the Collection Light Technology (CoLiTec) project and the developer of the Lemur software for detecting the moving space objects (asteroids/satellites) in a series of astronomical frames. Also, Dr. Khlamov has more than 12 years of experience in Test Automation and Quality Assurance in the different top IT-companies. Now he is a Team Lead Test Automation of the SoftServe Incorporation. Dr. Sergii Khlamov has more than 160 national and international publications, including 5 monographs, 22 patents. Currently he is a scientific supervisor of the Ukrainian project of fundamental scientific research “Development of computational methods for detecting objects with near-zero and locally constant motion by optical-electronic devices” #0124U000259 in 2024-2026 years.

---