BrainMap

Mr. Corneliu Florea

Professor (Prof. dr. ing.)

- UNIVERSITATEA NAŢIONALĂ DE ŞTIINŢĂ ŞI TEHNOLOGIE POLITEHNICA BUCUREŞTI

Researcher | Teaching staff | Scientific reviewer | PhD supervisor

Corneliu Florea was born in 1980 in Bucharest. He got his master degree from University “Politehnica”of Bucharest, Romania in 2004 and the PhD from the same university in 2009. After a stint with digital still camera software industry, currently he is Professor at the same university, within Image Processing and Analysis group. There he lectures on machine learning, computer vision, statistical signal and image processing . His research interests include statistical approaches to machine leaning and computer vision. He is author of more than 100 peer-reviewed papers and of more than 25 US patents.

Web of Science ResearcherID: B-5540-2012

Curriculum Vitae (25/04/2024)

Expertise & keywords

Early identification of agricultural crops using artificial intelligence and Sentinel-2 data Call name: PNCDI IV, SP 5.7.1 - Proiect experimental demonstrativ
PN-IV-P7-7.1-PED-2024-0375 2025 - 2026

Role in this project:

Coordinating institution: UNIVERSITATEA TRANSILVANIA BRASOV

Project partners: UNIVERSITATEA TRANSILVANIA BRASOV (RO); FIELD DATA ZOOM S.R.L. (RO)

Affiliation:

Project website:

Abstract:

The goal of the project is to increase the technology readiness level, from TRL 2 to TRL 4, of a method for early identification of agricultural crops based on artificial intelligence and Sentinel-2 data. The goal will be achieved by collaboration with a private company and thus ensuring the correct technology transfer and adopting of the proposed solution into a commercial product. The project includes fundamental research activities, collaborative industrial research, as well as intellectual property dissemination and protection activities. The early identification of crops will allow the generation, with a certain level of precision, of vegetation maps for the current agricultural season, which will lead to a reduction in the waiting time for statistics at the national level, as well as a reduction in the discrepancies between the statistics at the national level and those at the international level.

Platform for fusion and management of multi-source data collections exploitable by artificial intelligence models for the estimation and predictive analysis of risk situations Call name: P 5.6 - SP 5.6.3 - Solutii
PN-IV-P6-6.3-SOL-2024-2-0251 2024 - 2026

Role in this project:

Coordinating institution: UNIVERSITATEA NAȚIONALĂ DE ȘTIINȚĂ ȘI TEHNOLOGIE POLITEHNICA BUCUREȘTI

Project partners: UNIVERSITATEA NAȚIONALĂ DE ȘTIINȚĂ ȘI TEHNOLOGIE POLITEHNICA BUCUREȘTI (RO); Academia Tehnică Militară „FERDINAND I” (RO); UNIVERSITATEA TRANSILVANIA BRASOV (RO); ESRI ROMANIA SRL (RO)

Affiliation:

Project website: http://ai4risk.ro

Abstract:

AI4RISK aims to strengthen national security by using machine learning technologies to analyze and interpret complex data from diverse sources in order to identify and anticipate potential security risks. The project aims to achieve an innovative solution for automatic collection and processing of satellite and aerial data collections by implementing generative methods to improve the characteristics of aerial images; the creation and implementation of containerized explainable artificial intelligence models, the implementation of the platform for predictive analysis and risk estimation based on user questions, the integration and visualization of data in the platform, through the implementation of mapping services and the rendering of real-time image streams.

In addition to the requirements formulated by the beneficiary, the consortium proposes for implementation in the AI4RISK platform a system for preparing and improving operators' reactions to risk incidents, through the use of machine learning techniques. It will allow operators to adapt/enable specific methodologies for data processing and status assessment. Therefore, analyzing the results and taking into account the information density of the map of the region of interest, the system will propose recommendations for further actions - such as the integration of a wider range of news sources, including in several languages, the activation of an extensive network of cameras for monitoring, the addition of additional data relating to targets in the area – all aimed at optimizing the operator's intervention strategy.

IMINT for Black Sea, Frontiers and Naval Mines Call name: P 5.6 - SP 5.6.3 - Solutii
PN-IV-P6-6.3-SOL-2024-0124 2024 - 2026

Role in this project:

Coordinating institution: UNIVERSITATEA NAȚIONALĂ DE ȘTIINȚĂ ȘI TEHNOLOGIE POLITEHNICA BUCUREȘTI

Project partners: UNIVERSITATEA NAȚIONALĂ DE ȘTIINȚĂ ȘI TEHNOLOGIE POLITEHNICA BUCUREȘTI (RO); Academia Tehnică Militară „FERDINAND I” (RO); AGENTIA SPATIALA ROMANA (RO); Academia Navala "Mircea cel Batran" (RO); INTERGRAPH COMPUTER SERVICES S.R.L. (RO); UNIVERSITATEA TRANSILVANIA BRASOV (RO)

Affiliation:

Project website: https://sites.google.com/view/imint21/

Abstract:

The Black Sea, being located at the confluence of Europe and Asia, has a significant strategic importance both at regional and international levels. In the current international context, the security and stability of the Black Sea region is of uttermost importance. For this reason, significant efforts are focused on ensuring IMINT capabilities by integrating IT systems for monitoring maritime activities, control of maritime borders, identification of military infrastructure in the Black Sea area, as well as for naval mine detection. The project proposes an integrated and innovative approach for multi-source information retrieval (e.g. data collected by sensors placed on ships and aerial platforms, coastal stations, satellite data) through the development of a 5D IMINT Dispatch. The 5D IMINT Dispatch allows automated data collection, with incorporated location and timing, as well as an advanced data analysis in order to identify anomalies and trends in the presence of risk situations (e.g., the existence of floating naval mines). At the same time, the detection of potential risks and the prediction of future events using artificial intelligence tools is a important step forward towards early warning and timely support of decision makers. The development of the 5D IMINT Dispatcher contributes to the creation of a methodological framework for the standardization of IMINT data models and information flow, as well as for planning operations to prevent and eliminate maritime risks. Moreover, the identification of the main security risks in the maritime area leads to the definition of the main case studies, which are the basis of the IMINT proposal and which outline the demonstration of the concepts and methodologies that are implemented within the project, e.g. support for ensuring security for people and maritime transport of goods, monitoring and protection of oil platforms in the Black Sea.

6. Identify people in video streams, using silhouette biometrics Call name: P 2 - SP 2.1 - Soluţii - 2021
PN-III-P2-2.1-SOL-2021-0114 2021 - 2023

Role in this project:

Coordinating institution: UNIVERSITATEA POLITEHNICA DIN BUCURESTI

Project partners: UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO); SOFTRUST VISION ANALYTICS S.A. (RO); Ministerul Apararii Nationale prin Agentia de Cercetare pentru Tehnica si Tehnologii Militare (ACTTM) (RO)

Affiliation:

Project website: https://deepvision-romania.aimultimedialab.ro/

Abstract:

The DeepVisionRomania project comes to respond to the call-for-solutions launched for the development, testing and operationalization of software services for automatic identification and re-identification of people in video sequences with GDPR regulations in force, in order to increase operational competitiveness, economy and staff skills via “made-in-Romania” security software solutions. The developed technologies are able, by their degree of novelty, to ensure a strategic security advantage for Romania by supporting operations to prevent and combat terrorism or other categories of threats to the public order and safety, or emergency situations. The project proposes: (i) the development of new state-of-the-art AI algorithms based on deep neural networks capable of ensuring the identification (search in a database of identities) and the re-identification of persons (search for the appearance of a person in multiple sources ) in different video streams using both static and dynamic information; (ii) development of dedicated analysis software services but with a secure architecture and developed in accordance with operational standards / requirements; modular so that any new functionality does not require reprogramming of the system; scalable to multiple existing and new video sources, optimized to ensure all real-time processing; (iii) development of an integrated software and hardware platform that brings together all solutions; (iv) ensuring GDPR compliance through an advanced user management system with access security. The market for the applicability of such systems is particularly wide, from the operational institutions that process this information, critical infrastructures (protection and security, prisons, border, airports, hospitals, banks), to the civil area (transport, shopping centers, surveillance). urban, residential complexes, company headquarters, supermarket).

Transfer Learning for Image Aesthetic Assessment Call name: P 1 - SP 1.1 - Proiecte de cercetare pentru stimularea tinerelor echipe independente
PN-III-P1-1.1-TE-2019-0543 2020 - 2022

Role in this project:

Coordinating institution: UNIVERSITATEA POLITEHNICA DIN BUCURESTI

Project partners: UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO)

Affiliation: UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO)

Project website: http://imag.pub.ro/translate/

Abstract:

The project aims to develop transfer learning methods for deep convolutional networks. In this framework, besides the main database, which allows supervised learning, we will consider a second database, larger in volume, that contains examples either without annotations or with labels, but for another, related, problem. The transfer learning involves extracting useful information from the second database to increase performance on the primary set, in a simultaneous process.

For the development of the transfer learning algorithm, the characteristics of the convolutional networks to form descriptors on the intermediate layers, respectively to be sequentially trained are exploited. The central algorithmic principle is the partitioning by clustering (if the data does not have labels) or by classifying the secondary data set, where the descriptors of data are extracted on an intermediate layer. The network is required to create descriptors that provide wide margins between the data of one group and the centroid of another group. The groups can be associated with the classes from the primary database and the method of calculating the wide margin will be explored. Secondly, for the harmonization of the two databases we will use regularization by injecting a random perturbation, in an annealed process, into the global gradient.

From an application point of view, we aim to estimate the aesthetic value of a photograph. The aesthetic value aggregates, consciously or unconsciously, how pleasing an image looks. For high performance, in this direction we will exploit the previously developed transfer learning algorithm. The application is useful in commercial area to offer customers visual quality.

Technologies and innovative video systems for person re-identification and analysis of dissimulated behavior Call name: P 2 - SP 2.1 - SOLUTII - 2 - Tehnologii şi sisteme video/audio inovative
PN-III-P2-2.1-SOL-2016-02-0002 2017 - 2020

Role in this project: Key expert

Coordinating institution: UNIVERSITATEA POLITEHNICA DIN BUCURESTI

Project partners: UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO); UTI GRUP S.A. (RO); Ministerul Apararii Nationale prin Agentia de Cercetare pentru Tehnica si Tehnologii Militare (ACTTM) (RO)

Affiliation: UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO)

Project website: http://campus.pub.ro/lab7/spiava/

Abstract:

Within the actual context of terrorist threats, SPIA-VA aims to develop intelligent systems for automatic person re-identification, detection of dissimulated behavior and speech analysis. The proposed system is inherently multi-modal via processing of multiple and variable quality sources, e.g., video, audio, text, infrared, depth, thermal, as well as the fusion of these. The proposed solution pushes forward the state of the art in deep learning and is designed to respond to a large variety of functional and operational scenarios. It is to be validated in real world scenarios, developed in cooperation with the beneficiary military public institutions.

Stabilization Technology for Elimination of Absolute and DYnamic blur due to Camera Acquisition Motion Call name:
PN-II-PT-PCCA- 2013-4-0579 2014 - 2017

Role in this project: Partner team leader

Coordinating institution: FOTONATION SRL

Project partners: FOTONATION SRL (RO); UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO)

Affiliation: UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO)

Project website: http://imag.pub.ro/steadycam

Abstract:

Following their appearance in the early 90's, digital still cameras (DSC) have become the common way of acquiring images. Nowadays, the main development directions include that of decreasing the size and weight of imaging devices, which has reached a pinnacle with Mobile Camera Phones (MCP). The trend of miniaturization mentioned above imposes design modifications such as reducing the size of optics and of photo-sensible area, thus increasing the probability that the pictures be blurred by the hands' tremor. Since human hand tremor is always present, both the small viewing angle and a large value of the exposure time increase the chances that the relative motion between the camera and the scene during exposure time become larger than a pixel size, thus leading to a visible degradation of the image by the motion blur. The problem manifests in the form of blurred still images, leading to unpleasant image artifacts and acquisition limitations in areas as security or medicine.

Perceptual ANalysis and DescriptiOn of Romanian visual Art Call name: Projects for Young Research Teams - RUTE -2014 call
PN-II-RU-TE-2014-4-0733 2015 - 2017

Role in this project: Project coordinator

Coordinating institution: UNIVERSITATEA POLITEHNICA DIN BUCURESTI

Project partners: UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO)

Affiliation: UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO)

Project website: http://imag.pub.ro/pandora/

Abstract:

With the vast expansion of digital painting collections, the art historians and curators gain access to new investigation methods that can help them in comprehending significant larger collections and in more detail. Motivated by such opportunity, there has been a significant escalation in the demand for automatic description of visual art works. In this project we propose, first, the exploration of new description methods based on the isogeodesic curve in logarithmic space that could lead to increasingly accurate artistic genre and school of art recognition. Secondly, we propose to annotate the paintings in terms of gaze fixations, and, thus, to identify saliency regions; the regions distribution and construction should provide additional indications about adherence to a school of art or artistic genre. Thirdly, but, from a point of view, most importantly, we propose to introduce in the international digital analysis domain works from modern and contemporary Romanian artists. Using the previously developed tools, we hope that new correlations and connections between Romanian and international art will be revealed.

Stabilization Technology for Elimination of Absolute and DYnamic blur due to Camera Acquisition Motion Call name: Joint Applied Research Projects - PCCA 2013 - call
PN-II-PT-PCCA-2013-4-0579 2014 - 2017

Role in this project:

Coordinating institution: FOTONATION SRL

Project partners: FOTONATION SRL (RO); UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO)

Affiliation: UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO)

Project website: http://imag.pub.ro/steadycam/

Abstract:

Following their appearance in the early 90’s, digital still cameras (DSC) have become the common way of acquiring images. Nowadays, the main development directions include that of decreasing the size and weight of imaging devices, which has reached a pinnacle with Mobile Camera Phones (MCP). The trend of miniaturization mentioned above imposes design modifications such as reducing the size of optics and of photo-sensible area, thus increasing the probability that the pictures be blurred by the hands’ tremor. Since human hand tremor is always present, both the small viewing angle and a large value of the exposure time increase the chances that the relative motion between the camera and the scene during exposure time become larger than a pixel size, thus leading to a visible degradation of the image by the motion blur. The problem manifests in the form of blurred still images, leading to unpleasant image artifacts and acquisition limitations in areas as security or medicine.

The problem of the tremor influence on image/video acquisition has been divided in smaller problems, as being simpler to be addressed. It is possible to estimate the camera trajectory (absolute blur) and restore the picture by means of deconvolution (so called Digital Image Restoration - DIR technology) or to compensate the motion by mechanical/optical means (forming the Optical Image Stabilization - OIS technology); an alternative is to prevent blurring by reducing the exposure time, thus acquiring an underexposed image that is to be amplified (LowLight technology). As all above mentioned attempts have addressed the problem from a single point of view, we claim that by sharing the same cause, all of them may be addressed and solved by the complementary use of microelectromechanical (MEMS) motion sensors, optical processing and innovative image processing adapted to the particularity of the target device.

The main goal of the proposed project is to construct a solution that addresses the problems caused by the involuntary human hand tremor in a broad range of consumer imaging devices. We build the project plan on a bottom-up pyramidal schema. More precisely, we start by studying the specific parameters of the human hand involuntary tremor; we continue by defining a first layer of 2 basic themes that will be studied in the first part of the project; later, we interconnect them in a second layer of a general and complex solution that compensates the human tremor, allowing thus the increase of the reliable maximum exposure period by 8 times (3EV).

Burns Assessment by MultiSpectral Imaging Call name: Joint Applied Research Projects - PCCA 2013 - call
PN-II-PT-PCCA-2013-4-0357 2014 - 2017

Role in this project: Key expert

Coordinating institution: UNIV.DE MEDICINA SI FARMACIE - CAROL DAVILA

Project partners: UNIV.DE MEDICINA SI FARMACIE - CAROL DAVILA (RO); UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO); FOTONATION SRL (RO)

Affiliation: UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO)

Project website: http://alpha.imag.pub.ro/bamsi/

Abstract:

Burns represent a serious public health problem, as burn injury represents perhaps the most severe form of trauma. The incidence of burn injuries greatly varies from one region to another, but the median value is about 31,2/100.000 persons/year. There are about 330.000 deaths per year, world-wide due to thermal injuries. Burns are the fourth leading cause of death from unintentional injury.

Establishing the difference between superficial dermal burns and deep dermal burns is very important. Superficial burns heal with proper dressings in 14-21 days and leave no scars. Deep dermal burns and full thickness burns need early surgical excision and skin grafting, otherwise healing time is longer than 21 days and pathological scaring is the rule. Usually, this differentiation is made on clinical basis by the burns surgeon. The accuracy of a senior burn surgeon is 64-76%. It means that in 23-36% of the cases, the clinical judgment of the surgeon fails to differentiate between superficial burns and deep burns.

The goal of this project is to develop a tool that objectively determines the depth of burn wounds, based on the estimation of perfusion through the burn wound via the joint use of several imaging recordings the color, near-infrared and thermal infrared bands. This approach is cheaper, faster, and may produce images of higher resolution than the current state of the art Laser Doppler spectrometry.

Our approach is quite original through the use of multispectral imaging and represents a challenging scientific and technical issue. The combination of experimental, modelling, and advanced statistics and image processing will solve some important problems like (1) to develop protocols for fast and accurate in vivo medical examination of burns via photographic imaging; (2) to fuse visible images with information extracted from thermal images to highlight abnormalities and to offer diagnostic aid; (3) to develop an workable burn assessment apparatus for clinical use.

Orthopedic Prothesis Monitoring by Analysis of Digital X-ray Images Call name:
CEEX-VIASAN-69/2006 2006 - 2008

Role in this project: Key expert

Coordinating institution: UNIVERSITATEA POLITEHNICA DIN BUCURESTI

Project partners: UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO); UNIVERSITATEA DE MEDICINA SI FARMACIE "CAROL DAVILA" (RO)

Affiliation: UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO)

Project website: http://alpha.imag.pub.ro/sadirupo/

Abstract:

The idea of total hip prosthesis was born in the 1950s and evolved towards the nowadays total hip prosthesis with a stem and an acetabular component. The scoring of the prothesis fit is a major area of investigation, from both the theoretical and practical (need for revision surgery) point of view. A significant number of bone loss and bone defect classification systems have been developed; however, there is no universal agreement on these criteria, and there are no reports on the use of automatic, image-processing based radiographic scores.

This research project aims to establish a support system for the characterization of the hip prothesis status, based on the automatic interpretation, by digital image processing and analysis techniques, of digitized X-ray images of the prothesisez hip. The main advantage of the proposed approach is the use of a cheap medical imaging equipment (classical, film-based X-ray), largely available across medical facilities and some modern digital technologies in order to reduce the need of the hip prothesed patients to travel for regular control and follow-up at the clinics that performed the actual prothesis impantation. The final goal of the research is to develop a complete and coherent algorihm (starting from the digital acquisition of a hip X-ray and ending with the computation of an effective statistical analysis of the bone structure, bone-prothesis interface and prothesis surface characteristics), that can be used for reliably generating a quantitative, clinical-relevant grading of the prothesis status.

Analysis of the EnvironMent Influence over paintings visual Saliency Call name: P 3 - SP 3.1 - Proiecte de mobilități, România-Belgia (bilaterale)
PN-III-P3-3.1-PM-RO-BE-2016-0004 2016 -

Role in this project: Key expert

Coordinating institution: UNIVERSITATEA POLITEHNICA DIN BUCURESTI

Project partners: UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO); Universite de Mons (UMONS) (BE)

Affiliation: UNIVERSITATEA POLITEHNICA DIN BUCURESTI (RO)

Project website:

Abstract:

Plecând de la ideea că „fără artă brutalitatea realității ar face lumea de nesuportat”, acest proiect își propune să studieze efectele pe care le poate avea metoda de digitizare a tablourilor și prelucrări subtile asupra imaginilor digitale rezultante asupra zonelor de proeminență ale tablourilor. Ne propunem să construim o metodă de urmărire a privirii ne-invazivă (formată prin alăturarea metodelor deja dezvoltate de cele două echipe implicate în proiect) și cu ajutorul acesteia să colectăm o bază de date relevantă cu zone de proeminență în tablouri.

FILE DESCRIPTION	DOCUMENT
List of research grants as project coordinator
List of research grants as partner team leader
List of research grants as project coordinator or partner team leader
Significant R&D projects for enterprises, as project manager
R&D activities in enterprises
Peer-review activity for international programs/projects