PROJECTS

FORTIS - Multi-Modal and Multi-Aspect Holistic Human-Robot Interaction - is an EU-funded project that aims to provide a solution for enabling robots to interact with humans in a human-like way for long periods. FORTIS will (i) develop, integrate and provide an human-centric solution for modeling and analyzing humans, (ii) develop and provide a flexible and agile multi-robotic-centric solution interacting with humans, (iii) provide optimized operations for both humans and robots and (iv) demonstrate the solution for industrial pilots in construction, maintenance and logistics. FORTIS wants to create a comprehensive interaction between humans and robots, encompassing both physical and non-physical communication. FORTIS is funded under the call HORIZON-CL4-2023-DIGITAL-EMERGING-01-02.

TRACENET - Training Centre Network on 3D and VR - aims to develop geospatial technologies, VR solutions and innovations for prevention, preparedness and response of Civil Protections and emergency responders to support policy and decision-making, encourage scientific excellence and exchange state-of-the-art knowledge and expertise. Within TRACENET, planners, decision makers, experts, teams and civil protection modules will jointly operate, test and learn from each other interacting within an online and real-time 3D environment. Such an innovative framework will offer simulation scenarios and countless opportunities to build expertise, share knowledge and establish remote but real contacts among participants, reducing costs and the environmental impact of training activities. TRACENET is funded under the EU Programme Union Civil Protection Mechanism (UCPM) and coordinated by the Autonomous Province of Trento (Italy).

InCUBE - an INClUsive toolBox for accElerating and smartening deep renovation - envisions to unlock the EU renovation wave through cutting-edge standardized and integrated processes based on (i) industrialization, (ii) new technologies for renewable energy production and storage, (iii) digitization (including Digital Twins, BIM, etc.) and (iv) new market entrants, organized under novel business. The Santa Chiara Open Lab and Santa Chiara Cultural Services Center in Trento’s historic district will represent one of 3 demonstration sites. The project is in collaboration with the DSIP research group and Sustainable Energy centre of FBK. InCUBE is funded under the EU call HORIZON-CL5-2021-D4-01-02, Grant Agreement no. 101069610 and is coordinated by CERTH.

SEC4TD - Securing tailings dam infrastructure with an innovative monitoring system - proposes an integrated end-to-end solution composed of hardware (IoT, drones, etc.) and software innovative products for mining operators and service providers that will enable multi-scale multi-platform data collection and visualization, events prediction, effective information management and data traceability. The SEC4TD will provide benefits to mining sector by improving the safety, efficacy and environment impact of the current mining processes and will also protect the reputation of this sector as well as will facilitate the compliance with new stricter regulations. SEC4TD will leverage on advanced and emerging technologies such as IoT low-power devices, mathematical soil model mechanisms, drones, advanced visualization tools, data analytics techniques, among others. The project is funded by the EIT Raw Materials and in collaboration with the DSIP research group of FBK.

WildDrone – Autonomous Drones for Nature Conservation Missions - is an international training network funded by the EU Marie Skłodowska Curie Actions. WildDrone aims to revolutionize wildlife conservation practices by using autonomous drone technology as a unifying platform to monitor wildlife populations, track their movements, and manage human-wildlife conflicts. We believe that the rapid decline of wildlife populations has prompted an urgent need for more effective and feasible nature conservation practices. The project will operate from 1 January 2023 – 31 December 2026.  

V0T3D - Ventilation Optimizing Technology based on 3D-scanning -  wants to robotize the Raw Material sector by introducing modern methods for the optimization of underground ventilation with the main aim of reducing electricity consumption by 1,5%. The VOT3D project, funded by the EIT Raw Materials, will deliver an innovative solution based on 3D digitization of underground tunnels and the creation of their real digital twins. These detailed and precise 3D data will allow to improve airflow simulations in order to adjust the most favorable ventilation scenario for the digitized underground tunnels.

MIAMI is an industrial project which aims to develop a system able to integrate advanced metrology and artificial intelligence (AI) for: (i) helping an operator in setting up a laser micro-machining process and in checking the quality of the obtained result using 3D analytics; (ii) improving the production process by learning defects and their relation with machining parameters; (iii) developing new business opportunities related to design and sale of intelligent laser machines and related services. Two types of micro-machining are considered: LIPPS and micro-channels on steel. 3D data are collected with SEM and interferometers. MIAMI is in collaboration with Kirana and supported by the Autonomous Province of Trento.

UPDATE – fast aUtomatic insPection with 3D imAging and arTificial intElligence - is a metrology project, run in collaboration with EOPTIS and funded by Fondazione CARITRO. UPDATE aims to research, develop and demonstrate innovative solutions for quality control of industrial components. The inspection will be performed merging 3D imaging and Artificial Intelligence solutions and following international standards (VDI/VDE, ASTM, etc.). 

In order to preserve the memory of the Sycamore Gap tree, a symbol of the UNESCO Hadrian's Wall site that was felled in an act of vandalism in September 2023, deep-learning-based features trained specifically on challenging image datasets are employed. The research demonstrates how unordered crowdsourced images and UAV videos can be oriented and used for 3D reconstruction purposes. This allows the memory of the Sycamore Gap tree to live on and exhibits the potential of photogrammetric AI for reverse engineering lost heritage.

Classification and segmentation of 3D data has become a very active research field. Researchers are developing solutions to deliver enriched 3D information useful in many fields and applications. We are developing methodologies based on modern machine/deep learning methods useful for urban modeling, heritage documentation, etc. 

We explore supervised CNN architectures for monocular depth estimation and evaluated their potential in 3D reconstruction i photogrammetric scenarios and applications. Since most available datasets for training are not designed toward this goal and are limited to specific indoor scenarios, we also offer a new metric, large-scale synthetic benchmark (ArchDepth) renders near real-world scenarios of outdoor scenes (ca 24.000 images). 

Multiple view stereo algorithms still have difficulties to reconstruct the depth in textureless areas. Semantic information is used to support the standard MVS reconstruction and generate more complete and accurate depth maps and 3D point clouds. The semantic labels are inherited from the labelled images to the 3D output.

If the field of photogrammetry has witnessed incremental technological developments in the last years, the field of airborne laser scanning is now seeing a potentially revolutionary change through the introduction of Single Photon LiDAR (SPL) and Geiger-Mode LiDAR technologies. SPL / Geiger-Mode arguably provide a more efficient approach to perform high-resolution 3D mapping. They offer higher efficiency with respect to traditional linear LiDAR and additional 3D mapping capability in terms of swath width, spatial resolution, acquisition time and density of range returns. The project is in collaboration with EuroSDR, Newcastle University and TU Vienna.

Automation in image orientation and dense point cloud reconstruction is reaching very high standards, but image quality is a fundamental pre-requisite to produce successful and photo-realistic 3D products. We investigate a productive pre-processing strategy, the various detector/descriptor operators for tie point extraction and dense matching algorithms to find out all open issues in the actual photogrammetric pipeline for 3D reconstruction purposes. 

Various countries are digitizing their archives of aerial images as they are a unique and relatively unexplored means to analyze the territorial changes over the past 100 years with very high spatial resolution. This research activity wants to investigate the geo-processing of historical aerial images, including automatic geo-referencing, DSM generation, automatic detection of bomb craters, LULC maps generation, etc. The TIME dataset/benchmark was realized in collaboration with EuroSDR.