A5018368394- Underwater Vehicles and Communication Systems
- Robotics and Sensor-Based Localization
- Advanced Neural Network Applications
- Robotic Path Planning Algorithms
- Maritime Navigation and Safety
- Infrared Target Detection Methodologies
- Advanced Database Systems and Queries
- Water Quality Monitoring Technologies
- Natural Language Processing Techniques
- Inertial Sensor and Navigation
- Underwater Acoustics Research
- Impact of AI and Big Data on Business and Society
- Advanced Text Analysis Techniques
- Remote Sensing and LiDAR Applications
University of Ljubljana
2017-2021
The progress of obstacle detection via semantic segmentation on unmanned surface vehicles (USVs) has been significantly lagging behind the developments in related field autonomous cars. reason is lack large curated training datasets from USV domain required for development data-hungry deep CNNs. This paper addresses this issue by presenting MaSTr1325, a marine dataset tailored methods small-sized coastal USVs. contains 1325 diverse images captured over two year span with real USV, covering...
Small-sized unmanned surface vehicles (USV) are coastal water devices with a broad range of applications such as environmental control and surveillance. A crucial capability for autonomous operation is obstacle detection timely reaction collision avoidance, which has been recently explored in the context camera-based visual scene interpretation. Owing to curated datasets, substantial advances interpretation have made related field ground vehicles. However, current maritime datasets do not...
In this paper, we present a method for detecting and tracking waterborne obstacles from an unmanned surface vehicle (USV) the purpose of short-term obstacle avoidance. A stereo camera system provides point cloud scene in front vehicle. The water is estimated by fitting plane to outlying points are further processed find potential obstacles. We propose new algorithm detection that applies fast approximate semantic segmentation filter utilizes external IMU reading constrain orientation. novel...
Obstacle detection using semantic segmentation has become an established approach in autonomous vehicles. However, existing methods, primarily developed for ground vehicles, are inadequate aquatic environment as they produce many false positive (FP) detections the presence of water reflections and wakes. We propose a novel deep encoder-decoder architecture, refinement (WaSR) network, specifically designed marine to address these issues. A encoder based on ResNet101 with atrous convolutions...
Obstacle detection by semantic segmentation shows a great promise for autonomous navigation in unmanned surface vehicles (USV). However, existing methods suffer from poor estimation of the water edge presence visual ambiguities, small obstacles and high false-positive rate on reflections wakes. We propose new deep encoder-decoder architecture, water-obstacle separation refinement network (WaSR), to address these issues. Detection accuracy are improved novel decoder that gradually fuses...
We propose a stereo-based obstacle detection approach for unmanned surface vehicles. Obstacle is cast as scene semantic segmentation problem in which pixels are assigned probability of belonging to water or non-water regions. extend single-view model stereo system by adding constraint prefers consistent class labels assignment the left and right camera images corresponding same parts 3D scene. Our jointly fits both images, leading an improved class-label posterior map from obstacles edge...
We present a new semantic segmentation algorithm for obstacle detection in unmanned surface vehicles. The novelty lies the graphical model that incorporates boat tilt measurements from on-board inertial measurement unit (IMU). IMU readings are used to estimate location of horizon line image, and automatically adjusts priors probabilistic algorithm. derive necessary projection equations, an efficient optimization proposed model, practical IMU-camera-USV calibration. A challenging dataset,...
Obstacle detection by semantic segmentation shows a great promise for autonomous navigation in unmanned surface vehicles (USV). However, existing methods suffer from poor estimation of the water edge presence visual ambiguities, small obstacles and high false-positive rate on reflections wakes. We propose new deep encoder-decoder architecture, water-obstacle separation refinement network (WaSR), to address these issues. Detection accuracy are improved novel decoder that gradually fuses...
ROSUS 2019 – Računalniška obdelava slik in njena uporaba v Sloveniji je strokovna računalniška konferenca, ki jo od leta 2006 naprej vsako leto organizira Inštitut za računalništvo iz Fakultete elektrotehniko, informatiko, Univerze Mariboru. Konferenca povezuje strokovnjake raziskovalce s področij digitalne obdelave strojnega vida z uporabniki tega znanja, pri čemer prihajajo raznovrstnih industrijskih okolij, biomedicine, športa, zabavništva sorodnih področij. Zbornik konference združuje 14...
Razvoj segmentacijskih metod globokega učenja za detekcijo ovir na vodi je v precejšnjem zaostanku primerjavi z razvojem sorodni domeni avtonomnih vozil (AGV). Do nedavnega bil glavni razlog to pomanjkanje ustreznih podatkovnih zbirk ter dejstvo, da metode razvite AGV niso primerne aplikacijo vodno okolje zaradi domenskih specifik. Trenutno ni jasno, katere arhitekture so najprimernejše domeno. Zato smo izbrali tri popularne globoke semantične segmentacije (U-Net, PSPNet, DeepLab2), jih...