WODIS: Water Obstacle Detection Network Based on Image Segmentation for Autonomous Surface Vehicles in Maritime Environments

Chen, Xiang; Liu, Yuanchang; Achuthan, Kamalasudhan

doi:10.1109/tim.2021.3092070

Cited by 30 publications

(14 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Significant work has been completed in the past decade to use semantic networks in the ground and air domain for self-driving cars and UAVs [25,26]. However, only recently have semantic networks been applied to water detection in robotic applications for the purposes of obstacle detection for USV avoidance [9], river level monitoring [20], and general USV SLAM/Control [6][7][8]. The lack of uptake in the use of semantic networks in marine domains is directly tied to the data-hungry nature of the networks themselves.…”

Section: Related Workmentioning

confidence: 99%

“…WODIS [8] is a more traditional encoding and decoding framework that uses a U-Net inspired encoder-decoder structure to segment an input 512 × 384 image of marine scenes into water, sky, and obstacles. The encoder utilized is derived from the Xception network [48] that utilizes depth-wise separable convolutions in lieu of traditional convolution to extract deep features from the image.…”

Section: Water Obstacle Detection Network Based On Image Segmentation...mentioning

confidence: 99%

“…The network was trained on a NVIDIA RTX3080 with 8 GB of VRAM and a clock speed of 1.4 GHZ. The hyperparameters used for the training were as reported in the WODIS publication and associated code repository [8], except for the change in learning rate at 50 epochs, as performed in the author's online repository linked to in [8]. Instead of jumping from 0.0001 to 0.003 as in the publication, the learning rate is halved after 50 epochs, as shown in Equation ( 1).…”

Section: Network Implementation and Trainingmentioning

confidence: 99%

“…Semantic segmentation networks have seen exceptional real-world use in autonomous ground vehicles such as autonomous cars but have seen very limited experimental validatation for use in marine environmental perception in harbors and littoral zones [6][7][8] and even less real world implementation [9] fo the navigation task. To the authors' knowledge, semantic segmentation networks have not been explicitly applied to the river navigation and fluvial obstacle avoidance problem for USVs.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

ROSEBUD: A Deep Fluvial Segmentation Dataset for Monocular Vision-Based River Navigation and Obstacle Avoidance

Lambert

Chavez-Galaviz

et al. 2022

Sensors

View full text Add to dashboard Cite

Obstacle detection for autonomous navigation through semantic image segmentation using neural networks has grown in popularity for use in unmanned ground and surface vehicles because of its ability to rapidly create a highly accurate pixel-wise classification of complex scenes. Due to the lack of available training data, semantic networks are rarely applied to navigation in complex water scenes such as rivers, creeks, canals, and harbors. This work seeks to address the issue by making a one-of-its-kind River Obstacle Segmentation En-Route By USV Dataset (ROSEBUD) publicly available for use in robotic SLAM applications that map water and non-water entities in fluvial images from the water level. ROSEBUD provides a challenging baseline for surface navigation in complex environments using complex fluvial scenes. The dataset contains 549 images encompassing various water qualities, seasons, and obstacle types that were taken on narrow inland rivers and then hand annotated for use in semantic network training. The difference between the ROSEBUD dataset and existing marine datasets was verified. Two state-of-the-art networks were trained on existing water segmentation datasets and tested for generalization to the ROSEBUD dataset. Results from further training show that modern semantic networks custom made for water recognition, and trained on marine images, can properly segment large areas, but they struggle to properly segment small obstacles in fluvial scenes without further training on the ROSEBUD dataset.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Water Obstacle Detection Network Based On Image Segmentation...mentioning

confidence: 99%

Section: Network Implementation and Trainingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

ROSEBUD: A Deep Fluvial Segmentation Dataset for Monocular Vision-Based River Navigation and Obstacle Avoidance

Lambert

Chavez-Galaviz

et al. 2022

Sensors

View full text Add to dashboard Cite

show abstract

“…As an example [11] generates a dense 3D reconstruction with associated semantic labelling from stereo camera images, [12] and [13] propose a multimodal sensor-based semantic 3D mapping system using a 3D LiDAR combined with a camera. In the frame of maritime navigation, the authors of [14] present a Water Obstacle Detection network based on Image Segmentation (WODIS) for autonomous surface vehicles. Gan et al [15] describe a method that provides a unified probabilistic model for both occupancy and semantic probabilities, producing a Bayesian continuous 3D semantic occupancy map from noisy point clouds.…”

Section: Introductionmentioning

confidence: 99%

Occupancy grid mapping for rover navigation based on semantic segmentation

2021

View full text Add to dashboard Cite

Obstacle mapping is a fundamental building block of the autonomous navigation pipeline of many robotic platforms such as planetary rovers. Nowadays, occupancy grid mapping is a widely used tool for obstacle perception. It foreseen the representation of the environment in evenly spaced cells, whose posterior probability of being occupied is updated based on range sensors measurement. In more classic approaches, the cells are updated to occupied at the point where the ray emitted by the range sensor encounters an obstacle, such as a wall. The main limitation of this kind of methods is that they are not able to identify planar obstacles, such as slippery, sandy, or rocky soils. In this work, we use the measurements of a stereo camera combined with a pixel labeling technique based on Convolution Neural Networks to identify the presence of rocky obstacles in planetary environment. Once identified, the obstacles are converted into a scan-like model. The estimation of the relative pose between successive frames is carried out using ORB-SLAM algorithm. The final step consists of updating the occupancy grid map using the Bayes’ update Rule. To evaluate the metrological performances of the proposed method images from the Martian analogous dataset, the ESA Katwijk Beach Planetary Rover Dataset have been used. The evaluation has been performed by comparing the generated occupancy map with a manually segmented ortomosaic map, obtained by drones’ survey of the area used as reference.

show abstract

A method for detecting navigable areas in narrow rivers under complex reflection conditions

Zhang,

Hu,

et al. 2024

Engineering Reports

View full text Add to dashboard Cite

The perception of unmanned surface vehicles is significantly influenced by the detection of navigable areas in narrow rivers. Conventional semantic segmentation networks are unable to resolve the numerous interferences on the water's surface, including highlights and inverted images. To solve this problem, a river surface image reflection removal generative adversarial network (RRGAN) is proposed to eliminate the interference of harsh water surface environment. The proposed RRGAN only uses a single generator to reduce the number of parameters. By adding AdaLIN layers in the generator to enhance the ability to generate low‐reflection images, the AdaLIN encoder (AdaLINE) is proposed to automatically generate normalized affine parameters. In addition, a cycle semantic consistency loss function with a single generator is proposed to ensure that the water region of the generated images remains unchanged. Finally, a two‐stage method for detecting navigable areas is proposed. In the first stage, the RRGAN is used to remove the interference on the water surface environment. In the second stage, the semantic segmentation network is used to segment the water body from the denoised image to determine the navigable areas on the water surface. The experimental results demonstrate that, in the complex and varied narrow river environment, the suggested RRGAN method can significantly reduce the reflection interference of the water surface and improve the accuracy of the water segmentation after the reflection is removed.

show abstract

WODIS: Water Obstacle Detection Network Based on Image Segmentation for Autonomous Surface Vehicles in Maritime Environments

Cited by 30 publications

References 36 publications

ROSEBUD: A Deep Fluvial Segmentation Dataset for Monocular Vision-Based River Navigation and Obstacle Avoidance

ROSEBUD: A Deep Fluvial Segmentation Dataset for Monocular Vision-Based River Navigation and Obstacle Avoidance

Occupancy grid mapping for rover navigation based on semantic segmentation

A method for detecting navigable areas in narrow rivers under complex reflection conditions

Contact Info

Product

Resources

About