The Empirical Application of Automotive 3D Radar Sensor for Target Detection for an Autonomous Surface Vehicle’s Navigation

Stateczny, Andrzej; Kazimierski, Witold; Gronska-Sledz, Daria; Motyl, Weronika

doi:10.3390/rs11101156

Cited by 42 publications

(34 citation statements)

References 29 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Short prediction time is needed to meet the requirements for vehicle motion control time [17]. The challenges of anti-collision tasks in a water environment were presented in [18], but the test was provided for a small autonomous surface vehicle using radar sensors. When choosing a DLNN type [19], the following factors should be considered: accuracy, size and prediction time.…”

Section: Introductionmentioning

confidence: 99%

The Effectiveness of Using a Pretrained Deep Learning Neural Networks for Object Classification in Underwater Video

2020

View full text Add to dashboard Cite

Video image processing and object classification using a Deep Learning Neural Network (DLNN) can significantly increase the autonomy of underwater vehicles. This paper describes the results of a project focused on using DLNN for Object Classification in Underwater Video (OCUV) implemented in a Biomimetic Underwater Vehicle (BUV). The BUV is intended to be used to detect underwater mines, explore shipwrecks or observe the process of corrosion of munitions abandoned on the seabed after World War II. Here, the pretrained DLNNs were used for classification of the following type of objects: fishes, underwater vehicles, divers and obstacles. The results of our research enabled us to estimate the effectiveness of using pretrained DLNNs for classification of different objects under the complex Baltic Sea environment. The Genetic Algorithm (GA) was used to establish tuning parameters of the DLNNs. Three different training methods were compared for AlexNet, then one training method was chosen for fifteen networks and the tests were provided with the description of the final results. The DLNNs were trained on servers with six medium class Graphics Processing Units (GPUs). Finally, the trained DLNN was implemented in the Nvidia JetsonTX2 platform installed on board of the BUV, and one of the network was verified in a real environment.

show abstract

Section: Introductionmentioning

confidence: 99%

The Effectiveness of Using a Pretrained Deep Learning Neural Networks for Object Classification in Underwater Video

2020

View full text Add to dashboard Cite

show abstract

“…In other words, sensors that can detect the activity of other vehicles at a sufficient distance to prevent an accident considering driver reaction times and Vehicle speed can be too costly to widely penetrate the market. Many ADAS implementations therefore use in-vehicle sensors to produce warning signals based on information from a limited range of up to 100 or 150 meters from the vehicle [14]. However, this range may not be sufficient to anticipate a possible collision risk arising from traffic further downstream from the subject vehicle in time for the driver to safely conduct necessary actions to prevent a dangerous situation, especially in a free flow traffic state.…”

Section: Introductionmentioning

confidence: 99%

Sectional Information-Based Collision Warning System Using Roadside Unit Aggregated Connected-Vehicle Information for a Cooperative Intelligent Transport System

Tak

Yoon

Woo

et al. 2020

Journal of Advanced Transportation

View full text Add to dashboard Cite

Vehicular collision and hazard warning is an active field of research that seeks to improve road safety by providing an earlier warning to drivers to help them avoid potential collision danger. In this study, we propose a new type of a collision warning system based on aggregated sectional information, describing vehicle movement processed by a roadside unit (RSU). The proposed sectional information-based collision warning system (SCWS) overcomes the limitations of existing collision warning systems such as the high installation costs, the need for high market penetration rates, and the lack of consideration of traffic dynamics. The proposed SCWS gathers vehicle operation data through on-board units (OBUs) and shares this aggregated information through an RSU. All the data for each road section are locally processed by the RSU using edge computing, allowing the SCWS to effectively estimate the information describing the vehicles surrounding the subject vehicle in each road section. The performance of the SCWS was evaluated through comparison with other collision warning systems such as the vehicle-to-vehicle communication-based collision warning system (VCWS), which solely uses in-vehicle sensors; the hybrid collision warning system (HCWS), which uses information from both infrastructure and in-vehicle sensors; and the infrastructure-based collision warning system (ICWS), which only uses data from infrastructure. In this study, the VCWS with a 100% market penetration rate was considered to provide the most theoretically similar result to the actual collision risk. The comparison results show that in both aggregation and disaggregation level analyses, the proposed SCWS exhibits a similar collision risk trend to the VCWS. Furthermore, the SCWS shows a high potential for practical application because it provides acceptable performance even with a low market penetration rate (30%) at the relatively low cost of OBU installation, compared to the VCWS requirement of a high market penetration rate at a high installation cost.

show abstract

“…Nevertheless, environment perception is the foundation to achieve all these above-mentioned requirements. Various sensors are applied for the USV perception, such as camera [1] and radar [2,3]. Due to size limitations, such USVs have a small load capacity and cannot be mounted many sensors.…”

Section: Introductionmentioning

confidence: 99%

Adaptive Semantic Segmentation for Unmanned Surface Vehicle Navigation

et al. 2020

View full text Add to dashboard Cite

The intelligentization of unmanned surface vehicles (USVs) has recently attracted intensive interest. Visual perception of the water scenes is critical for the autonomous navigation of USVs. In this paper, an adaptive semantic segmentation method is proposed to recognize the water scenes. A semantic segmentation network model is designed to classify each pixel of an image into water, land or sky. The segmentation result is refined by the conditional random field (CRF) method. It is further improved accordingly by referring to the superpixel map. A weight map is generated based on the prediction confidence. The network trains itself with the refined pseudo label and the weight map. A set of experiments were designed to evaluate the proposed method. The experimental results show that the proposed method exhibits excellent performance with few-shot learning and is quite adaptable to a new environment, very efficient for limited manual labeled data utilization.

show abstract

The Empirical Application of Automotive 3D Radar Sensor for Target Detection for an Autonomous Surface Vehicle’s Navigation

Cited by 42 publications

References 29 publications

The Effectiveness of Using a Pretrained Deep Learning Neural Networks for Object Classification in Underwater Video

The Effectiveness of Using a Pretrained Deep Learning Neural Networks for Object Classification in Underwater Video

Sectional Information-Based Collision Warning System Using Roadside Unit Aggregated Connected-Vehicle Information for a Cooperative Intelligent Transport System

Adaptive Semantic Segmentation for Unmanned Surface Vehicle Navigation

Contact Info

Product

Resources

About