Forward-Looking Sonar CNN-based Automatic Target Recognition: an experimental campaign with FeelHippo AUV

Zacchini, Leonardo; Franchi, Matteo; Manzari, Vincenzo; Pagliai, Marco; Secciani, Nicola; Topini, Alberto; Stifani, Mirko; Ridolfi, Alessandro

doi:10.1109/auv50043.2020.9267902

Cited by 23 publications

(10 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, some algorithms use CNN to extract better features, but the detection time increases because of the sliding window used to extract proposals. Zacchini et al [ 31 ] realize the recognition and location of potential objects in FLS images through the existing Mask R-CNN, and achieve good accuracy and recall, but the inference time of Mask R-CNN is about 200 ms, which is not suitable for real-time detection. By comparing the performance of several different object detection algorithms on the same FLS image dataset, Kvasic et al [ 32 ] find a robust and reliable object detection network for the detection and tracking of human divers.…”

Section: Related Workmentioning

confidence: 99%

Multiple Receptive Field Network (MRF-Net) for Autonomous Underwater Vehicle Fishing Net Detection Using Forward-Looking Sonar Images

Qin

Zhao

Zhu

et al. 2021

Sensors

View full text Add to dashboard Cite

Underwater fishing nets represent a danger faced by autonomous underwater vehicles (AUVs). To avoid irreparable damage to the AUV caused by fishing nets, the AUV needs to be able to identify and locate them autonomously and avoid them in advance. Whether the AUV can avoid fishing nets successfully depends on the accuracy and efficiency of detection. In this paper, we propose an object detection multiple receptive field network (MRF-Net), which is used to recognize and locate fishing nets using forward-looking sonar (FLS) images. The proposed architecture is a center-point-based detector, which uses a novel encoder-decoder structure to extract features and predict the center points and bounding box size. In addition, to reduce the interference of reverberation and speckle noises in the FLS image, we used a series of preprocessing operations to reduce the noises. We trained and tested the network with data collected in the sea using a Gemini 720i multi-beam forward-looking sonar and compared it with state-of-the-art networks for object detection. In order to further prove that our detector can be applied to the actual detection task, we also carried out the experiment of detecting and avoiding fishing nets in real-time in the sea with the embedded single board computer (SBC) module and the NVIDIA Jetson AGX Xavier embedded system of the AUV platform in our lab. The experimental results show that in terms of computational complexity, inference time, and prediction accuracy, MRF-Net is better than state-of-the-art networks. In addition, our fishing net avoidance experiment results indicate that the detection results of MRF-Net can support the accurate operation of the later obstacle avoidance algorithm.

show abstract

Section: Related Workmentioning

confidence: 99%

Multiple Receptive Field Network (MRF-Net) for Autonomous Underwater Vehicle Fishing Net Detection Using Forward-Looking Sonar Images

Qin

Zhao

Zhu

et al. 2021

Sensors

View full text Add to dashboard Cite

show abstract

“…Furthermore, marine robots are also being exploited to collect optical and acoustic imagery that can be analyzed by a human operator to monitor areas of interest and detect particular patterns that highlight the ongoing physical-chemical processes. Modern Automatic Target Recognition (ATR) strategies, based on Deep Neural Networks (DNN), applied to both optical and acoustic images [56][57][58], represent a possible and promising solution to help (or even replace) human interventions in these particular tasks.…”

Section: Payloadsmentioning

confidence: 99%

Marine Robotics for Recurrent Morphological Investigations of Micro-Tidal Marine-Coastal Environments. A Point of View

Ridolfi

Secciani

Stroobant

et al. 2021

JMSE

Self Cite

View full text Add to dashboard Cite

Coastal zones are subjected to a wide range of phenomena acting on very different temporal and spatial scales: from decades to days and from hundreds of kilometers to tens of meters. Planning the management of such areas, thus, requires an accurate and updated knowledge of the ongoing processes. While standard monitoring activities are functional for the medium-long time scale and medium-large spatial scale, they struggle to provide adequate information concerning the short period (i.e., days) and small range (i.e., few meters). In addition, such operations are affected by high costs and logistic complexity since they generally involve the deployment of specific aircraft or maritime vehicles. On the contrary, the employment of robotic devices can represent a solution to these issues. Their proper use can allow for frequent surveys and enhance the coverage of the acquired data due to optimized mission strategies. Marine robotics has the potential to arise as an efficient complementary tool to standard monitoring techniques. Nevertheless, the use of marine robots is still limited and should be improved. The purpose of this paper is to discuss the current state of robotic technology, identifying both the benefits and shortcomings of its use for micro-tidal marine-coastal monitoring. The discussion will be supported by actual results, taken as an example, achieved using FeelHippo AUV, the compact Autonomous Underwater Vehicle (AUV) developed by the Department of Industrial Engineering at the University of Florence, Italy.

show abstract

“…In these tasks, the vehicle is used to collect optical as well as acoustic images, using cameras and imaging SONARs. Then, the data could be used for underwater surveillance purposes (Terracciano et al, 2020), where image processing techniques and modern Deep Learning (DL) methodologies can accurately find targets of interest (see Jin et al, 2019; Zacchini, Franchi, et al, 2020; Zacchini, Ridolfi, Topini, et al, 2020), and for archeological investigations, where optical (Allotta, Costanzi, et al, 2016) and acoustic reconstructions (Franchi et al, 2018) emerged as an essential tool to correctly classifying historical finds.…”

Section: Introductionmentioning

confidence: 99%

Sensor‐driven autonomous underwater inspections: A receding‐horizon RRT‐based view planning solution for AUVs

Zacchini

Franchi

Ridolfi

2022

Journal of Field Robotics

Self Cite

View full text Add to dashboard Cite

Autonomous Underwater Vehicles (AUVs) are used by the scientific community for various applications, from collecting well-distributed high-quality data to mapping the seafloor or exploring unknown areas. Nonpredictable environmental conditions and sensor acquisitions make the design of AUV surveys challenging even for expert operators. Multiple attempts are required, and the collected data quality is not guaranteed: The AUV passively stores the sensors' acquisitions that are then analyzed offline after its recovery. In Forward-Looking SONAR (FLS) seabed inspections, the vehicle follows lawnmower paths designed by an expert operator that considers the sensor characteristics. The performance of FLSs is affected by several environmental conditions and possible protruding objects. This paper presents a probabilistic framework for FLS-based seabed inspections that endow the AUV with the ability to autonomously conducting the survey and ensure adequate coverage of the target area. A three-dimensional probabilistic occupancy mapping system for FLS reconstructions to update the covered area map was developed. The map is used by the Coverage Path Planning (CPP) algorithm to evaluate the visibility of the viewpoints that are generated as nodes of a random tree. The Next-Best Viewpoint (NBV) is selected as the first node in the branch expected to collect more data, and the path to reach the NBV is computed using the rapidly exploring random tree (RRT*) algorithm. The sensor-driven coverage approach is used in a receding-horizon manner. The proposed Receding-Horizon Coverage Approach was validated with simulations and real prerecorded data. Finally, the framework was used online during an experimental campaign where several FLS seabed inspections were performed.

show abstract

Forward-Looking Sonar CNN-based Automatic Target Recognition: an experimental campaign with FeelHippo AUV

Cited by 23 publications

References 11 publications

Multiple Receptive Field Network (MRF-Net) for Autonomous Underwater Vehicle Fishing Net Detection Using Forward-Looking Sonar Images

Multiple Receptive Field Network (MRF-Net) for Autonomous Underwater Vehicle Fishing Net Detection Using Forward-Looking Sonar Images

Marine Robotics for Recurrent Morphological Investigations of Micro-Tidal Marine-Coastal Environments. A Point of View

Sensor‐driven autonomous underwater inspections: A receding‐horizon RRT‐based view planning solution for AUVs

Contact Info

Product

Resources

About