Underwater Target Detection and Localization with Feature Map and CNN-Based Classification

Guo, Tiantian; Song, Yunze; Kong, Zejian; Lim, Enggee; López-Benítez, Miguel; Ma, Fei; Yu, Limin

doi:10.1109/ctisc54888.2022.9849785

Cited by 5 publications

(3 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In 2022, Guo et al [49] applied acoustic features, Mel Frequency Cepstrum Coefficients (MFCC), and Gamma Pass Frequency Cepstrum Coefficients (GFCC) to underwater signal classification and proposed a model combining deterministic and statistical models. The geometric channel model helps to generate databases for different geometric set- In 2009, Wu et al [45] proposed a convolutional network (ECNet) for the semantic segmentation of side-scan sonar images that was fast and had few parameters.…”

Section: Convolutional Neural Network (Cnn)mentioning

confidence: 99%

“…In 2022, Guo et al [49] applied acoustic features, Mel Frequency Cepstrum Coefficients (MFCC), and Gamma Pass Frequency Cepstrum Coefficients (GFCC) to underwater signal classification and proposed a model combining deterministic and statistical models. The geometric channel model helps to generate databases for different geometric settings, and the effectiveness of its systematic framework is verified by comparing it with continuous wavelet transform (CWT) and short-time Fourier transform (STFT) using a CNN as a classifier.…”

Section: Convolutional Neural Network (Cnn)mentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning-Based Classification of Raw Hydroacoustic Signal: A Review

Dong

2022

JMSE

View full text Add to dashboard Cite

Underwater target recognition is a research component that is crucial to realizing crewless underwater detection missions and has significant prospects in both civil and military applications. This paper provides a comprehensive description of the current stage of deep-learning methods with respect to raw hydroacoustic data classification, focusing mainly on the variety and recognition of vessels and environmental noise from raw hydroacoustic data. This work not only aims to describe the latest research progress in this field but also summarizes three main elements of the current stage of development: feature extraction in the time and frequency domains, data enhancement by neural networks, and feature classification based on deep learning. In this paper, we analyze and discuss the process of hydroacoustic signal processing; demonstrate that the method of feature fusion can be used in the pre-processing stage in classification and recognition algorithms based on raw hydroacoustic data, which can significantly improve target recognition accuracy; show that data enhancement algorithms can be used to improve the efficiency of recognition in complex environments in terms of deep learning network structure; and further discuss the field’s future development directions.

show abstract

Section: Convolutional Neural Network (Cnn)mentioning

confidence: 99%

“…In 2022, Guo et al [49] applied acoustic features, Mel Frequency Cepstrum Coefficients (MFCC), and Gamma Pass Frequency Cepstrum Coefficients (GFCC) to underwater signal classification and proposed a model combining deterministic and statistical models. The geometric channel model helps to generate databases for different geometric settings, and the effectiveness of its systematic framework is verified by comparing it with continuous wavelet transform (CWT) and short-time Fourier transform (STFT) using a CNN as a classifier.…”

Section: Convolutional Neural Network (Cnn)mentioning

confidence: 99%

Deep Learning-Based Classification of Raw Hydroacoustic Signal: A Review

Dong

2022

JMSE

View full text Add to dashboard Cite

show abstract

“…Wang et al [10] proposed an underwater target image edge detection algorithm based on ant colony optimization and reinforcement learning, which can effectively extract underwater contour information, better maintain image texture, and has ideal anti-interference performance. Guo et al [11] proposed an underwater target detection and localization method using feature maps and CNN-based classification, which is superior in underwater signal classification and target localization. Zeng et al [12] proposed an underwater target detection based on Faster R-CNN and adversarial occlusion network, which can obtain better robustness for underwater seafood.…”

Section: Introductionmentioning

confidence: 99%

An Improved YOLOv5s-Based Scheme for Target Detection in a Complex Underwater Environment

Hou

Guan

Guo

et al. 2023

JMSE

View full text Add to dashboard Cite

At present, sea cucumbers, sea urchins, and other seafood products have become increasingly significant in the seafood aquaculture industry. In traditional fishing operations, divers go underwater for fishing, and the complex underwater environment can cause harm to the divers’ bodies. Therefore, the use of underwater robots for seafood fishing has become a current trend. During the fishing process, underwater fishing robots rely on vision to accurately detect sea cucumbers and sea urchins. In this paper, an algorithm for the target detection of sea cucumbers and sea urchins in complex underwater environments is proposed based on the improved YOLOv5s. The following improvements are mainly carried out in YOLOv5s: (1) To enhance the feature extraction ability of the model, the gnConv-based self-attentive sublayer HorBlock module is proposed to be added to the backbone network. (2) To obtain the optimal hyperparameters of the model for underwater datasets, hyperparameter evolution based on the genetic algorithm is proposed. (3) The underwater dataset is extended using offline data augmentation. The dataset used in the experiment is created in a real underwater environment. The total number of created datasets is 1536, and the training, validation, and test sets are randomly divided according to the ratio of 7:2:1. The divided dataset is input to the improved YOLOv5s network for training. The experiment shows that the mean average precision (mAP) of the algorithm is 94%, and the mAP of the improved YOLOv5s model rises by 4.5% compared to the original YOLOv5s. The detection speed increases by 4.09 ms, which is in the acceptable range compared to the accuracy improvement. Therefore, the improved YOLOv5s has better detection accuracy and speed in complex underwater environments, and can provide theoretical support for the underwater operations of underwater fishing robots.

show abstract