Ensemble Neural Network Based Fish Species Identification for Emerging Aquaculture Application DOI

Siddika Sabahath Anjum,

K M Shruthi,

Naziya Farheen H S

et al.

Published: Nov. 23, 2023

Precision fish farming is a novel idea in the aquaculture research and development industry that integrates modern technology pre-data processing methodologies inorder to allow data-driven decisions precise farming. The objectives are use contactless technologies automate process of monitoring fish, improve infrastructure, marine environment. Individualized treatment, biomass measurement, status assessment all depend on ability identify specific same species within group frame. Recent studies have demonstrated body patterns used for individual identification, but there currently no system automation this process. We here introduced method entirely identification based their characteristics. Therefore, we adopted Ensemble MobileNetV3 NASNets, also known as (Ens_MobV3+NasNet), excusively classification present Deep Vision trawl camera images. Speeded-up robust features, which realistic simulations images, increase speed. Fish4Knowledge dataset study, it discovered suggested Ens_MobV3+NasNet achieves 98.0% accuracy, 95% precision, 94.3% recall, 91% F1-score.

Language: Английский

Efficient One-Dimensional Network Design Method for Underwater Acoustic Target Recognition DOI Creative Commons

Qing Xue Huang,

Xiaoyan Zhang, Anqi Jin

et al.

Journal of Marine Science and Engineering, Journal Year: 2025, Volume and Issue: 13(3), P. 599 - 599

Published: March 18, 2025

Many studies have used various time-frequency feature extraction methods to convert ship-radiated noise into three-dimensional (3D) data suitable for computer vision (CV) models, which shown good results in public datasets. However, traditional engineering (FE) has been enhanced interface matching–feature (IM-FE). This approach requires considerable effort design, larger sample duration, or a higher upper limit of frequency. In this context, paper proposes one-dimensional network design underwater acoustic target recognition (UATR-ND1D), only combined with fast Fourier transform (FFT), can effectively alleviate the problem IM-FE. method is abbreviated as FFT-UATR-ND1D. FFT-UATR-ND1D was applied network, named ResNet1D. Experiments were conducted on two mainstream datasets, using ResNet1D 4320 and 360 tests, respectively. The lightweight model ResNet1D_S, 0.17 M parameters 3.4 floating point operations (FLOPs), achieved average accuracies 97.2% 95.20%. model, ResNet1D_B, 2.1 5.0 FLOPs, both reached optimal accuracies, 98.81% 98.42%, Compared existing methods, those similar parameter sizes performed 3–5% worse than proposed paper. Additionally, achieving rates require more 1 2 orders magnitude FLOPs.

Language: Английский

Citations

0
Fish Disease Detection in Aquaculture using Pseudo Hamiltonian Neural network Optimized with Philippine Eagle Optimization Algorithm DOI

Mr. Prasanna kumar M,

Saravana Kumar K,

K. P.

et al.

Knowledge-Based Systems, Journal Year: 2025, Volume and Issue: unknown, P. 113374 - 113374

Published: March 1, 2025

Language: Английский

Citations

0
The capacity of imaging sonar for quantifying the abundance, species richness, and size of reef fish assemblages DOI Creative Commons

ECP Sibley,

Alethea S. Madgett,

TS Elsdon

et al.

Marine Ecology Progress Series, Journal Year: 2023, Volume and Issue: 717, P. 157 - 179

Published: July 19, 2023

Reef fish assemblages across tropical and temperate latitudes are increasingly threatened by human impacts climate change. Accurate efficient survey methods essential for quantifying these communities to inform management strategies. Imaging sonars (ISs) high-frequency acoustic devices that produce camera-like images of objects. Unlike optical instruments, IS functions effectively in turbid dark water has proven valuable detecting fishes poor visibility at night. Abundance, species richness, size desirable metrics most reef surveys. This review investigates previous attempts quantify using different habitats. These often quantified comparison with alternative (e.g. cameras, extractive techniques). examines the causes agreement or incongruence between estimates from alternatives. Any instrument employed should be able operate structurally complex habitats, thus ability function circumstances is also reviewed. Finally, 5 notable limitations described solutions discussed. Overall, this underlines net value surveying but advises complement abundance, size.

Language: Английский

Citations

8
Boosting fish counting in sonar images with global attention and point supervision DOI
Yunhong Duan, Shubin Zhang, Yang Liu

et al.

Engineering Applications of Artificial Intelligence, Journal Year: 2023, Volume and Issue: 126, P. 107093 - 107093

Published: Sept. 11, 2023

Language: Английский

Citations

7
Enhanced deep learning models for automatic fish species identification in underwater imagery DOI Creative Commons

Dharmapuri Siri,

Gopikrishna Vellaturi,

Shaik Hussain Shaik Ibrahim

et al.

Heliyon, Journal Year: 2024, Volume and Issue: 10(15), P. e35217 - e35217

Published: July 27, 2024

Underwater cameras are crucial in marine ecology, but their data management needs automatic species identification. This study proposes a two-stage deep learning approach. First, the Unsharp Mask Filter (UMF) preprocesses images. Then, an enhanced region-based fully convolutional network (R-FCN) detects fish using two-order integrals for position-sensitive score maps and precise region of interest (PS-Pr-RoI) pooling accuracy. The second stage integrates ShuffleNetV2 with Squeeze Excitation (SE) module, forming Improved model, enhancing classification focus. Hyperparameters optimized Enhanced Northern Goshawk Optimization Algorithm (ENGO). improved R-FCN model achieves 99.94 % accuracy, 99.58 precision recall, 99.27 F-measure on Fish4knowledge dataset. Similarly, ENGO-based evaluated same dataset, shows 99.93 99.19 precision, 98.29 98.71 F-measure, highlighting its superior

Language: Английский

Citations

2
The Convergence of AI and animal-inspired robots for ecological conservation DOI Creative Commons
Naqash Afzal, Mobeen Ur Rehman, Lakmal Seneviratne

et al.

Ecological Informatics, Journal Year: 2024, Volume and Issue: unknown, P. 102950 - 102950

Published: Dec. 1, 2024

Language: Английский

Citations

2
Applications and analytical approaches using imaging sonar for quantifying behavioural interactions among aquatic organisms and their environment DOI Creative Commons
Ryan T. Munnelly, José Carlos Castillo, Nils Olav Handegard

et al.

ICES Journal of Marine Science, Journal Year: 2023, Volume and Issue: 81(2), P. 207 - 251

Published: Nov. 30, 2023

Abstract For many aquatic animals, distributional patterns, trophodynamic relationships, and reproductive activity are driven by complex biotic abiotic ecosystem interactions that influence behaviour. Linking behavioural information to environmental stimuli stressors can, therefore, help anticipate population responses changing conditions inform management. However, is challenging obtain because sampling gears do not provide adequate spatial or temporal resolution, potentially alter behaviours. Traditionally, most studies have been laboratory experiments, while behaviour in situ has often inferred indirectly. Advancements imaging sonar technology enable the study of behaviours with potential address understudied relationships. In this review we discuss applications among a meta-analysis 155 organisms their environments. We evaluate performance for studying inter- intra-specific interactions, associations sensitive habitats low-visibility environments, traditional fisheries gears. examine data processing analytical methods used refine taxonomic manage time use autocorrelation through sub-sampling, extract metrics applied ecological processes, automating abundance estimates image classification goal providing resource researchers.

Language: Английский

Citations

6
Fish as a Deformable Solid: An Innovative Method to Characterize Fish Swimming Behavior on Acoustic Videos DOI Creative Commons
Azénor Le Quinio, François Martignac, Alexandre Girard

et al.

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 134486 - 134497

Published: Jan. 1, 2024

Acoustic cameras are increasingly used for continuous, non-intrusive recording and counting of fish passage in natural environments artificial structures such as fishways. However, analysing the large number videos recorded is time consuming. Although automatic reading processes have been developed, poor quality acoustic images, including discontinuity signal a single object, challenging. We developed an innovative method videos. Unlike previous methods, it focuses on swimming locomotion instead morphological properties fish. Each image pre-processed to remove discontinuities restore entire body cluster pixels. The set pixels then tracked retrieve movement, independent displacement fish, using mesh solid deformable model. deformation which subjected between each pair frames (i.e., body) summarised map passage. Testing dataset four species strongly suggested that maps species-dependent. These results must be extended other confirm effectiveness identification characterisation their behaviour camera records.

Language: Английский

Citations

1
Combining Imaging Sonar Counting and Underwater Camera Species Apportioning to Estimate the Number of Atlantic Salmon and Striped Bass in the Miramichi River, New Brunswick, Canada DOI Creative Commons
Jani Helminen, Tommi Linnansaari

North American Journal of Fisheries Management, Journal Year: 2023, Volume and Issue: 43(3), P. 743 - 757

Published: March 8, 2023

Abstract A combined method incorporating an imaging sonar and underwater cameras was tested for assessing the size of adult Atlantic Salmon Salmo salar Striped Bass Morone saxatilis populations in one main tributaries Miramichi River, New Brunswick, Canada. The number fish recorded with October 2019 apportioned using species ratio from camera data. estimated 358 255 when applied every day 274 337 monthly applied. counts were compared to catches a downstream index trap net values trap‐net catchability proportion ascending same tributary. Depending on values, sonar–camera between 40% 190% numbers net. For Bass, produced lower agreement (sonar–camera count = 5–24% adjusted catch) because unlike Salmon, do not deterministically migrate up tributary autumn. mostly detected overnight, trends daily similar net, most being mid‐month at end month. similarity catch new information about migration timing demonstrate that sonar–underwater combination can provide tributary‐specific timely population River.

Language: Английский

Citations

3
Diadromous fish run assessment: a double-observer model using acoustic cameras to correct imperfect detection and improve population abundance estimates DOI Creative Commons

Clarisse Boulenger,

Jean‐Marc Roussel, Laurent Beaulaton

et al.

Frontiers in Ecology and Evolution, Journal Year: 2024, Volume and Issue: 11

Published: Jan. 15, 2024

Introduction Diadromous fish populations have strongly declined over decades, and many species are protected through national international regulations. They account for less than 1% of biodiversity worldwide, but they among the most perceptible linkages between freshwater marine ecosystems. During their migration back forth, diadromous subjected to anthropogenic threats, which river damming can severely limit access vital habitats jeopardize population sustainability. Here, we developed a method based on double-observer modeling approach estimating abundance during in rivers. Methods The relies two independent synchronous records counts that were analyzed jointly thanks hierarchical Bayesian model estimate detection efficiencies daily passage. We used simulated data test robustness identify conditions under be used. was then applied empirical annual silver eel run Touques River, France. Results analysis datasets study case gives evidence provide robust,accurate, precise estimates probabilities total set dependent information provided (annual distribution passage, number observation, pairing period, etc.). Discussion Then, various counting systems, including nomad acoustic camera devices. discuss its relevance programs continuity restoration, notably quantify restoration associated with dam removals.

Language: Английский

Citations

0