Search

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

Acoustic cameras are increasingly used in monitoring studies of diadromous fish populations, even though analyzing them is time-consuming. In complex in situ contexts, anguilliform fish may be especially difficult to identify automatically using acoustic camera data because the undulation of their body frequently results in fragmented targets. Our study aimed to develop a method based on a succession of computer vision techniques, in order to automatically detect, identify and count anguilliform fish using data from multiple models of acoustic cameras. Indeed, several models of cameras, owning specific technical characteristics, are used to monitor fish populations, causing major differences in the recorded data shapes and resolutions. The method was applied to two large datasets recorded at two distinct monitoring sites with populations of European eels with different length distributions. The method yielded promising results for large eels, with more than 75% of eels automatically identified successfully using datasets from ARIS and BlueView cameras. However, only 42% of eels shorter than 60 cm were detected, with the best model performances observed for detection ranges of 4–9 m. Although improvements are required to compensate for fish-length limitations, our cross-camera method is promising for automatically detecting and counting large eels in long-term monitoring studies in complex environments.

Acoustic cameras are increasingly used in monitoring studies of diadromous fish populations, even though analyzing them is time-consuming. In complex in situ contexts, anguilliform fish may be especially difficult to identify automatically using acoustic camera data because the undulation of their body frequently results in fragmented targets. Our study aimed to develop a method based on a succession of computer vision techniques, in order to automatically detect, identify and count anguilliform fish using data from multiple models of acoustic cameras. Indeed, several models of cameras, owning specific technical characteristics, are used to monitor fish populations, causing major differences in the recorded data shapes and resolutions. The method was applied to two large datasets recorded at two distinct monitoring sites with populations of European eels with different length distributions. The method yielded promising results for large eels, with more than 75% of eels automatically identified successfully using datasets from ARIS and BlueView cameras. However, only 42% of eels shorter than 60 cm were detected, with the best model performances observed for detection ranges of 4–9 m. Although improvements are required to compensate for fish-length limitations, our cross-camera method is promising for automatically detecting and counting large eels in long-term monitoring studies in complex environments.

Acoustic cameras, or imaging sonars, are high-potential devices for many applications in aquatic ecology, notably for fisheries management and population monitoring. However, how to extract such data into high-value information without a time-consuming entire data set reading by an operator is still a challenge. Moreover, the analysis of acoustic imaging, due to its low signal-to-noise ratio, is a perfect training ground for experimenting with new approaches, especially concerning deep learning techniques. We present hereby a novel approach that takes advantage of both convolutional neural network (CNN) and classical computer vision (CV) techniques, able to detect fish passages in acoustic video streams. The pipeline pre-treats the acoustic images to localise the signals of interest and to improve the detection performances. The YOLOv3-based model was trained with fish data from multiple species recorded by the two most frequently used models of acoustic cameras, the DIDSON and ARIS, including species of high ecological interest, as Atlantic salmon or European eels. The pre-treatment of images greatly improves the model performance, increasing its F1-score from 0.52 to 0.69. The model we developed provides satisfying results detecting almost 80% of fish passages and minimising the false-positive rate. On a validation data set, 40 h of videos and around 1 800 fish passages, the efficiency increases with the fish sizes, notably reaching a recall higher than 95% for Atlantic salmon. Conversely, the model appears much less efficient for eel detections on ARIS videos than on DIDSON data (31% recall vs 75%).

Underwater acoustic cameras are high potential devices for many applications in ecology, notably for fisheries management and monitoring. However how to extract such data into high value information without a time-consuming entire dataset reading by an operator is still a challenge. Moreover the analysis of acoustic imaging, due to its low signal-to-noise ratio, is a perfect training ground for experimenting with new approaches, especially concerning Deep Learning techniques. We present hereby a novel approach that takes advantage of both CNN (Convolutional Neural Network) and classical CV (Computer Vision) techniques, able to detect a generic class ''fish'' in acoustic video streams. The pipeline pre-treats the acoustic images to extract 2 features, in order to localise the signals and improve the detection performances. To ensure the performances from an ecological point of view, we propose also a two-step validation, one to validate the results of the trainings and one to test the method on a real-world scenario. The YOLOv3-based model was trained with data of fish from multiple species recorded by the two common acoustic cameras, DIDSON and ARIS, including species of high ecological interest, as Atlantic salmon or European eels. The model we developed provides satisfying results detecting almost 80% of fish and minimizing the false positive rate, however the model is much less efficient for eel detections on ARIS videos. The first CNN pipeline for fish monitoring exploiting video data from two models of acoustic cameras satisfies most of the required features. Many challenges are still present, such as the automation of fish species identification through a multiclass model. 1 However the results point a new solution for dealing with complex data, such as sonar data, which can also be reapplied in other cases where the signal-to-noise ratio is a challenge.