PNNL

Abstract

1. Adult American eels (Anguilla rostrata) are vulnerable to hydropower turbine mortality during outmigration from inland waters to the sea. Due to the protracted, episodic nature of adult eel outmigration, cost-effective monitoring requires a high degree of automation for data analysis. 2. The imaging sonar technology is a reliable and proven technology for fish passage and migration monitoring, but there doesn’t exist an efficient, automated detection framework. We designed a machine-learning based method for automated detection of adult American eels from the Adaptive Resolution Imaging Sonar (ARIS) data. The method employs Convolution Neural Network (CNN), a powerful deep learning method for image classification, to distinguish between images of eels and non-eel moving objects. 3. Prior to image classification with CNN, sonar images were preprocessed with background subtraction and wavelet denoising to enhance sonar images. Background subtraction removed the static background structures and part of the environmental noise. Wavelet transform denoised sonar images further, by removing high frequency components. The image preprocessing helped increase the overall image classification accuracy significantly. 4. The designed CNN model was first trained and tested on data obtained from a controlled laboratory experiment, which yielded overall accuracy of >98% for image-based classification. Then, the model was applied to distinguish eels from non-eel objects with field data obtained near the Iroquois Dam located on the St. Lawrence River. The CNN model achieved accuracies commensurate with human experts. 5. The developed framework, by integrating image enhancement, object separation, and CNN classification, can be generalized for automatic monitoring of fish passage and migration using other imaging sonars like ARIS, and will benefit the design of ecology-friendly hydroelectric projects. The developed wavelet and CNN model configuration parameters can also be transferred to applications of identifying eels in similar riverine environments.