Research on water quality classification of unbalanced river image data under deep learning framework

Water quality is an important indicator of current river ecological monitoring. Water quality monitoring based on image recognition can greatly save monitoring costs and time. But the river image data are mostly unbalanced data, which will affect the accuracy of image recognition. Aiming at this problem, this paper constructs a deep learning framework for water quality classification of unbalanced river image data. According to the imbalance degree of samples, adaptive unbalanced sampling processing is carried out, and then a convolutional neural network model is constructed to classify water quality of balanced data sets. Through comparison experiments with a variety of deep learning models, this framework has achieved good results, indicating that the method can effectively classify water quality based on images.