Applicability Evaluation of the Hydrological Image and Convolution Neural Network for Prediction of the Biochemical Oxygen Demand and Total Phosphorus Loads in Agricultural Areas

Abstract

This study employed a convolution neural network (CNN) model, hitherto used only for solving classification problems, with two-dimensional input data to predict the pollution loads and evaluate the CNN model’s applicability. A CNN model generally requires two-dimension input data, such as photographs in previous studies. However, this study’s CNN model necessitates the numerical images that reflect hydrological phenomena due to the nature of the study. A hydrological image was used as the input data for the CNN model in this study to address this issue. The last layer of the CNN model was also transformed into a linear function to derive the continuous variable. As a result, the Pearson correlation coefficient, which represents the relationship between the measured and predicted values, demonstrated a Biochemical Oxygen Demand (BOD) load model of 0.94 and a Total Phosphorus (TP) load model of 0.87. Nash–Sutcliffe efficiency was used to evaluate the model performance; the BOD load model was 0.83, while the TP load model was 0.79, respectively, indicating good performance. These results demonstrate that the hydrological images led to stable model learning and generalization, and the proposed CNN model is suitable for predicting the pollution load, with potential future applications in various fields.

Detecting the Early Flowering Stage of Tea Chrysanthemum Using the F-YOLO Model

“…Each prediction module had three prior anchors of different scales. Through k-means clustering [39], nine prior anchors- (21,10), (31,16), (34,41), (45,21), (60,28), (60,78), (76,38), (108,52), and (190,101)-of the chrysanthemum dataset were obtained. At the same time, the data of four images were computed, so that the minibatch size did not need to be large, and one GPU could achieve good computation results.…”

“…The detection results not only serve as a guide for the manipulator to harvest chrysanthemum in the subsequent operation, but also determine the detection accuracy in chrysanthemum harvesting. Although in recent years, methods based on deep convolutional neural networks (CNNs) have made remarkable achievements in object detection tasks [4][5][6][7][8][9][10], under agricultural application scenarios, it is still difficult to build a lightweight network for a selective harvesting robot that can adapt to complex unstructured scenarios.…”

A Study on TiO2 Surface Texturing Effect for the Enhancement of Photocatalytic Reaction in a Total Phosphorous Concentration Measurement System

“…Accordingly, early detection and response to eutrophication are priorities for water quality management in rivers and water supply sources. Furthermore, phosphorus, one of the causes of eutrophication, is also an indicator of water pollution [8][9][10]. Therefore, research is being actively conducted worldwide to develop a small total phosphorus monitoring system with real-time measurement to prevent eutrophication in advance [11].…”