Estimation of Water Level Changes Using Image Processing and Bee Colony Algorithm

To determine the status of water resources management, water level estimation is essential. Determining water level is important to reduce the effects of floods. Given that determining water level is expensive and time-consuming, in many cases it is difficult to measure it. The aim of this research is to use a common smartphone camera and evaluate the effectiveness of an image-based water level detection system and add knowledge to this topic. In this study, the capabilities of various machine learning models, support vector machine model and three hybrid models including invasive weeds (IWO), bee colony optimization algorithm (ABC), cultural algorithm (CA), which are the most applicable algorithms in modeling, were used. . The results of intelligent modeling showed that the best water level estimation results are related to SVM-ABC with more accuracy and less error (RMSE=5.49 MAE=4.29). These results show that metaheuristic models increase the speed of calculations and reduce the error rate in estimating water levels in areas that are difficult to navigate and difficult to perform correctly and with high accuracy.
Introduction
Due to the importance of continuous control of water level in rivers and irrigation canals and preventing any critical events like floods and erosion, it is necessary to create a system to satisfy the limitations of manual measurement and have high capability in detection and measurement(Abdi et al. 2024 ).
Ashori et al. (2021) in order to measure the amount of copper in drinking water , image processing based on deep neural network was used .in addition , 70 percent of the data in the training phase , 15 percent validation phase and 15 percent of the proposed system test phase with python software and Py Torch and open CV . results showed a high accuracy of 97 % for detecting heavy metal contamination in drinking water. Akhoni pourhosseini et al.(2023) water quality of Babolrood river was investigated in Iran .In this study , a number of hybrid models using support vector machine including cultural optimization , Harmony algorithm and learning and learning were used to predict tds in Babolrood river , Iran .The measured and unpublished monthly data included Ca , Mg , Na , HCO₃ , SO₄ , cl , pH , and TDS from 1968 to 2016 . the results were analyzed using several statistical indices . SVM model was used to predict the water quality in the basin , compared to other models .
The purpose of this study is to use an ordinary smartphone camera and evaluate the effectiveness of an image - based water - level detection system and adding knowledge about this subject .The main focus of this research is using capabilities of different machine learning models , support vector machine model and invasive weed control, optimization algorithm of honey bee colony and cultural algorithm for accurate estimation of water surface by processing and image analysis .
Materials and Methods
The first step involves visualization of images. Conversion of graphic images to images is one of the processing processes in image processing .This process is a fundamental aspect of digital image processing, in which graphical data are converted into digital form and are used to extract useful information or enhance image quality .In this research , the measurements and observations in the hydraulic experiment of hydraulic engineering department of tabriz university were studied .In this study , 244 images were used in two cases, smooth and opaque water ( with sediment ) to model and estimate water surface. 201 images related to training and 43 images are tested .the methods used for the simulation include support vector machine and metaheuristic algorithms .The basis of the proposed method is based on the statistical learning theory which follows the minimization of structural risk .Artificial bee colony is an optimization method inspired by the social behavior of honey bees in which bees are collectively looking for optimal food resources in the search space .Genetic algorithms are a genetic algorithm that in addition to the population space based on darwinian evolution , have a belief space .this belief space represents knowledge and experience collected from the best members of society and acts as a reference to guide the process of evolution .The method of optimization of invasive weed by Mehrabiyan and lucas was introduced and used as an inspiration to nature. This algorithm is very effective in finding the optimal locations and is based on the natural and natural features of weeds in a colony. After data analysis with different methods and models ,it is necessary to evaluate its performance. In this research ,the most applicable criteria in water engineering including determination coefficient , root mean square error and mean absolute error .

Results and Discussion
After omitting the outliers , the data were converted to normal data to increase the accuracy of modeling. Then, by using the metaheuristic methods , support vector machine including invasive weed, optimization algorithm of honey bee colony, cultural algorithm was simulated .we can find that the meta - arithmetic model of SVM-ABC has the best performance with respect to evaluation metrics .with respect to the measured criteria used in the paper , SVM-ABC model has been able to perform the simulation of the water level in a satisfactory way than other models .In the validation section , the coefficient of determination of 0.999 and root mean square error ( RMSE ) 5 .49 mm and mean absolute error of 4.29 mm was obtained which is acceptable
Conclusion
Due to the importance of automatic and continuous control of water level for planning , decision making and management of critical situations such as flood and river flooding , in this research we have tried to identify the water level in the images by image processing and simulation by using image processing and simulation using the support vector machine including SVM - IWO , SVM - CA and SVM - ABC .In the modeling section , the SVM-ABC model in the validation section , the coefficient of determination of R2 of 0.999 and root mean square error of 5.49 mm and mean absolute error of 4.29 mm , respectively.our proposed approach is based on the proposed approach , which is based on the proposed approach .also , we can say that meta - heuristic models can be used to estimate water level in comparison with the LS - SVM and BNN models .