Interpretation of georadar data based on machine learning technologies

Abstract

In the laboratory analysis of geological data, a number of problems arise due to the insufficient accuracy of the results. For example, different laboratories may provide different end results for the same samples, which creates a problem. This can lead to unreliable results, which can ultimately reduce the quality of the assessment. Machine learning allows to speed up the processing of radar data, as well as avoid the above-mentioned "misunderstandings". The problem of conducting scientific research at specialized landfills for a comprehensive assessment of the possibilities of using computer technology in the interpretation of georadar profiles is urgent. This makes it possible to objectively evaluate the result of automatic interpretation of georadar data. The several machine-learning algorithms described in the article are designing to improve the analysis and interpretation of data by incorporating various methods for optimizing georadar data processing processes. These methods include regression, classification and clustering. By incorporating these methods of optimizing the processing of georadar data into several machine-learning algorithms, the software can provide a comprehensive analysis and interpretation of the data obtained. This allows for a better understanding of the relationships, patterns and trends in the data, which ultimately leads to more informed decision-making and improved understanding. To improve the understanding of the results, the following quantitative indicators were obtained: correlation coefficient – 0.7072, determination coefficient – 0.5001, all these indicators correspond to these models. The deviation from the regression line is on average 22.37 units. Based on the classification results, the soil was determined to be wet. Errors in the sets do not exceed 1 %