Automatic Aquifer Detection Using YOLOv8 Algorithms and Image Processing Techniques Amir Tag Departement of Electrical and Electronics Engineering Sorbonne Universit´ e Paris, France amir.tag@etu.sorbonne-universite.fr Tamer Khattab Departement of Electrical and Computer Engineering Qatar University Doha, Qatar tkhattab@qu.edu.qa Abstract—This study leverages the YOLOv8m computer vision algorithm with Ground Penetrating Radar (GPR) for enhanced aquifer detection in simulated Qatari terrains. Using image pro- cessing techniques like the Canny operator, a detection accuracy of 72% was achieved. Different radar resolutions were tested, highlighting the potential of Transfer Learning for improvement. While simulations show promise, real-world deployment remains an essential next step. Index Terms—Ground Penetrating Radar (GPR),Computer Vision Algorithms (CVA),Image Processing I. I NTRODUCTION The ground-penetrating radar (GPR) is a RADAR used for non-destructive and non-invasive geophysical imaging. It is mainly used to characterize the electrical and magnetic prop- erties of the subsurface, as well as for locating and mapping various buried/hidden artifacts [1]. The GPR emits electro- magnetic (EM) waves toward the surface of the structure being studied. When these EM waves encounter discontinuities in the electromagnetic properties of the medium, they are reflected and refracted according to Snell’s law. The reflected waves return to the surface to be captured by the GPR, while the refracted waves continue to propagate deeper into the medium until all energy is dissipated or all the waves are reflected back to the surface. The time interval between the emission and reception of an electromagnetic wave provides information about the depth and location of the discontinuity. Most GPR applications use time-domain radars, with particular emphasis on pulsed radar [2]. The reflected/scattered electromagnetic waves are sampled by the receiver and are represented in an amplitude trace over time. Freshwater scarcity is a pressing issue in arid regions like Qatar [3]. In these regions, water is a vital but limited resource. Moreover, due to severe climatic conditions, surface water is scarce, making groundwater even more essential to support populations and economic activities. However, detecting groundwater can be a challenge in these regions. Traditional methods like drilling wells can be expensive, have environmental impacts, and only provide information on specific sites. Therefore, it is necessary to Identify applicable funding agency here. If none, delete this. develop and adopt more efficient and precise techniques to detect groundwater. In this context, this paper proposes the application of ground-penetrating radar (GPR) as a potential technique for groundwater detection in Qatar. The primary objective of this research would be to apply Deep Learning to Ground Penetrating Radar (GPR) data to assist GPR experts in their in- terpretations. These algorithms have the potential to recognize complex and subtle patterns in data that might be difficult to perceive. Moreover, acquiring GPR images is time-consuming and costly [4]. The second objective would be to develop an algorithm that creates realistic scenarios to generate B-scans for use as input to the computer vision algorithm. Computer vision algorithms (CVA) have the ability to provide predictions and increase the effectiveness of interpreting GPR data. The goal is to optimize a CVA to detect aquifers in GPR images with at least 70% accuracy, as the traditional technique for detecting the presence of an object/material on a GPR image is by the human eye. Manually detecting material presence is time-consuming, and detection accuracy without expertise is around 65% [5]. By training these algorithms on a large amount of GPR data, they could learn to detect the presence of groundwater with high accuracy, significantly improving water resource management in arid regions like Qatar. This could also have applications for searching for aquifers on other planets with geological structures similar to arid regions. II. THE GROUND PENETRATING RADAR (GPR) A. Workings Of The GPR A Ground Penetrating Radar (GPR) system is conceptually very simple. It is important to list and analyze its main components. A GPR consists of two antennas, a transmitter module, a receiver module, a control unit, and a display/in- terface/data storage unit. Figure 1 shows how the components are connected to each other. The electrical connections to the transmitter and receiver modules are made with optical cables to ensure isolation from electromagnetic-related issues. The heart of the system is the synchronization unit, which controls the generation of the radar signal and then detects the received signals based on time.