The location and characterization of water runoff locations in canal networks is of critical importance in order to properly forecast and manage floods. Heavy rains in upstream areas can suddenly increase the rate of discharge resulting in events such as overflow, seepage losses, erosion, or flooding. The inability to simulate floods in the actual terrain often results in the actual floods developing in unexpected patterns. Thus, actual floods have been documented to occur in the inhabited side of a canal as opposed to the uninhabited embankment where managers had planned to occur. Also, warning alarms have only triggered once the flood had occurred. Additionally, nature- and human-made activities (e.g., driving trucks or cars) result in the loss of soil, creation of uneven surfaces, and erosion of edges on the canal embankment, and, overall, change the embankment profile and can alter the water runoff outlet over time. Currently, though, the manual localization of water runoff escape points is often overseen in large infrastructure networks since it demands a time-consuming, labor intensive, and prone-to-error surveying effort. The efforts in the ongoing study presented in this paper introduce a methodology to automatically detect the lowest points along canal embankments. High-resolution raster images and 3D point cloud representation of the existing canal infrastructure and surrounding areas, produced with above-the-ground photogrammetric sensors, are collected along the canals. Then, geometric algorithm, such as random sample consensus (RANSAC) is used to analyze the sensed data. This paper presents the preliminary results of an ongoing research study, showing the elevation and coordinates for the lowest and near-lowest escape outlets. Such results promise to minimize soil erosion and improve the predictability and effectiveness of flood monitoring approaches.