Python-based Computer Vision for Non-Intrusive River Level Measurement

The level of the rivers in some highly populated cities grows rapidly, endangering the nearby communities, due to the intense rains that fall in a short time and raise these levels several meters in a matter of a few minutes. Civil protection authorities request an early warning system for potential floods. Taking advantage of the potential of Python, a program has been developed that processes the images obtained by a camera, without the need to install any type of additional sensor in the river as is the case with other projects. To process the images, flood-filling algorithms, contour algorithms and an own design related to the trajectory of expansion points were applied to detect the areas that have water. The system, in operation, has given very satisfactory results with high congruence between the point indicating the water limit in the processed image and the value of the recorded levels.