Urban Road Flooding Detection System based on SVM Algorithm

Abstract:
In order to solve the frequent flooding in urban areas caused by extreme rainfall weather, the traditional solution is manual detection of waterlogging, which is inefficient and risky. While the detection system based on deep learning convolution neural networks not only takes too long to operate, but also has very high requirements for CPU computing power. Thus, it is not suitable for grassroots areas characterized by insufficient hardware base. Given this, a method of detecting urban flood based on SVM algorithm in machine learning for extraction and matching of cumulative features on the input road image dataset is applied in this research. Through multiple training and validation of the model, the SVM algorithm can achieve 96% accuracy in the binary classification of the presence or absence of standing water, which can accurately and quickly identify the presence or absence of standing water in road images captured from the actual road environment, providing technical support to city managers for monitoring floods and having application value for future urban planning and construction.