• crack detection
• UAV
• real-time structural inspection method
• Structural Health Monitoring

One of the paramount facets of Structural Health Monitoring (SHM) entails the automated detection of structural damage in buildings. Since 2012, several methods have been studied on the use of UAVs for automatic damage detection. The proposed approach offers an alternative solution to the problem of automatic damage detection by enabling the UAV to adjust its motion during the inspection process autonomously, exploiting information from onboard sensors. The proposed method is entirely implemented and tested in the ROS environment. The building inspection algorithm relies on signal processing from a neural network that performs real-time damage detection on images captured by a camera installed onboard the UAV. This method differs from classic approaches in that it defines two levels of distance from the wall during the building inspection: the “coverage distance” and the “damage exploration distance”. The first level corresponds to the distance at which a coverage roadmap is defined, which serves as a reference path for the UAV to follow during the inspection. The second level, closest to the wall, corresponds to the distance at which the UAV searches for damage boundaries and acquires images for post-flight analysis. The UAV automatically switches between these two levels, guided by the damage identified by the neural network. The proposed approach is compared with classic methods, which use a pre-defined roadmap at a fixed distance from the wall, in a simulation scenario involving a structure with vertical and planar walls. The results confirm that the proposed approach provides an improvement in the image quality of the captured damages and it is more efficient in terms of inspection time for the same number of damages identified.