This paper presents a novel deep learning method, called DC-Fire, for recognizing wildland fires using aerial infrared images. Experimental results showed that DC-Fire achieved high performance with an accuracy of 100% and an F1-score of 100% using a large dataset and data augmentation techniques, better than classical machine learning and baseline CNN methods. In addition, DC-Fire demonstrated its potential in detecting smoke and flames, surpassing challenges including small areas of fire/smoke, background complexity, and the variability of forest fires in terms of size, shape, and intensity.

This work presents a new concept for permanent markers consisting of RFID tags built into aircraft components made from fibre reinforced plastics. The novelty of this concept lies in the antenna design, which represents a geometric point that can accessed using commonly employed non-destructive testing (NDT) methods, specifically suited for thermovision. These geometric points provide a reliable frame of reference for accurate data localization throughout the entire life cycle of the components. Initial results obtained with an antenna design incorporated into a glass-reinforced epoxy laminate quantify the anticipated localization accuracy achieved through our concept. We utilize a geometrically calibrated, off-the-shelf bolometer and standard lock-in thermography in our approach.

As wind power running, problems such as damage may occur, in order to keep the system running normally and safely, wind turbine need to be regularly maintained. The current main treatment is to detect the blades by carrying an infrared camera on board a UAV. However, as the infrared images are suffering interference in the field environment, causing of inhomogeneity, this paper proposes a fast way of segmenting the inhomogeneous images, by combining the local and global information of the infrared images, to achieve a correction of the bias field and an accurate segmentation of the defects.