Threshold-Based BRISQUE-Assisted Deep Learning for Enhancing Crack Detection in Concrete Structures

Automated visual inspection has made significant advancements in the detection of cracks on the surfaces of concrete structures. However, low-quality images significantly affect the classification performance of convolutional neural networks (CNNs). Therefore, it is essential to evaluate the suitability of image datasets used in deep learning models, like Visual Geometry Group 16 (VGG16), for accurate crack detection. This study explores the sensitivity of the BRISQUE method to different types of image degradations, such as Gaussian noise and Gaussian blur. By evaluating the performance of the VGG16 model on these degraded datasets with varying levels of noise and blur, a correlation is established between image degradation and BRISQUE scores. The results demonstrate that images with lower BRISQUE scores achieve higher accuracy, F1 score, and Matthew's correlation coefficient (MCC) in crack classification. The study proposes the implementation of a BRISQUE score threshold (BT) to optimise training and testing times, leading to reduced computational costs. These findings have significant implications for enhancing accuracy and reliability in automated visual inspection systems for crack detection and structural health monitoring (SHM).

Autonomous Vision-Based Primary Distribution Systems Porcelain Insulators Inspection Using UAVs

The early detection of damaged (partially broken) outdoor insulators in primary distribution systems is of paramount importance for continuous electricity supply and public safety. Unmanned aerial vehicles (UAVs) present a safer, autonomous, and efficient way to examine the power system components without closing the power distribution system. In this work, a novel dataset is designed by capturing real images using UAVs and manually generated images collected to overcome the data insufficiency problem. A deep Laplacian pyramid-based super-resolution network is implemented to reconstruct high-resolution training images. To improve the visibility of low-light images, a low-light image enhancement technique is used for the robust exposure correction of the training images. A different fine-tuning strategy is implemented for fine-tuning the object detection model to increase detection accuracy for the specific faulty insulators. Several flight path strategies are proposed to overcome the shuttering effect of insulators, along with providing a less complex and time- and energy-efficient approach for capturing a video stream of the power system components. The performance of different object detection models is presented for selecting the most suitable one for fine-tuning on the specific faulty insulator dataset. For the detection of damaged insulators, our proposed method achieved an F1-score of 0.81 and 0.77 on two different datasets and presents a simple and more efficient flight strategy. Our approach is based on real aerial inspection of in-service porcelain insulators by extensive evaluation of several video sequences showing robust fault recognition and diagnostic capabilities. Our approach is demonstrated on data acquired by a drone in Swat, Pakistan.