Advanced Classification Technique for Defect and Fault Detection in the Electrical Industry Using Image Data

Effective defect and fault detection in image data is essential for quality assurance, particularly in industries reliant on high-performing infrastructure, like power transmission insulator defects. So far, various methods based on machine learning techniques have been developed to detect defects in images captured using UAVs. In this study the advanced techniques used in image-based fault and defect detection evaluates with a primary focus on the electrical industry. A wide range of approaches are discussed for identifying faults and defect in power transmission line images and underscoring their impact on detection accuracy and adaptability to complex environments. In this paper, a new version of the Mask R-CNN network is presented, which has been redesigned in its head architecture to satisfy these limitations. To increase the accuracy of detection, in the classification branch, a series of fully connected layers with a rhombus structure have been used. The results of the experiments show that an accuracy rate of 98.81% and precision rate of 98.89% for detecting insulator defects in power transmission lines, which is higher than the existing compared methods and the cost and time of repairs will reduce.