Abstract: With the expansion of transmission lines, traditional methods for detecting insulator faults have become inadequate due to their complexity and inefficiency, failing to meet the modern power grid's demand for efficient and accurate detection. This paper proposes a method for detecting insulator faults in transmission lines based on visible light optical sensing and a multi-scale detection unit network. High-resolution images of insulators are captured within the visible light spectrum, and a convolutional neural network-based multi-scale detection unit network is employed for automated analysis and fault identification of the insulator images. This method effectively identifies common insulator faults in complex backgrounds. Experimental results demonstrate that the proposed method maintains a high true positive rate under most conditions of false positive rates, achieving an AUC value of 0.94. Therefore, this method exhibits high detection accuracy and reliability, showcasing its broad application prospects in intelligent insulator detection for power grids.