Multi-object Detection by Using CNN for Power Transmission Line Inspection

Multi-object detection for power transmission line is one of the key tasks to control and monitor quality of the system. In the past, defective objects were found out by naked-eye inspection through aerial images and relied on the experienced workers. Due to harsh environmental conditions, manual observation might be a time-consuming and dangerous task. Recently, this task has been supported by machine learning where deep-learning algorithms are applied to increase the efficiency of detection/recognition phases. This paper discusses different approaches for multi-object detection based on Convolutional Neural Network (CNN) model to investigate the quality and condition of power lines in Vietnam. Our proposed system outperforms the state-of-the-art methods on our dataset.