Recognition of Damage Types on the Great Wall Surface Based on Machine Learning: Taking the Shanhaiguan Great Wall as an Example(Training set for machine learning)
The Shanhaiguan Great Wall is a part of the Ming dynasty Great Wall, which is a world heritage site. Its basic structure is filled with rammed earth and gray bricks on both sides. Due to environmental influences, gray bricks on the surface will catch some damages, resulting in a decline in the quality of their structure and even threatening their safety. Traditional surface damage detection is mainly based on manual identification or manual identification after UAV aerial photography, and this will be costly in human resources. This paper uses the YOLOv4 machine learning model, taking the surface gray brick of the plain Great Wall of Shanhaiguan as an example. By slicing and labeling the photos, creating a training set, and then training the model, it automatically finds four types of damage (chalking, plant, ubiquinol, and cracking) on the surface of the Great Wall, which will solve the problem of costly human resources for manual identification after aerial photography, allowing the work to progress faster.