Structural Health Monitoring (SHM) is accelerating the calling process for innovative methodologies to address its increasing demands efficiently. Traditional SHM approaches often involve time-intensive and costly processes, highlighting the need for automated, deep learning-based solutions. In this research, we created an automated system to detect and classify structural damage, focusing on three specific types of structural integrity concerns: cracks, dampness, and corrosion. Our approach utilizes image classification to distinguish between intact and compromised structures, with an innovative focus on estimating the likelihood of each damage type within a single image. This probability-based assessment provides insight into which type of damages, cracks, dampness, or corrosion is most pronounced, offering a prioritized understanding of structural health. This research is based on a self-created dataset of 815 mobile camera images having three different types of structural damage with seven different combinations. We applied both multi-class and multi-label classification to categorize structural integrity based on the type of damage. For multi-class classification, we tested three methods: Random Forest, Multilayer Perceptron, and CNN. In the multi-label classification context, we employed ResNet, Xception, and Inception to effectively identify and differentiate between overlapping damage types within each image. This research aims to enhance automated damage detection in SHM, delivering a comprehensive and accessible solution for real-time structural assessment through image-based classification.