This paper explores the integration of ensemble learning approaches for automated defect detection in industrial and infrastructure applications. By combining multiple machine learning models and computer vision techniques, we demonstrate how ensemble methods can overcome the limitations of individual models and achieve superior performance in detecting surface defects. Our approach leverages the complementary strengths of Convolutional Neural Networks (CNNs) and Vision Transformers (ViTs) enhanced with various image processing techniques. Experimental results across multiple datasets, including steel surface defects and concrete cracks, show that ensemble models consistently outperform individual models in terms of accuracy, precision, and recall. The proposed methodology demonstrates significant potential for real-world applications in quality control and structural health monitoring, contributing to reduced waste, enhanced safety, and improved efficiency in industrial and infrastructure maintenance processes. Our ensemble method achieved 99.5% accuracy while eliminating the seven forms of waste inherent in manual inspection processes: defects, overproduction, waiting, non-utilized talent, transportation, inventory, and excess processing. The framework transforms infrastructure health monitoring from a labor-intensive, error-prone process into a lean, automated system that optimizes maintenance resources and provides consistent, reliable results.