Improvement of the Control and Surveillance Process for Automotive Diagnostic Centers through the Development of a Computer Vision and Deep Learning-based Software

This paper presents the development and implementation of the “Manhattan” system, a proprietary software solution designed to enhance control and surveillance processes in Automotive Diagnostic Centers (CDAs) in Colombia carried out by Compaa Internacional de Integracin S.A. (CI2). The primary objective of this project was to address critical inefficiencies and fraud risks in vehicle inspections by leveraging Industry 4.0 technologies, including Computer Vision, Deep Learning, and Artificial Intelligence. The system integrates modular applications such as Oslo (for real-time license plate recognition using YOLO-Darknet with 98.8% accuracy), Platino (for video consolidation), and Gambito (for secure data transmission), ensuring seamless coordination with regulatory entities like the Superintendence of Ports and Transport. Validated across 2 CDAs that involbe 2.094 characters, and later deployed in 419 centers (covering 60% of the national market), the system demonstrated a 70% reduction in operational costs, a 40% improvement in video processing speed, and a 45% increase in data transmission capacity. These results highlight its scalability and effectiveness in automating inspections, reducing fraud, and optimizing resource utilization. The project underscores the transformative potential of AI-driven solutions in regulatory compliance and industrial automation.