Drone AI Crack Detection from DLSU Reaches NAST Finals as Deep‑Tech Interest Rises

MANILA, Philippines — A De La Salle University team has pushed a drone-based AI crack detection prototype into the 2026 NAST Talent Search for Young Scientists after the system posted accuracy scores above 90 percent. The research, led by assistant professor Timothy Scott C. Chu, is one of the few Philippine-developed tools aimed at automated structural assessment, an area that is gaining attention as the government pushes over PHP 1 trillion a year into infrastructure.

A prototype built for earthquake and typhoon damage

Chu’s work combines convolutional neural networks with camera-equipped unmanned aerial vehicles. The system uses an AlexNet model to detect cracks and a YOLOv4 model to measure them. During his April 22 presentation at Admiral Hotel Manila, Chu said the models still run on a proof-of-concept setup and have not yet been integrated into larger UAV platforms. For now, testing has been limited to controlled environments.

That limitation is common among local deep‑tech projects. While universities have produced UAV experiments for air quality mapping and renewable energy surveys, few projects advance beyond early prototypes. Chu’s research is one of the first Philippine attempts to apply AI directly to infrastructure inspection.

Why the system matters for infrastructure and disaster planning

The country faces more than 20 typhoons each year and sits on multiple fault lines. Government inspections remain mostly manual, and even large bridge audits rely on personnel rappelling or using rented heavy equipment. A drone system capable of detecting cracks of a few millimeters at heights beyond 20 meters could cut inspection times and reduce worker risk.

Other countries treat this as a mature industry. Firms in the United States, Japan, and Israel already sell integrated platforms for utility plants and bridge monitoring. Southeast Asia has also moved faster; Indonesia and Vietnam have produced drone companies serving agriculture and logistics at commercial scale. The Philippines is still largely in research mode.

Barriers that slow down local deep‑tech development

Chu noted that the prototype cannot be fielded yet. Hardware integration, UAV calibration, and regulatory compliance remain unresolved. The Civil Aviation Authority of the Philippines limits many types of flights, especially those beyond visual line of sight. Any startup attempting to commercialize a similar system would need capital for hardware fabrication, testing, and certification.

Local venture data shows that more than 80 percent of Philippine tech investment still goes to software. Hardware and robotics receive a fraction of that, often through small university grants. Neither DOST, DICT, nor DTI has publicly recorded support for Chu’s project. Without stronger channels between academic research and procurement, prototypes tend to stall.

What this means for founders and investors

For founders watching the space, the market gap is obvious. Government engineers are responsible for thousands of bridges, buildings, and public facilities. The Department of Public Works and Highways alone manages over 15,000 national bridges. Inspection cycles are slow, and safety audits are costly. A startup offering drone-based surveying or AI-driven condition monitoring could find real demand.

The harder part is financing. UAV-based deep tech has long development timelines, and investors who prefer fast-scaling software may hesitate. Field testing can take months, and hardware iterations often cost more than PHP 1 million each. Without dedicated grant funding or public-sector pilots, early teams face steep risks.

A signal for the country’s research pipeline

DLSU has been expanding its UAV and computer vision work for at least five years, including its IAH-NET sensors for heat and air quality tracking. Chu’s recognition in the NAST finals suggests growing interest in these domains even if the path to commercialization remains uncertain. The Philippines continues to lag behind its neighbors in robotics and drone manufacturing, but academic work is slowly building a foundation.

What comes next for the project

There have been no recent grants, partnerships, or pilot runs tied to the system. The next steps are mostly technical: integrating the models into larger drones, running field trials with public agencies, and securing CAAP approval. Whether it becomes a commercial product depends on funding and whether government agencies are willing to test AI-driven inspection tools.

The project sits at an early but telling point. Philippine deep‑tech activity is still thin, yet the problems the country faces — aging infrastructure, heavy climate exposure, and limited engineering manpower — are pushing demand for automated tools. Academic teams like Chu’s are doing the early work. The question is whether investors and policymakers will meet them halfway.