Wildfires · Edition 01

Qixing Zhang

Associate Professor, University of Science and Technology of China.

University of Science and Technology of China, China

Field
Fire and smoke detection
Institution
University of Science and Technology of China
Location
Hefei, China
Lab
State Key Laboratory of Fire Science

The research

The public may believe that wildfire detection mainly relies on satellites. Zhang's correction is that ground-based video surveillance networks identify early-stage fires more effectively than satellites do, and that advances in infrastructure are allowing more of these networks to be deployed. He is an associate professor at the State Key Laboratory of Fire Science at the University of Science and Technology of China in Hefei, where his research areas are smoke detection and fire alarm, wildfire detection, and the characterisation of smoke aerosol and its effect on the environment and climate. He completed his PhD at USTC in 2011 and worked as an engineer at China Electronics Technology Group Corporation before returning to USTC in 2015.

His published work with his group covers smoke detection in video and, more recently, systems built for wildfire. A 2019 Fire Technology paper led by Gaohua Lin, with Zhang as senior author, applies 3D convolutional neural networks to smoke detection on video sequences. A 2025 paper in the International Journal of Intelligent Systems, led by Chong Wang with Zhang as senior author, sets out a wildfire smoke detection system, its model architecture, training mechanism and dataset, and releases a real-world wildfire test set from the laboratory, SKLFS-WildFire. SmokeFormer, a 2026 Fire Technology paper led by Chen Xu with Zhang again as senior author, is a decision-ensemble transformer for early wildfire smoke detection in videos. His group now plans a large-scale wildfire-smoke image dataset so that evaluation benchmarks for detection algorithms can be established and more accurate models developed.

In his words

What does the public still get most wrong about wildfire?

The public may believe that wildfire detection today mainly relies on satellites. In reality, ground-based video surveillance networks can identify early-stage fires more effectively than satellites. Advances in infrastructure have also enabled the growing deployment of these technologies.

Which finding in your own work surprised you most — including one that went against what you expected?

The enhancement of artificial intelligence in wildfire smoke-detection capabilities.

What would change in your field if one currently missing dataset existed?

We are planning to construct a large-scale wildfire-smoke image dataset for wildfire detection. With this dataset, evaluation benchmarks for detection algorithms in this field can be established, and more accurate models can be developed.

Whose work outside your own subfield should more wildfire researchers be reading?

Progress in AI deserves attention, among which are applications of world models for wildfire prevention and control.

If the next decade of fire seasons looks like the last three, what should the research community be doing differently?

We hope for closer cooperation, both internationally and between researchers and management agencies, to cope with wildfire-control challenges. In addition, research should focus more on technologies and real-world applications to better support front-line fire-protection staff.

Responses submitted in writing to SusanHub · August 2026

Selected publications

Links

Back to Wildfires · Edition 01