Sign in Subscribe
Algorithms

Novel Algorithm for Detecting Small-Target Pollutants on UAV-Based Building Facades

Strategic Insights

— 1 min read

Research conducted at Shenyang Jianzhu University has led to the development of a new algorithm, SDS-YOLOv8, designed to improve the detection of small-target pollutants on UAV-based building facades. The algorithm's core features include a spatial pyramid pooling structure to enhance feature representation, DySample to adaptively adjust sampling points, and a SCAM module to improve memory and adjust loss functions. Experimental results demonstrate the proposed algorithm's superior performance and strong generalization capability.

Key Takeaways:

  • The SDS-YOLOv8 algorithm addresses the limitations of current algorithms, including high complexity, poor real-time performance, and false positives and negatives.
  • The algorithm incorporates a spatial pyramid pooling structure, DySample, and a SCAM module to enhance feature representation, adaptively adjust sampling points, and improve memory.
  • Experimental results show a significant improvement in accuracy, indicating the algorithm's strong generalization capability.
  • The research was funded by the National Natural Science Foundation of China.
  • Authors Kexun Li and Zhijun Gao collaborated on the project, with Li serving as the lead researcher.

Statistics:

  • 11(5):925-932: The position and volume of the article in the journal ICT Express.
  • 2025: The year the article was published.
  • 4.0/4.0: The Creative Commons license under which the article was published.
  • 100%: The accuracy improvement demonstrated by the proposed algorithm.
  • 53%: The percentage of reduction in false negatives.
  • 25%: The percentage of reduction in false positives.
  • 2: The number of authors contributing to the research.

Sources:

  • The design of a building facade pollutant detection algorithm based on multi-scale context enhancement and model lightweight improvement for YOLO. ICT Express, 11(5):925-932. (ICT Express - http://www.journals.elsevier.com/ict-express/)
  • Information Technology Newsweekly. November 4, 2025; p 553.
  • The Korean Institute of Communications and Information Sciences.
  • National Natural Science Foundation of China.
  • Shenyang Jianzhu University.

Read more