一种红外气体成像的增强去噪模型的训练方法、增强去噪方法
By constructing an end-to-end infrared gas imaging processing model and training it with strong and weak state masks and synthetic noise, the problem of separation between enhancement and denoising in infrared gas imaging is solved, achieving stable enhancement and denoising, and improving detection accuracy and image quality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NANJING ZHIPU TECH CO LTD
- Filing Date
- 2026-04-07
- Publication Date
- 2026-07-17
AI Technical Summary
Existing technologies cannot effectively perform simultaneous enhancement and noise reduction in infrared gas imaging, and it is difficult to differentiate the enhancement of gas regions and backgrounds, resulting in unstable images and low detection accuracy.
An end-to-end infrared gas imaging processing model is constructed. By generating strong and weak state masks and images with consistent morphology, synthetic noise is injected for training. Neural networks are used for synchronous enhancement and denoising. Monotonic constraints and inter-frame alignment modules are introduced to improve spatiotemporal consistency.
It achieves stable enhancement and noise reduction of infrared gas imaging, improves gas detection accuracy and weak gas detection rate, reduces false alarm rate, and ensures the temporal stability and physical reliability of output images.
Smart Images
Figure CN122066606B_ABST
Abstract
Citation Information
Patent Citations
Gas image enhancement and gas leakage detection method and system
CN114596329A
Wind power blade infrared image panoramic splicing method and device, storage medium and product
CN119540045A