A water quality detection index prediction method based on contrast learning
By using comparative learning methods to select high-quality wavelengths and generate enhanced samples, a predictive model for water quality testing indicators is constructed. This solves the problems of insufficient data representativeness and inaccurate mapping in traditional water quality testing, and improves prediction accuracy and efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NANYANG XIHU MUYUAN SYNTHETIC BIOLOGY RESEARCH INSTITUTE
- Filing Date
- 2026-05-25
- Publication Date
- 2026-07-17
AI Technical Summary
Existing water quality testing methods rely on building models using a single wavelength, resulting in insufficient data representativeness and low prediction accuracy; the extraction cycle for spectral measurement data is long, and high-quality data is lacking, making it difficult to meet the training requirements of deep learning; traditional linear methods are difficult to achieve accurate mapping from absorbance vector to index vector.
A contrastive learning-based approach was adopted to select high-quality detection wavelengths through principal component analysis, generate enhanced samples, and train a feature extraction network using a Transformer model and InfoNCE loss function to construct a water quality detection index prediction model.
It improves the prediction accuracy and efficiency of water quality testing, reduces labor and time costs, and achieves accurate mapping from absorbance vector to water quality test index vector.
Smart Images
Figure CN122409548A_ABST