一种基于构建卷积核的卷积神经网络水质参数测量方法
By constructing spectral convolution kernels for the target and interfering substances, and combining a neural network model with singular value decomposition and the sigmoid function, the modeling difficulties of convolutional neural networks in water quality parameter measurement were solved, achieving rapid modeling and improving anti-interference capabilities.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- OCEANOGRAPHIC INSTR RES INST SHANDONG ACAD OF SCI
- Filing Date
- 2023-12-21
- Publication Date
- 2026-07-17
AI Technical Summary
Existing convolutional neural networks require a large amount of data for training in water quality parameter measurement and are difficult to model effectively. Furthermore, traditional methods are sensitive to interfering substances, leading to difficulties in modeling and training.
By constructing spectral convolution kernels for the target and interfering substances, extracting key convolution kernels using singular value decomposition, and combining them with a neural network model using the Sigmoid function, convolution operations and predictions are performed on the spectral data.
It enables rapid modeling and calibration, improves the anti-interference capability and environmental adaptability of spectroscopic water quality analysis instruments, and reduces the sample requirement and training difficulty.
Smart Images
Figure CN117894383B_ABST