Multi-sensor System Calibration Approach Based on Forward-Model and Inverse-Model

详细信息    查看全文

• 关键词：Multi ; sensor system ; Cross ; sensitivity ; Calibration approach ; Neural network ; Additional samples
• 刊名：Lecture Notes in Computer Science
• 出版年：2017
• 出版时间：2017
• 年：2017
• 卷：10135
• 期：1
• 页码：248-255
• 丛书名：Smart Computing and Communication
• ISBN：978-3-319-52015-5
• 卷排序：10135

文摘

Aimed at the calibration problem of multi-sensor system because of cross-sensitivity, a novel approach based on forward-model and inverse-model was proposed. For forward-model, measurants were taken as inputs, sensor outputs were taken as outputs. For inverse-model, its inputs and outputs were just the opposite. The forward-model was built with calibration samples by interpolation at first. And with this model, “additional samples” were produced. Then, inverse model was built with calibration samples and “additional samples” by training neural networks for every component of analyte. Finally, spectral analysis of light alkane gas mixture was taken as an example to test the calibration approach. The results showed that almost same accuracy, when the approach was applied, could be obtained with only 200 sets of samples as that obtained with 7000 sets of samples, which meaned the proposed approach could reduce the number of calibration sample. And calibration cost could be reduced.