Data De-noising Based on PCA-KNN Algorithm in Billet Surface Temperature Measurement

This paper first analyzes the soft-sensor technology and temperature measurement technology used in recent soft measurements, focuses on the accuracy and efficiency requirements of soft measurement in slab surface temperature detection. In this paper, we collects images noise processes of continuous casting slabs, and uses component-based OTSU segmentation method to extract the slab area and then implements Hough transform for edge correction; then measures the selected regions of interest in pixels and extracts color features using PCA for feature reduction; we extracts the improved data set with KNN algorithm for noise reduction, and the removal of contradictory data; the final regression models are used in prediction.