Float glass melting furnace flame identification method based on linear group and generalized characteristic optimization

Abstract

The invention relates to a float glass melting furnace flame identification method based on linear group and generalized characteristic optimization, and belongs to an advanced manufacturing field, an automation field, and an image processing technology field. By aiming at characteristics of flame images of glass melting furnace severe production environment (such as dust, oil contamination, and high temperature) such as large noises, pixel loss, and low definition, the float glass melting furnace flame identification method is provided. Linear group conversion of incomplete glass melting furnace flame information is carried out, and the incomplete glass information after the linear group conversion is recovered by using a matrix completing method, and therefore the complete flame image is acquired; and then the completed flame image information is mapped in the generalized characteristic space, and the flame state is identified in the generalized characteristic space according to a nearest-neighbor rule. The float glass melting furnace flame identification method is effectively used to identify the glass melting furnace flame, and then the glass melting furnace combustion condition is optimized.

Description

Technical field [0001] The invention belongs to the technical field of advanced manufacturing, automation and image processing, and specifically relates to a flame monitoring image recognition method for a float glass furnace in a complex and non-bad production environment. Background technique [0002] At present, there are several types of image processing methods for noisy image recognition problems: methods based on partial differential equations, methods based on variation, and methods based on pixel blocks; methods based on partial differential equations and methods based on variation are only applicable to a small range In general, the range of incomplete flame information in the float glass furnace is relatively large, so this type of method cannot adapt to this problem; the pixel block-based method needs to iteratively match the damaged area of ​​the image with the known area. In order to obtain the best matching block, filling is performed. Each filling requires travers...

AI Technical Summary

Problems solved by technology

[0002] At present, for the problem of image recognition with noise, there are the following types of image processing methods: methods based on partial differential equations, methods based on variation, methods based on pixel blocks; methods based on partial differential equations and methods based on variation are only applicable to a small range However, the range of incomplete flame information in the float glass melting furnace is generally relatively large, so this type of method cannot be adapted to this problem; the method based on pixel blocks needs to iteratively match the similarity between the damaged area of ​​the image and the known area degree to get the best matching block for filling, and each filling needs to traverse the entire image, resulting in high computational complexity, which cannot meet the requirements of float glass enterprises for glass flame identification

Images