Research Of Consistency Determination System For High Polymer Materials Based On Spectrogram Recognition

High polymer materials have been widely used in the manufacture of many products because of their abundant sources and multiple properties. The consistency monitoring of high polymer materials has been an important part of products quality supervision. Experts from quality supervision department have been determining the consistency between approval and in-use high polymer materials through manually comparing three spectrograms of IR, DSC and TGA which results in huge workloads and costs. Aiming for an automatic consistency determination system of high polymer materials, the researches in this paper are:1. Algorithm for curve recognition is proposed, this algorithm contains two parts:(1) Based on pixels projection, the positions of the coordinate axes and origin are confirmed. Then, the numbers among the coordinate origin are recognized based on feature number recognition introduced in this paper. Depending on these numbers, the scales of coordinate axes which are conversion relations between measurand and pixel are calculated. (2) Utilizing color feature, the pixels that curve occupies are extracted. According to the scales, the physical values of these pixels are calculated.2. Algorithm for feature number recognition is proposed, this algorithm contains two parts:(1) According to the condition that disturbances in IR are relatively few, feature numbers in IR are oriented and cut out based on pixels projection. As to DSC and TGA, the unit symbols of feature numbers are first oriented based on connected domain labeling, and then feature numbers are oriented and cut out according to their locations relative to the unit symbols. (2) To identify a feature number, the number is first split into many characters, then the characters are identified based on uniform grid feature extraction and template matching, and at last the characters are recombined to identify the feature number.3. Algorithms for IR, DSC and TGA consistency determination are proposed respectively based on curve data and feature numbers, which apply high-low thresholds. Tests of 1318 group spectrograms indicate that the probability of incorrect determinations given by these algorithms is rather low.4. Algorithm for sample consistency determination is proposed based on IR, DSC and TGA consistency determination results. This algorithm could reach certain determination results for more than 91% samples under which circumstance the accuracy rate are more than 99%.5. An application system for automatic consistency determination of high polymer materials is developed based on the above algorithms. The system employs B/S framework and provides a series of functions such as "User Management", "Default parameter setting" and "Consistency Determination". The performance of the system has been excellent during operation.