Research And Application On Model Based Faint Feature Extraction In Digital Image Processing

In the metallurgical industry, adopting traditional detection methods to inspect some unpredictable process parameters suffered from severe noise and external disturbances will cause the results have too big deviation or make the detection failed. Compared with above, visual inspection has many advantages, such as non-contact measurement, real-time detection, large quantity, small amount of system maintenance etc., which have good applicability for inspection of metallurgical industry parameters. But when characteristics of the object are too weak, only using of image information can not completely guarantee the reliability of faint characteristic information extraction.To solve the problem above, this paper puts forward a method of faint feature extraction in digital image processing based on mechanism model. It can realize object's characteristics extract in complicated environment. Mechanism model can be used to analyze and estimate the state of the object, and finally get criterion for extracting faint characteristic information of the target. Combining the criterion with the results of visual inspection, fake information will be inhibited and faint characteristic information will be enhanced.The main researches in this paper are as follows:(1) Characteristics of detected target extraction by using the technology of visual inspection.Some metallurgy industry procedure's parameters are not suitable or even can't be measured directly because of the high temperature, danger or space restrictions in industrial field. To solve this question, this paper adopts threshold de-noising algorithm based on wavelet transform to filter the noise and uses peak value to replace the information in image to inhibit the fake information interference.(2) Getting the criterion based on mechanism model.Considering the results of a large number of image experiments, when the target's characteristics are too weak, only using visual inspection can not completely guarantee the reliability of characteristic extraction. By analyzing and estimating the state of the object from the mechanism model of the object, the criterion which will bring accuracy improvement is obtained and used to eliminate the fake objects formed by the noise and interferences.(3) Faint feature extraction in visual inspection based on model.This method is been used in the interface location between the slag layer and the molten steel layer in tundish steel level measurement. It can overcome the interferences of both noises and slag, and the measuring error is less than 5mm. In a word, visual inspection based on mechanism model can get reliable measuring result in harsh environment of metallurgical industry by improving information content and noise immunity. Proved by some field applications, the method is reliable and will have a bright application prospects in metallurgical process parameter measurement.